RTG Command Reference

This chapter describes RTG commands with a generic description of parameter options and usage. This section also includes expected operation and output results.

Command line interface (CLI)

RTG is installed as a single executable in any system subdirectory where permissions authorize a particular community of users to run the application. RTG commands are executed through the RTG command-line interface (CLI). Each command has its own set of parameters and options described in this section. The availability of each command may be determined by the RTG license that has been installed. Contact support@realtimegenomics.com to discuss changing the set of commands that are enabled by your license.

Results are organized in results directories defined by command parameters and settings. The command line shell environment should include a set of familiar text post-processing tools, such as grep, awk, or perl. Otherwise, no additional applications such as databases or directory services are required.

RTG command syntax

Usage:

rtg COMMAND [OPTIONS] <REQUIRED>

To run an RTG command at the command prompt (either DOS window or Unix terminal), type the product name followed by the command and all required and optional parameters. For example:

$ rtg format -o human_REF_SDF human_REF.fasta

Typically results are written to output files specified with the -o option. There is no default filename or filename extension added to commands requiring specification of an output directory or format.

Many times, unfiltered output files are very large; the built-in compression option generates block compressed output files with the .gz extension automatically unless the parameter -Z or --no-gzip is issued with the command.

Many command parameters require user-supplied information of various types, as shown in the following:

Type

Description

DIR, FILE

File or directory name(s)

SDF

Sequence data that has been formatted to SDF

INT

Integer value

FLOAT

Floating point decimal value

STRING

A sequence of characters for comments, filenames, or labels

REGION

A genomic region specification (see below)

Genomic region parameters take one of the following forms:

  • sequence_name (e.g.: chr21) corresponds to the entirety of the named sequence.

  • sequence_name:start (e.g.: chr21:100000) corresponds to a single position on the named sequence.

  • sequence_name:start-end (e.g.: chr21:100000-110000) corresponds to a range that extends from the specified start position to the specified end position (inclusive). The positions are 1-based.

  • sequence_name:position+length (e.g.: chr21:100000+10000) corresponds to a range that extends from the specified start position that includes the specified number of nucleotides.

  • sequence_name:position~padding (e.g.: chr21:100000~10000) corresponds to a range that spans the specified position by the specified amount of padding on either side.

To display all parameters and syntax associated with an RTG command, enter the command and type --help. For example: all parameters available for the RTG format command are displayed when rtg format --help is executed, the output of which is shown below.

Usage: rtg format [OPTION]... -o SDF FILE+
                  [OPTION]... -o SDF -I FILE
                  [OPTION]... -o SDF -l FILE -r FILE

Converts the contents of sequence data files (FASTA/FASTQ/SAM/BAM) into the RTG
Sequence Data File (SDF) format.

File Input/Output
  -f, --format=FORMAT            format of input. Allowed values are [fasta,
                                 fastq, sam-se, sam-pe, cg-fastq, cg-sam]
                                 (Default is fasta)
  -I, --input-list-file=FILE     file containing a list of input read files (1
                                 per line)
  -l, --left=FILE                left input file for FASTA/FASTQ paired end
                                 data
  -o, --output=SDF               name of output SDF
  -p, --protein                  input is protein. If this option is not
                                 specified, then the input is assumed to
                                 consist of nucleotides
  -q, --quality-format=FORMAT    format of quality data for fastq files (use
                                 sanger for Illumina 1.8+). Allowed values are
                                 [sanger, solexa, illumina]
  -r, --right=FILE               right input file for FASTA/FASTQ paired end
                                 data
      FILE+                      input sequence files. May be specified 0 or
                                 more times

Filtering
      --duster                   treat lower case residues as unknowns
      --exclude=STRING           exclude input sequences based on their name.
                                 If the input sequence contains the specified
                                 string then that sequence is excluded from the
                                 SDF. May be specified 0 or more times
      --select-read-group=STRING when formatting from SAM/BAM input, only
                                 include reads with this read group ID
      --trim-threshold=INT       trim read ends to maximise base quality above
                                 the given threshold

Utility
      --allow-duplicate-names    disable checking for duplicate sequence names
  -h, --help                     print help on command-line flag usage
      --no-names                 do not include name data in the SDF output
      --no-quality               do not include quality data in the SDF output
      --sam-rg=STRING|FILE       file containing a single valid read group SAM
                                 header line or a string in the form
                                 "@RG\tID:READGROUP1\tSM:BACT_SAMPLE\tPL:ILLUMINA"

Required parameters are indicated in the usage display; optional parameters are listed immediately below the usage information in organized categories.

Use the double-dash when typing the full-word command option, as in --output:

$ rtg format --output human_REF_SDF human_REF.fasta

Commonly used command options provide an abbreviated single-character version of a full command parameter, indicated with only a single dash, (Thus --output is the same as specifying the command option with the abbreviated character -o):

$ rtg format -o human_REF human_REF.fasta

A set of utility commands are provided through the CLI: version, license, and help. Start with these commands to familiarize yourself with the software.

The rtg version command invokes the RTG software and triggers the launch of RTG product commands, options, and utilities:

$ rtg version

It will display the version of the RTG software installed, RAM requirements, and license expiration, for example:

$rtg version
Product: RTG Core 3.5
Core Version: 6236f4e (2014-10-31)
RAM: 40.0GB of 47.0GB RAM can be used by rtg (84%)
License: Expires on 2015-09-30
License location: /home/rtgcustomer/rtg/rtg-license.txt
Contact: support@realtimegenomics.com

Patents / Patents pending:
US: 7,640,256, 13/129,329, 13/681,046, 13/681,215, 13/848,653,
13/925,704, 14/015,295, 13/971,654, 13/971,630, 14/564,810
UK: 1222923.3, 1222921.7, 1304502.6, 1311209.9, 1314888.7, 1314908.3
New Zealand: 626777, 626783, 615491, 614897, 614560
Australia: 2005255348, Singapore: 128254

Citation:
John G. Cleary, Ross Braithwaite, Kurt Gaastra, Brian S. Hilbush, Stuart
Inglis, Sean A. Irvine, Alan Jackson, Richard Littin, Sahar
Nohzadeh-Malakshah, Mehul Rathod, David Ware, Len Trigg, and Francisco
M. De La Vega. "Joint Variant and De Novo Mutation Identification on
Pedigrees from High-Throughput Sequencing Data." Journal of
Computational Biology. June 2014, 21(6): 405-419.
doi:10.1089/cmb.2014.0029.
(c) Real Time Genomics Inc, 2014

To see what commands you are licensed to use, type rtg license:

$rtg license
License: Expires on 2015-03-30
Licensed to: John Doe
License location: /home/rtgcustomer/rtg/rtg-license.txt

    Command name    Licensed?  Release Level

Data formatting:
    format          Licensed   GA
    sdf2fasta       Licensed   GA
    sdf2fastq       Licensed   GA

Utility:
    bgzip           Licensed   GA
    index           Licensed   GA
    extract         Licensed   GA
    sdfstats        Licensed   GA
    sdfsubset       Licensed   GA
    sdfsubseq       Licensed   GA
    mendelian       Licensed   GA
    vcfstats        Licensed   GA
    vcfmerge        Licensed   GA
    vcffilter       Licensed   GA
    vcfannotate     Licensed   GA
    vcfsubset       Licensed   GA
    vcfeval         Licensed   GA
    pedfilter       Licensed   GA
    pedstats        Licensed   GA
    rocplot         Licensed   GA
    version         Licensed   GA
    license         Licensed   GA
    help            Licensed   GA

To display all commands and usage parameters available to use with your license, type rtg help:

$ rtg help
Usage: rtg COMMAND [OPTION]...
       rtg RTG_MEM=16G COMMAND [OPTION]...  (e.g. to set maximum memory use to 16 GB)

Type ``rtg help COMMAND`` for help on a specific command. The
following commands are available:

Data formatting:
      format                 convert a FASTA file to SDF
      cg2sdf                 convert Complete Genomics reads to SDF
      sdf2fasta              convert SDF to FASTA
      sdf2fastq              convert SDF to FASTQ
      sdf2sam                convert SDF to SAM/BAM
Read mapping:
      map                    read mapping
      mapf                   read mapping for filtering purposes
      cgmap                  read mapping for Complete Genomics data
Protein search:
      mapx                   translated protein search
Assembly:
      assemble               assemble reads into long sequences
      addpacbio              add Pacific Biosciences reads to an assembly
Variant detection:
      calibrate              create calibration data from SAM/BAM files
      svprep                 prepare SAM/BAM files for sv analysis
      sv                     find structural variants
      discord                detect structural variant breakends using discordant reads
      coverage               calculate depth of coverage from SAM/BAM files
      snp                    call variants from SAM/BAM files
      family                 call variants for a family following Mendelian inheritance
      somatic                call variants for a tumor/normal pair
      population             call variants for multiple potentially-related individuals
      lineage                call de novo variants in a cell lineage
      avrbuild               AVR model builder
      avrpredict             run AVR on a VCF file
      cnv                    call CNVs from paired SAM/BAM files
Metagenomics:
      species                estimate species frequency in metagenomic samples
      similarity             calculate similarity matrix and nearest neighbor tree
Simulation:
      genomesim              generate simulated genome sequence
      cgsim                  generate simulated reads from a sequence
      readsim                generate simulated reads from a sequence
      readsimeval            evaluate accuracy of mapping simulated reads
      popsim                 generate a VCF containing simulated population variants
      samplesim              generate a VCF containing a genotype simulated from a population
      childsim               generate a VCF containing a genotype simulated as a child of two parents
      denovosim              generate a VCF containing a derived genotype containing de novo variants
      samplereplay           generate the genome corresponding to a sample genotype
      cnvsim                 generate a mutated genome by adding CNVs to a template
Utility:
      bgzip                  compress a file using block gzip
      index                  create a tabix index
      extract                extract data from a tabix indexed file
      sdfstats               print statistics about an SDF
      sdfsplit               split an SDF into multiple parts
      sdfsubset              extract a subset of an SDF into a new SDF
      sdfsubseq              extract a subsequence from an SDF as text
      sam2bam                convert SAM file to BAM file and create index
      sammerge               merge sorted SAM/BAM files
      samstats               print statistics about a SAM/BAM file
      samrename              rename read id to read name in SAM/BAM files
      mapxrename             rename read id to read name in mapx output files
      mendelian              check a multi-sample VCF for Mendelian consistency
      vcfstats               print statistics from about variants contained within a VCF file
      vcfmerge               merge single-sample VCF files into a single multi-sample VCF
      vcffilter              filter records within a VCF file
      vcfannotate            annotate variants within a VCF file
      vcfsubset              create a VCF file containing a subset of the original columns
      vcfeval                evaluate called variants for agreement with a baseline variant set
      pedfilter              filter and convert a pedigree file
      pedstats               print information about a pedigree file
      avrstats               print statistics about an AVR model
      rocplot                plot ROC curves from vcfeval ROC data files
      usageserver            run a local server for collecting RTG command usage information
      version                print version and license information
      license                print license information for all commands
      help                   print this screen or help for specified command

The help command will only list the commands for which you have a license to use.

To display help and syntax information for a specific command from the command line, type the command and then the –help option, as in:

$ rtg format --help

Note

The following commands are synonymous: rtg help format and rtg format --help

See also

Refer to Installation and deployment for information about installing the RTG product executable.

Data Formatting Commands

format

Synopsis:

The format command converts the contents of sequence data files (FASTA/FASTQ/SAM/BAM) into the RTG Sequence Data File (SDF) format. This step ensures efficient processing of very large data sets, by organizing the data into multiple binary files within a named directory. The same SDF format is used for storing sequence data, whether it be genomic reference, sequencing reads, protein sequences, etc.

Syntax:

Format one or more files specified from command line into a single SDF:

$ rtg format [OPTION] -o SDF FILE+

Format one or more files specified in a text file into a single SDF:

$ rtg format [OPTION] -o SDF -I FILE

Format mate pair reads into a single SDF:

$ rtg format [OPTION] -o SDF -l FILE -r FILE

Examples:

For FASTA (.fa) genome reference data:

$ rtg format -o maize_reference maize_chr*.fa

For FASTQ (.fq) sequence read data:

$ rtg format -f fastq -q sanger -o h1_reads -l h1_sample_left.fq -r h1_sample_right.fq

Parameters:

File Input/Output

-f

--format=FORMAT

The format of the input file(s). Allowed values are [fasta, fastq, fastq-interleaved, sam-se, sam-pe] (Default is fasta).

-I

--input-list-file=FILE

Specifies a file containing a list of sequence data files (one per line) to be converted into an SDF.

-l

--left=FILE

The left input file for FASTA/FASTQ paired end data.

-o

--output=SDF

The name of the output SDF.

-p

--protein

Set if the input consists of protein. If this option is not specified, then the input is assumed to consist of nucleotides.

-q

--quality-format=FORMAT

The format of the quality data for fastq format files. (Use sanger for Illumina1.8+). Allowed values are [sanger, solexa, illumina].

-r

--right=FILE

The right input file for FASTA/FASTQ paired end data.

FILE+

Specifies a sequence data file to be converted into an SDF. May be specified 0 or more times.

Filtering

--duster

Treat lower case residues as unknowns.

--exclude=STRING

Exclude individual input sequences based on their name. If the input sequence name contains the specified string then that sequence is excluded from the SDF. May be specified 0 or more times.

--select-read-group=STRING

Set to only include only reads with this read group ID when formatting from SAM/BAM files.

--trim-threshold=INT

Set to trim the read ends to maximise the base quality above the given threshold.

Utility

--allow-duplicate-names

Set to disable duplicate name detection.

-h

--help

Prints help on command-line flag usage.

--no-names

Do not include sequence names in the resulting SDF.

--no-quality

Do not include sequence quality data in the resulting SDF.

--sam-rg=STRING|FILE

Specifies a file containing a single valid read group SAM header line or a string in the form @RG\tID:RG1\tSM:G1_SAMP\tPL:ILLUMINA.

Usage:

Formatting takes one or more input data files and creates a single SDF. Specify the type of file to be converted, or allow default to FASTA format. To aggregate multiple input data files, such as when formatting a reference genome consisting of multiple chromosomes, list all files on the command line or use the --input-list-file flag to specify a file containing the list of files to process.

For input FASTA and FASTQ files which are compressed, they must have a filename extension of .gz (for gzip compressed data) or .bz2 (for bzip2 compressed data).

When formatting human reference genome data, it is recommended that the resulting SDF be augmented with chromosome reference metadata, in order to enable automatic sex-aware features during mapping and variant calling. The format command will automatically recognize several common human reference genomes and install a reference configuration file. If your reference genome is not recognized, a configuration can be manually adapted from one of the examples provided in the RTG distribution and installed in the SDF directory. The reference configuration is described in RTG reference file format.

When using FASTQ input files you must specify the quality format being used as one of sanger, solexa or illumina. As of Illumina pipeline version 1.8 and higher, quality values are encoded in Sanger format and so should be formatted using --quality-format=sanger. Output from earlier Illumina pipeline versions should be formatted using --quality-format=illumina for Illumina pipeline versions starting with 1.3 and before 1.8, or --quality-format=solexa for Illumina pipeline versions less than 1.3.

For FASTQ files that represent paired-end read data, indicate each side respectively using the --left=FILE and --right=FILE flags. Sometimes paired-end reads are represented in a single FASTQ file by interleaving each side of the read. This type of input can be formatted by specifying fastq-interleaved as the format type.

The mapx command maps translated DNA sequence data against a protein reference. You must use the -p, --protein flag to format the protein reference used by mapx.

Use the sam-se format for single end SAM/BAM input files and the sam-pe format for paired end SAM/BAM input files. Note that if the input SAM/BAM files are sorted in coordinate order (for example if they have already been aligned to a reference), it is recommended that they be shuffled before formatting, so that subsequent mapping is not biased by processing reads in chromosome order. For example, a BAM file can be shuffled using samtools collate as follows:

$ samtools collate -uOn 256 reads.bam tmp-prefix >reads_shuffled.bam

And this can be carried out on the fly during formatting using bash process redirection in order to reduce intermediate I/O, for example:

$ rtg format --format sam-pe <(samtools collate -uOn 256 reads.bam temp-prefix) ...

The SDF for a read set can contain a SAM read group which will be automatically picked up from the input SAM/BAM files if they contain only one read group. If the input SAM/BAM files contain multiple read groups you must select a single read group from the SAM/BAM file to format using the --select-read-group flag or specify a custom read group with the --sam-rg flag. The --sam-rg flag can also be used to add read group information to reads given in other input formats. The SAM read group stored in an SDF will be automatically used during mapping the reads it contains to provide tracking information in the output BAM files.

The --trim-threshold flag can be used to trim poor quality read ends from the input reads by inspecting base qualities from FASTQ input. If and only if the quality of the final base of the read is less than the threshold given, a new read length is found which maximizes the overall quality of the retained bases using the following formula.

\arg \max x\left({\sum_{i=x+1}^l (T - q(i))}\right) \text{ if } q(l) < T

Where l is the original read length, x is the new read length, T is the given threshold quality and q(n) is the quality of the base at the position n of the read.

Note

Sequencing system read files and reference genome files often have the same extension and it may not always be obvious which file is a read set and which is a genome. Before formatting a sequencing system file, open it to see which type of file it is. For example:

$ less pf3.fa

In general, a read file typically begins with an @ or + character; a genome reference file typically begins with the characters chr.

Normally when the input data contains multiple sequences with the same name the format command will fail with an error. The --allow-duplicate-names flag will disable this check conserving memory, but if the input data has multiple sequences with the same name you will not be warned. Having duplicate sequence names can cause problems with other commands, especially for reference data since the output many commands identifies sequences by their names.

sdf2fasta

Synopsis:

Convert SDF data into a FASTA file.

Syntax:

$ rtg sdf2fasta [OPTION]... -i SDF -o FILE

Example:

$ rtg sdf2fasta -i humanSDF -o humanFASTA_return

Parameters:

File Input/Output

-i

--input=SDF

SDF containing sequences.

-o

--output=FILE

Output filename (extension added if not present). Use ‘-‘ to write to standard output.

Filtering

--end-id=INT

Only output sequences with sequence id less than the given number. (Sequence ids start at 0).

--start-id=INT

Only output sequences with sequence id greater than or equal to the given number. (Sequence ids start at 0).

-I

--id-file=FILE

Name of a file containing a list of sequences to extract, one per line.

--names

Interpret any specified sequence as names instead of numeric sequence ids.

--taxons

Interpret any specified sequence as taxon ids instead of numeric sequence ids. This option only applies to a metagenomic reference species SDF.

STRING+

Specify one or more explicit sequences to extract, as sequence id, or sequence name if –names flag is set.

Utility

-h

--help

Prints help on command-line flag usage.

--interleave

Interleave paired data into a single output file. Default is to split to separate output files.

-l

--line-length=INT

Set the maximum number of nucleotides or amino acids to print on a line of FASTA output. Should be nonnegative, with a value of 0 indicating that the line length is not capped. (Default is 0).

-Z

--no-gzip

Set this flag to create the FASTA output file without compression. By default the output file is compressed with blocked gzip.

Usage:

Use the sdf2fasta command to convert SDF data into FASTA format. By default, sdf2fasta creates a separate line of FASTA output for each sequence. These lines will be as long as the sequences themselves. To make them more readable, use the -l, --line-length flag and define a reasonable record length like 75.

By default all sequences will be extracted, but flags may be specified to extract reads within a range, or explicitly specified reads (either by numeric sequence id or by sequence name if --names is set). Additionally, when the input SDF is a metagenomic species reference SDF, the --taxons option, any supplied id is interpreted as a taxon id and all sequences assigned directly to that taxon id will be output. This provides a convenient way to extract all sequence data corresponding to a single (or multiple) species from a metagenomic species reference SDF.

Sequence ids are numbered starting at 0, the --start-id flag is an inclusive lower bound on id and the --end-id flag is an exclusive upper bound. For example if you have an SDF with five sequences (ids: 0, 1, 2, 3, 4) the following command:

$ rtg sdf2fasta --start-id=3 -i mySDF -o output

will extract sequences with id 3 and 4. The command:

$ rtg sdf2fasta --end-id=3 -i mySDF -o output

will extract sequences with id 0, 1, and 2. And the command:

$ rtg sdf2fasta --start-id=2 --end-id=4 -i mySDF -o output

will extract sequences with id 2 and 3.

sdf2fastq

Synopsis:

Convert SDF data into a FASTQ file.

Syntax:

$ rtg sdf2fastq [OPTION]... -i SDF -o FILE

Example:

$ rtg sdf2fastq -i humanSDF -o humanFASTQ_return

Parameters:

File Input/Output

-i

--input=SDF

Specifies the SDF data to be converted.

-o

--output=FILE

Specifies the file name used to write the resulting FASTQ output.

Filtering

--end-id=INT

Only output sequences with sequence id less than the given number. (Sequence ids start at 0).

--start-id=INT

Only output sequences with sequence id greater than or equal to the given number. (Sequence ids start at 0).

-I

--id-file=FILE

Name of a file containing a list of sequences to extract, one per line.

--names

Interpret any specified sequence as names instead of numeric sequence ids.

STRING+

Specify one or more explicit sequences to extract, as sequence id, or sequence name if –names flag is set.

Utility

-h

--help

Prints help on command-line flag usage.

-q

--default-qualty=INT

Set the default quality to use if the SDF does not contain sequence quality data (0-63).

--interleave

Interleave paired data into a single output file. Default is to split to separate output files.

-l

--line-length=INT

Set the maximum number of nucleotides or amino acids to print on a line of FASTQ output. Should be nonnegative, with a value of 0 indicating that the line length is not capped. (Default is 0).

-Z

--no-gzip

Set this flag to create the FASTQ output file without compression. By default the output file is compressed with blocked gzip.

Usage:

Use the sdf2fastq command to convert SDF data into FASTQ format. If no quality data is available in the SDF, use the -q, --default-quality flag to set a quality score for the FASTQ output. The quality encoding used during output is sanger quality encoding. By default, sdf2fastq creates a separate line of FASTQ output for each sequence. As with sdf2fasta, there is an option to use the -l, --line-length flag to restrict the line lengths to improve readability of long sequences.

By default all sequences will be extracted, but flags may be specified to extract reads within a range, or explicitly specified reads (either by numeric sequence id or by sequence name if --names is set).

It may be preferable to extract data to unaligned SAM/BAM format using sdf2sam, as this preserves read-group information stored in the SDF and may also be more convenient when dealing with paired-end data.

The --start-id and --end-id flags behave as in sdf2fasta.

sdf2sam

Synopsis:

Convert SDF read data into unaligned SAM or BAM format file.

Syntax:

$ rtg sdf2sam [OPTION]... -i SDF -o FILE

Example:

$ rtg sdf2sam -i samplereadsSDF -o samplereads.bam

Parameters:

File Input/Output

-i

--input=SDF

Specifies the SDF data to be converted.

-o

--output=FILE

Specifies the file name used to write the resulting SAM/BAM to. The output format is automatically determined based on the filename specified. If ‘-‘ is given, the data is written as uncompressed SAM to standard output.

Filtering

--end-id=INT

Only output sequences with sequence id less than the given number. (Sequence ids start at 0).

--start-id=INT

Only output sequences with sequence id greater than or equal to the given number. (Sequence ids start at 0).

-I

--id-file=FILE

Name of a file containing a list of sequences to extract, one per line.

--names

Interpret any specified sequence as names instead of numeric sequence ids.

STRING+

Specify one or more explicit sequences to extract, as sequence id, or sequence name if –names flag is set.

Utility

-h

--help

Prints help on command-line flag usage.

-Z

--no-gzip

Set this flag when creating SAM format output to disable compression. By default SAM is compressed with blocked gzip, and BAM is always compressed.

Usage:

Use the sdf2sam command to convert SDF data into unaligned SAM/BAM format. By default all sequences will be extracted, but flags may be specified to extract reads within a range, or explicitly specified reads (either by numeric sequence id or by sequence name if --names is set). This command is a useful way to export paired-end data to a single output file while retaining any read group information that may be stored in the SDF.

The output format is either SAM/BAM depending on the specified output file name. e.g. output.sam or output.sam.gz will output as SAM, whereas output.bam will output as BAM. If neither SAM or BAM format is indicated by the file name then BAM will be used and the output file name adjusted accordingly. e.g output will become output.bam. However if standard output is selected (-) then the output will always be in uncompressed SAM format.

The --start-id and --end-if behave as in sdf2fasta.

fastqtrim

Synopsis:

Trim reads in FASTQ files.

Syntax:

$ rtg fastqtrim [OPTION]... -i FILE -o FILE

Example:

Apply hard base removal from the start of the read and quality-based trimming of terminal bases:

$ rtg fastqtrim -s 12 -E 18 -i S12_R1.fastq.gz -o S12_trimmed_R1.fastq.gz

Parameters:

File Input/Output

-i

--input=FILE

Input FASTQ file, Use ‘-‘ to read from standard input.

-o

--output=FILE

Output filename. Use ‘-‘ to write to standard output.

-q

--quality-format=FORMAT

Quality data encoding method used in FASTQ input files (Illumina 1.8+ uses sanger). Allowed values are [sanger, solexa, illumina] (Default is sanger)

Filtering

--discard-empty-reads

Discard reads that have zero length after trimming. Should not be used with paired-end data.

-E

--end-quality-threshold=INT

Trim read ends to maximise base quality above the given threshold (Default is 0)

--min-read-length=INT

If a read ends up shorter than this threshold it will be trimmed to zero length (Default is 0)

-S

--start-quality-threshold=INT

Trim read starts to maximise base quality above the given threshold (Default is 0)

-e

--trim-end-bases=INT

Always trim the specified number of bases from read end (Default is 0)

-s

--trim-start-bases=INT

Always trim the specified number of bases from read start (Default is 0)

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

-r

--reverse-complement

If set, output in reverse complement.

--seed=INT

Seed used during subsampling.

--subsample=FLOAT

If set, subsample the input to retain this fraction of reads.

-T

--threads=INT

Number of threads (Default is the number of available cores)

Usage:

Use fastqtrim to apply custom trimming and preprocessing to raw FASTQ files prior to mapping and alignment. The format command contains some limited trimming options, which are applied to all input files, however in some cases different or specific trimming operations need to be applied to the various input files. For example, for paired-end data, different trimming may need to be applied for the left read files compared to the right read files. In these cases, fastqtrim should be used to process the FASTQ files first.

The --end-quality-threshold flag can be used to trim poor quality bases from the ends of the input reads by inspecting base qualities from FASTQ input. If and only if the quality of the final base of the read is less than the threshold given, a new read length is found which maximizes the overall quality of the retained bases using the following formula:

\arg \max x\left({\sum_{i=x+1}^l (T - q(i))}\right) \text{ if } q(l) < T

where l is the original read length, x is the new read length, T is the given threshold quality and q(n) is the quality of the base at the position n of the read. Similarly, --start-quality-threshold can be used to apply this quality-based thresholding to the start of reads.

Some of the trimming options may result in reads that have no bases remaining. By default, these are output as zero-length FASTQ reads, which RTG commands are able to handle normally. It is also possible to remove zero-length reads altogether from the output with the --discard-empty-reads option, however this should not be used when processing FASTQ files corresponding to paired-end data, otherwise the pairs in the two files will no longer be matched.

Similarly, when using the --subsample option to down-sample a FASTQ file for paired-end data, you should specify an explicit randomization seed via --seed and use the same seed value for the left and right files.

Formatting with filtering on the fly

Running custom filtering with fastqtrim need not mean that additional disk space is required or that formatting be slowed down due to additional disk I/O. It is possible when using standard unix shells to perform the filtering on the fly. The following example demonstrates how to apply different trimming options to left and right files while formatting to SDF:

$ rtg format -f fastq -o S12_trimmed.sdf \
    -l <(rtg fastqtrim -s 12 -E 18 -i S12_R1.fastq.gz -o -)
    -r <(rtg fastqtrim -E 18 -i S12_R2.fastq.gz -o -)

See also

format

petrim

Synopsis:

Trim paired-end read FASTQ files based on read arm alignment overlap.

Syntax:

$ rtg petrim [OPTION]... -l FILE -o FILE -r FILE

Parameters:

File Input/Output

-l

--left=FILE

Left input FASTQ file (AKA R1)

-o

--output=FILE

Output filename prefix. Use ‘-‘ to write to standard output.

-q

--quality-format=FORMAT

Quality data encoding method used in FASTQ input files (Illumina 1.8+ uses sanger). Allowed values are [sanger, solexa, illumina] (Default is sanger)

-r

--right=FILE

Right input FASTQ file (AKA R2)

Sensitivity Tuning

--aligner-band-width=FLOAT

Aligner indel band width scaling factor, fraction of read length allowed as an indel (Default is 0.5)

--gap-extend-penalty=INT

Penalty for a gap extension during alignment (Default is 1)

--gap-open-penalty=INT

Penalty for a gap open during alignment (Default is 19)

-P

--min-identity=INT

Minimum percent identity in overlap to trigger overlap trimming (Default is 90)

-L

--min-overlap-length=INT

Minimum number of bases in overlap to trigger overlap trimming (Default is 25)

--mismatch-penalty=INT

Penalty for a mismatch during alignment (Default is 9)

--soft-clip-distance=INT

Soft clip alignments if indels occur INT bp from either end (Default is 5)

--unknowns-penalty=INT

Penalty for unknown nucleotides during alignment (Default is 5)

Filtering

--discard-empty-pairs

If set, discard pairs where both reads have zero length (after any trimming)

--discard-empty-reads

If set, discard pairs where either read has zero length (after any trimming)

--left-probe-length=INT

Assume R1 starts with probes this long, and trim R2 bases that overlap into this (Default is 0)

-M

--midpoint-merge

If set, merge overlapping reads at midpoint of overlap region. Result is in R1 (R2 will be empty)

-m

--midpoint-trim

If set, trim overlapping reads to midpoint of overlap region.

--min-read-length=INT

If a read ends up shorter than this threshold it will be trimmed to zero length (Default is 0)

--mismatch-adjustment=STRING

Method used to alter bases/qualities at mismatches within overlap region. Allowed values are [none, zero-phred, pick-best] (Default is none)

--right-probe-length=INT

Assume R2 starts with probes this long, and trim R1 bases that overlap into this (Default is 0)

Utility

-h

--help

Print help on command-line flag usage.

--interleave

Interleave paired data into a single output file. Default is to split to separate output files.

-Z

--no-gzip

Do not gzip the output.

--seed=INT

Seed used during subsampling.

--subsample=FLOAT

If set, subsample the input to retain this fraction of reads.

-T

--threads=INT

Number of threads (Default is the number of available cores)

Usage:

Paired-end read sequencing with read lengths that are long relative to the typical library fragment size can often result in the same bases being sequenced by both arms. This repeated sequencing of bases within the same fragment can skew variant calling, and so it can be advantageous to remove such read overlap.

In some cases, complete read-through can occur, resulting in additional adaptor or non-genomic bases being present at the ends of reads.

In addition, some library preparation methods rely on the ligation of synthetic probe sequence to attract target DNA, which is subsequently sequenced. Since these probe bases do not represent genomic material, they must be removed at some point during the analytic pipeline prior to variant calling, otherwise they could act as a reference bias when calling variants. Removal from the primary arm where the probe is attached is typically easy enough (e.g. via fastqtrim), however in cases of high read overlap, probe sequence can also be present in the other read arm.

petrim aligns each read arm against it’s mate with high stringency in order to identify cases of read overlap. The sensitivity of read overlap detection is primarily controlled through the use of --min-identity and --min-overlap-length, although it is also possible to adjust the penalties used during alignment.

The following types of trimming or merging may be applied.

  • Removal of non-genomic bases due to complete read-through. This removal is always applied.

  • Removal of overlap bases impinging into regions occupied by probe bases. For example, if the left arms contain 11-mer probes, using --left-probe-length=11 will result in the removal of any right arm bases that overlap into the first 11 bases of the left arm. Similar trimming is available for situations where probes are ligated to the right arm by using --right-probe-length.

  • Adjustment of mismatching read bases inside areas of overlap. Such mismatches indicate that one or other of the bases has been incorrectly sequenced. Alteration of these bases is selected by supplying the --mismatch-adjustment flag with a value of zero-phred to alter the phred quality score of both bases to zero, or pick-best to choose whichever base had the higher reported quality score.

  • Removal of overlap regions by trimming both arms back to a point where no overlap is present. An equal number of bases are removed from each arm. This trimming is enabled by specifying --midpoint-trim and takes place after any read-through or probe related trimming.

  • Merging non-redundant sequence from both reads to create a single read, enabled via --midpoint-merge. This is like --midpoint-trim with a subsequent moving of the R2 read onto the end of the the R1 read (thus the R2 read becomes empty).

After trimming or merging it is possible that one or both of the arms of the pair have no bases remaining, and a strategy is needed to handle these pairs. The default is to retain such pairs in the output, even if one or both are zero-length. When both arms are zero-length, the pair can be dropped from output with the use of --discard-empty-pairs. If downstream processing cannot handle zero-length reads, --discard-empty-reads will drop a read pair if either of the arms is zero-length.

petrim also provides the ability to down-sample a read set by using the --subsample option. This will produce a different sampling each time, unless an explicit randomization seed is specified via --seed.

Formatting with paired-end trimming on the fly

Running custom filtering with petrim can be done in standard Unix shells without incurring the use of additional disk space or unduly slowing down the formatting of reads. The following example demonstrates how to apply paired-end trimming while formatting to SDF:

$ rtg format -f fastq-interleaved -o S12_trimmed.sdf \
    <(rtg petrim -l S12_R1.fastq.gz -r S12_R2.fastq.gz -m -o - --interleaved)

This can even be combined with fastqtrim to provide extremely flexible trimming:

$ rtg format -f fastq-interleaved -o S12_trimmed.sdf \
    <(rtg petrim -m -o - --interleave \
         -l <(rtg fastqtrim -s 12 -E 18 -i S12_R1.fastq.gz -o -) \
         -r <(rtg fastqtrim -E 18 -i S12_R2.fastq.gz -o -) \
     )

Note

petrim currently assumes Illumina paired-end sequencing, and aligns the reads in FR orientation. Sequencing methods which produce arms in a different orientation can be processed by first converting the input files using fastqtrim --reverse-complement, running petrim, followed by another fastqtrim --reverse-complement to restore the reads to their original orientation.

See also

fastqtrim, format

Simulation Commands

RTG includes some simulation commands that may be useful for experimenting with effects of various RTG command parameters or when getting familiar with RTG work flows. A simple simulation series might involve the following commands:

$ rtg genomesim --output sim-ref-sdf --min-length 500000 --max-length 5000000 \
  --num-contigs 5
$ rtg popsim --reference sim-ref-sdf --output population.vcf.gz
$ rtg samplesim --input population.vcf.gz --output sample1.vcf.gz \
  --output-sdf sample1-sdf --reference sim-ref-sdf --sample sample1
$ rtg readsim --input sample1-sdf --output reads-sdf --machine illumina_pe \
  -L 75 -R 75 --coverage 10
$ rtg map --template sim-ref-sdf --input reads-sdf --output sim-mapping \
  --sam-rg "@RG\tID:sim-rg\tSM:sample1\tPL:ILLUMINA"
$ rtg snp --template sim-ref-sdf --output sim-name-snp sim-mapping/alignments.bam

genomesim

Synopsis:

Use the genomesim command to simulate a reference genome, or to create a baseline reference genome for a research project when an actual genome reference sequence is unavailable.

Syntax:

Specify number of sequences, plus minimum and maximum lengths:

$ rtg genomesim [OPTION]... -o SDF --max-length INT --min-length INT -n INT

Specify explicit sequence lengths (one more sequences):

$ rtg genomesim [OPTION]... -o SDF -l INT

Example:

$ rtg genomesim -o genomeTest -l 500000

Parameters:

File Input/Output

-o

--output=SDF

The name of the output SDF.

Utility

--comment=STRING

Specify a comment to include in the generated SDF.

--freq=STRING

Set the relative frequencies of A,C,G,T in the generated sequence. (Default is 1,1,1,1).

-h

--help

Prints help on command-line flag usage.

-l

--length=INT

Specify the length of generated sequence. May be specified 0 or more times, or as a comma separated list.

--max-length=INT

Specify the maximum sequence length.

--min-length=INT

Specify the minimum sequence length.

-n

--num-contigs=INT

Specify the number of sequences to generate.

--prefix=STRING

Specify a sequence name prefix to be used for the generated sequences. The default is to name the output sequences ‘simulatedSequenceN’, where N is increasing for each sequence.

-s

--seed=INT

Specify seed for the random number generator.

Usage:

The genomesim command allows one to create a simulated genome with one or more contiguous sequences - exact lengths of each contig or number of contigs with minimum and maximum lengths provided. The contents of an SDF directory created by genomesim can be exported to a FASTA file using the sdf2fasta command.

This command is primarily useful for providing a simple randomly generated base genome for use with subsequent simulation commands.

Each generated contig is named by appending an increasing numeric index to the specified prefix. For example --prefix=chr --num-contigs=10 would yield contigs named chr1 through chr10.

cgsim

Synopsis:

Simulate Complete Genomics Inc sequencing reads. Supports the original 35 bp read structure (5-10-10-10), and the newer 29 bp read structure (10-9-10).

Syntax:

Generation by genomic coverage multiplier:

$ rtg cgsim [OPTION]... -V INT -t SDF -o SDF -c FLOAT

Generation by explicit number of reads:

$ rtg cgsim [OPTION]... -V INT -t SDF -o SDF -n INT

Example:

$ rtg cgsim -V 1 -t HUMAN_reference -o CG_3x_readst -c 3

Parameters:

File Input/Output

-t

--input=SDF

SDF containing input genome.

-o

--output=SDF

Name for reads output SDF.

Fragment Generation

--abundance

If set, the user-supplied distribution represents desired abundance.

-N

--allow-unknowns

Allow reads to be drawn from template fragments containing unknown nucleotides.

-c

--coverage=FLOAT

Coverage, must be positive.

-D

--distribution=FILE

File containing probability distribution for sequence selection.

--dna-fraction

If set, the user-supplied distribution represents desired DNA fraction.

-M

--max-fragment-size=INT

Maximum fragment size (Default is 500)

-m

--min-fragment-size=INT

Minimum fragment size (Default is 350)

--n-rate=FLOAT

Rate that the machine will generate new unknowns in the read (Default is 0.0)

-n

--num-reads=INT

Number of reads to be generated.

--taxonomy-distribution=FILE

File containing probability distribution for sequence selection expressed by taxonomy id.

Complete Genomics

-V

--cg-read-version=INT

Select Complete Genomics read structure version, 1 (35 bp) or 2 (29 bp)

Utility

--comment=STRING

Comment to include in the generated SDF.

-h

--help

Print help on command-line flag usage.

--no-names

Do not create read names in the output SDF.

--no-qualities

Do not create read qualities in the output SDF.

-q

--qual-range=STRING

Set the range of base quality values permitted e.g.: 3-40 (Default is fixed qualities corresponding to overall machine base error rate)

--sam-rg=STRING|FILE

File containing a single valid read group SAM header line or a string in the form @RG\tID:READGROUP1\tSM:BACT_SAMPLE\tPL:ILLUMINA

-s

--seed=INT

Seed for random number generator.

Usage:

Use the cgsim command to set either --coverage or --num-reads in simulated Complete Genomics reads. For more information about Complete Genomics reads, refer to http://www.completegenomics.com

RTG simulation tools allows for deterministic experiment repetition. The --seed parameter, for example, allows for regeneration of exact same reads by setting the random number generator to be repeatable (without supplying this flag a different set of reads will be generated each time).

The --distribution parameter allows you to specify the probability that a read will come from a particular named sequence for use with metagenomic databases. Probabilities are numbers between zero and one and must sum to one in the file.

readsim

Synopsis:

Use the readsim command to generate single or paired end reads of fixed or variable length from a reference genome, introducing machine errors.

Syntax:

Generation by genomic coverage multiplier:

$ rtg readsim [OPTION]... -t SDF --machine STRING -o SDF -c FLOAT

Generation by explicit number of reads:

$ rtg readsim [OPTION]... -t SDF --machine STRING -o SDF -n INT

Example:

$ rtg readsim -t genome_ref -o sim_reads -r 75 --machine illumina_se  -c 30

Parameters:

File Input/Output

-t

--input=SDF

SDF containing input genome.

--machine=STRING

Select the sequencing technology to model. Allowed values are [illumina_se, illumina_pe, complete_genomics, complete_genomics_2, 454_pe, 454_se, iontorrent]

-o

--output=SDF

Name for reads output SDF.

Fragment Generation

--abundance

If set, the user-supplied distribution represents desired abundance.

-N

--allow-unknowns

Allow reads to be drawn from template fragments containing unknown nucleotides.

-c

--coverage=FLOAT

Coverage, must be positive.

-D

--distribution=FILE

File containing probability distribution for sequence selection.

--dna-fraction

If set, the user-supplied distribution represents desired DNA fraction.

-M

--max-fragment-size=INT

Maximum fragment size (Default is 250)

-m

--min-fragment-size=INT

Minimum fragment size (Default is 200)

--n-rate=FLOAT

Rate that the machine will generate new unknowns in the read (Default is 0.0)

-n

--num-reads=INT

Number of reads to be generated.

--taxonomy-distribution=FILE

File containing probability distribution for sequence selection expressed by taxonomy id.

Illumina PE

-L

--left-read-length=INT

Target read length on the left side.

-R

--right-read-length=INT

Target read length on the right side.

Illumina SE

-r

--read-length=INT

Target read length, must be positive.

454 SE/PE

--454-max-total-size=INT

Maximum 454 read length (in paired end case the sum of the left and the right read lengths)

--454-min-total-size=INT

Minimum 454 read length (in paired end case the sum of the left and the right read lengths)

IonTorrent SE

--ion-max-total-size=INT

Maximum IonTorrent read length.

--ion-min-total-size=INT

Minimum IonTorrent read length.

Utility

--comment=STRING

Comment to include in the generated SDF.

-h

--help

Print help on command-line flag usage.

--no-names

Do not create read names in the output SDF.

--no-qualities

Do not create read qualities in the output SDF.

-q

--qual-range=STRING

Set the range of base quality values permitted e.g.: 3-40 (Default is fixed qualities corresponding to overall machine base error rate)

--sam-rg=STRING|FILE

File containing a single valid read group SAM header line or a string in the form @RG\tID:READGROUP1\tSM:BACT_SAMPLE\tPL:ILLUMINA

-s

--seed=INT

Seed for random number generator.

Usage:

Create simulated reads from a reference genome by either specifying coverage depth or a total number of reads.

A typical use case involves creating a mutated genome by introducing SNPs or CNVs with popsim and samplesim generating reads from the mutated genome with readsim, and mapping them back to the original reference to verify the parameters used for mapping or variant detection.

RTG simulation tools allows for deterministic experiment repetition. The --seed parameter, for example, allows for regeneration of exact same reads by setting the random number generator to be repeatable (without supplying this flag a different set of reads will be generated each time).

The --distribution parameter allows you to specify the sequence composition of the resulting read set, primarily for use with metagenomic databases. The distribution file is a text file containing lines of the form:

<probability><space><sequence name>

Probabilities must be between zero and one and must sum to one in the file. For reference databases containing taxonomy information, where each species may be comprised of more than one sequence, it is instead possible to use the --taxonomy-distribution option to specify the probabilities at a per-species level. The format of each line in this case is:

<probability><space><taxon id>

When using --distribution or --taxonomy-distribution, the interpretation must be specified one of --abundance or --dna-fraction. When using --abundance each specified probability reflects the chance of selecting the specified sequence (or taxon id) from the set of sequences, and thus for a given abundance a large sequence will be represented by more reads in the resulting set than a short sequence. In contrast, with --dna-fraction each specified probability reflects the chance of a read being derived from the designated sequence, and thus for a given fraction, a large sequence will have a lower depth of coverage than a short sequence.

popsim

Synopsis:

Use the popsim command to generate a VCF containing simulated population variants. Each variant allele generated has an associated frequency INFO field describing how frequent in the population that allele is.

Syntax:

$ rtg popsim [OPTION]... -o FILE -t SDF

Example:

$ rtg popsim -o pop.vcf -t HUMAN_reference

Parameters:

File Input/Output

-o

--output=FILE

Output VCF file name.

-t

--reference=SDF

SDF containing the reference genome.

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--seed=INT

Seed for the random number generator.

Usage:

The popsim command is used to create a VCF containing variants with frequency in population information that can be subsequently used to simulate individual samples using the samplesim command. The frequency in population is contained in a VCF INFO field called AF. The types of variants and the allele-frequency distribution has been drawn from observed variants and allele frequency distribution in human studies.

samplesim

Synopsis:

Use the samplesim command to generate a VCF containing a genotype simulated from population variants according to allele frequency.

Syntax:

$ rtg samplesim [OPTION]... -i FILE -o FILE -t SDF -s STRING

Example:

From a population frequency VCF:

$ rtg samplesim -i pop.vcf -o 1samples.vcf -t HUMAN_reference -s person1 --sex male

From an existing simulated VCF:

$ rtg samplesim -i 1samples.vcf -o 2samples.vcf -t HUMAN_reference -s person2 \
  --sex female

Parameters:

File Input/Output

-i

--input=FILE

Input VCF containing population variants.

-o

--output=FILE

Output VCF file name.

--output-sdf=SDF

If set, output an SDF containing the sample genome.

-t

--reference=SDF

SDF containing the reference genome.

-s

--sample=STRING

Name for sample.

Utility

--allow-missing-af

If set, treat variants without allele frequency annotation as uniformly likely.

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--ploidy=STRING

Ploidy to use. Allowed values are [auto, diploid, haploid] (Default is auto)

--seed=INT

Seed for the random number generator.

--sex=SEX

Sex of individual. Allowed values are [male, female, either] (Default is either)

Usage:

The samplesim command is used to simulate an individuals genotype information from a population variant frequency VCF generated by the popsim command or by previous samplesim or childsim commands. The new output VCF will contain all the existing variants and samples with a new column for the new sample. The genotype at each record of the VCF will be chosen randomly according to the allele frequency specified in the AF field.

If input VCF records do not contain an AF annotation, by default any ALT allele in that record will not be selected and so the sample will be genotyped as 0/0. Alternatively for simple simulations the --allow-missing-af flag will treat each allele in such records as being equally likely (i.e.: effectively equivalent to AF=0.5 for a biallelic variant, AF=0.33,0.33 for a triallelic variant, etc).

The ploidy for each genotype is automatically determined according to the ploidy of that chromosome for the specified sex of the individual, as defined in the reference genome reference.txt file. For more information see RTG reference file format. If the reference SDF does not contain chromosome configuration information, a default ploidy can be specified using the --ploidy flag.

The --output-sdf flag can be used to optionally generate an SDF of the individuals genotype which can then be used by the readsim command to simulate a read set for the individual.

denovosim

Synopsis:

Use the denovosim command to generate a VCF containing a derived genotype containing de novo variants.

Syntax:

$ rtg denovosim [OPTION]... -i FILE --original STRING -o FILE -t SDF -s STRING

Example:

$ rtg denovosim -i sample.vcf --original personA -o 2samples.vcf \
  -t HUMAN_reference -s personB

Parameters:

File Input/Output

-i

--input=FILE

The input VCF containing parent variants.

--original=STRING

The name of the existing sample to use as the original genotype.

-o

--output=FILE

The output VCF file name.

--output-sdf=FILE

Set to output an SDF of the genome generated.

-t

--reference=SDF

The SDF containing the reference genome.

-s

--sample=STRING

The name for the new derived sample.

Utility

-h

--help

Prints help on command-line flag usage.

-Z

--no-gzip

Set this flag to create the VCF output file without compression.

--num-mutations=INT

Set the expected number of de novo mutations per genome (Default is 70).

--ploidy=STRING

The ploidy to use when the reference genome does not contain a reference text file. Allowed values are [auto, diploid, haploid] (Default is auto)

--seed=INT

Set the seed for the random number generator.

--show-mutations

Set this flag to display information regarding de novo mutation points.

Usage:

The denovosim command is used to simulate a derived genotype containing de novo variants from a VCF containing an existing genotype.

The output VCF will contain all the existing variants and samples, along with additional de novo variants. If the original and derived sample names are different, the output will contain a new column for the mutated sample. If the original and derived sample names are the same, the sample in the output VCF is updated rather than creating an entirely new sample column. When a sample receives a de novo mutation, the sample DN field is set to “Y”.

If de novo variants were introduced without regard to neighboring variants, a situation could arise where it is not possible to unambiguously determine the haplotype of the simulated sample. To prevent this, denovosim will not output a de novo variant that overlaps existing variants. Since denovosim chooses candidate de novo locations before reading the input VCF, this occasionally mandates skipping a candidate de novo so the target number of mutations may not always be reached.

The --output-sdf flag can be used to optionally generate an SDF of the derived genome which can then be used by the readsim command to simulate a read set for the new genome.

childsim

Synopsis:

Use the childsim command to generate a VCF containing a genotype simulated as a child of two parents.

Syntax:

$ rtg childsim [OPTION]... --father STRING -i FILE --mother STRING -o FILE -t SDF \
  -s STRING

Example:

$ rtg childsim --father person1 --mother person2 -i 2samples.vcf -o 3samples.vcf \
  -t HUMAN_reference -s person3

Parameters:

File Input/Output

--father=STRING

Name of the existing sample to use as the father.

-i

--input=FILE

Input VCF containing parent variants.

--mother=STRING

Name of the existing sample to use as the mother.

-o

--output=FILE

Output VCF file name.

--output-sdf=SDF

If set, output an SDF containing the sample genome.

-t

--reference=SDF

SDF containing the reference genome.

-s

--sample=STRING

Name for new child sample.

Utility

--extra-crossovers=FLOAT

Probability of extra crossovers per chromosome (Default is 0.01)

--genetic-map-dir=DIR

If set, load genetic maps from this directory for recombination point selection.

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--ploidy=STRING

Ploidy to use. Allowed values are [auto, diploid, haploid] (Default is auto)

--seed=INT

Seed for the random number generator.

--sex=SEX

Sex of individual. Allowed values are [male, female, either] (Default is either)

--show-crossovers

If set, display information regarding haplotype selection and crossover points.

Usage:

The childsim command is used to simulate an individuals genotype information from a VCF containing the two parent genotypes generated by previous samplesim or childsim commands. The new output VCF will contain all the existing variants and samples with a new column for the new sample.

The ploidy for each genotype is generated according to the ploidy of that chromosome for the specified sex of the individual, as defined in the reference genome reference.txt file. For more information see RTG reference file format. The generated genotypes are all consistent with Mendelian inheritance (de novo variants can be simulated with the denovosim command).

The --output-sdf flag can be used to optionally generate an SDF of the child’s genotype which can then be used by the readsim command to simulate a read set for the child.

By default positions for crossover events are chosen according to a uniform distribution. However, if linkage information is available, then this can be used to inform the crossover selection procedure. The expected format for this information is described in Genetic map directory, and the directory containing the relevant files can be specified by using the --genetic-map-dir flag.

pedsamplesim

Synopsis:

Generates simulated genotypes for all members of a pedigree. pedsamplesim automatically simulates founder individuals, inheritance by children, and de novo mutations.

Syntax:

$ rtg pedsamplesim [OPTION]... -i FILE -o DIR -p FILE -t SDF

Example:

$ rtg pedsamplesim -t reference.sdf -p family.ped -i popvars.vcf \
  -o family_sim --remove-unused

Parameters:

File Input/Output

-i

--input=FILE

Input VCF containing parent variants.

-o

--output=DIR

Directory for output.

--output-sdf

If set, output an SDF for the genome of each simulated sample.

-p

--pedigree=FILE

Genome relationships PED file.

-t

--reference=SDF

SDF containing the reference genome.

Utility

--extra-crossovers=FLOAT

Probability of extra crossovers per chromosome (Default is 0.01)

--genetic-map-dir=DIR

If set, load genetic maps from this directory for recombination point selection.

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--num-mutations=INT

Expected number of mutations per genome (Default is 70)

--ploidy=STRING

Ploidy to use. Allowed values are [auto, diploid, haploid] (Default is auto)

--remove-unused

If set, output only variants used by at least one sample.

--seed=INT

Seed for the random number generator.

Usage:

The pedsamplesim uses the methods of samplesim, denovosim, and childsim to greatly ease the simulation of multiple samples. The input VCF should contain standard allele frequency INFO annotations that will be used to simulate genotypes for any sample identified as a founder. Any samples present in the pedigree that are already present in the input VCF will not be regenerated. To simulate genotypes for a subset of the members of the pedigree, use pedfilter to create a filtered pedigree file that includes only the subset required.

The supplied pedigree file is first examined to identify any individuals that cannot be simulated according to inheritance from other samples in the pedigree. Note that simulation according to inheritance requires both parents to be present in the pedigree. These samples in the pedigree are treated as founder individuals.

Founder individuals are simulated using samplesim, where the genotypes are chosen according to the allele frequency annotation in the input VCF.

All newly generated samples may have de novo mutations introduced according to the --num-mutations setting. As with the denovosim command, any de novo mutations introduced in a sample will be genotyped as homozygous reference in other pre-existing samples, and introduced variants will not overlap any pre-existing variant loci.

Samples that can be simulated according to Mendelian inheritance are then generated, using childsim. As expected, as well as inheriting de novo variants from parents, each child may obtain new de novo mutations of their own.

If the simulated samples will be used for subsequent simulated sequencing, such as via readsim, it is possible to automatically output an SDF containing the simulated genome for each sample by specifying the --output-sdf option, obviating the need to separately use samplereplay.

By default positions for crossover events are chosen according to a uniform distribution. However, if linkage information is available, then this can be used to inform the crossover selection procedure. The expected format for this information is described in Genetic map directory, and the directory containing the relevant files can be specified by using the --genetic-map-dir flag.

samplereplay

Synopsis:

Use the samplereplay command to generate the genome SDF corresponding to a sample genotype in a VCF file.

Syntax:

$ rtg samplereplay [OPTION]... -i FILE -o SDF -t SDF -s STRING

Example:

$ rtg samplereplay -i 3samples.vcf -o child.sdf -t HUMAN_reference -s person3

Parameters:

File Input/Output

-i

--input=FILE

Input VCF containing the sample genotype.

-o

--output=SDF

Name for output SDF.

-t

--reference=SDF

SDF containing the reference genome.

-s

--sample=STRING

Name of the sample to select from the VCF.

Utility

-h

--help

Print help on command-line flag usage.

Usage:

The samplereplay command can be used to generate an SDF of a genotype for a given sample from an existing VCF file. This can be used to generate a genome from the outputs of the samplesim and childsim commands. The output genome can then be used in simulating a read set for the sample using the readsim command.

Every chromosome for which the individual is diploid will have two sequences in the resulting SDF.

Utility Commands

bgzip

Synopsis:

Block compress a file or decompress a block compressed file. Block compressed outputs from the mapping and variant detection commands can be indexed with the index command. They can also be processed with standard gzip tools such as gunzip and zcat.

Syntax:

$ rtg bgzip [OPTION]... FILE+

Example:

$ rtg bgzip alignments.sam

Parameters:

File Input/Output

-l

--compression-level=INT

The compression level to use, between 1 (least but fast) and 9 (highest but slow) (Default is 5)

-d

--decompress

Decompress.

-f

--force

Force overwrite of output file.

--no-terminate

If set, do not add the block gzip termination block.

-c

--stdout

Write on standard output, keep original files unchanged. Implied when using standard input.

FILE+

File to (de)compress, use ‘-‘ for standard input. Must be specified 1 or more times.

Utility

-h

--help

Print help on command-line flag usage.

Usage:

Use the bgzip command to block compress files. Files such as VCF, BED, SAM, TSV must be block-compressed before they can be indexed for fast retrieval of records corresponding to specific genomic regions.

See also

index

index

Synopsis:

Create tabix index files for block compressed TAB-delimited genome position data files or BAM index files for BAM files.

Syntax:

Multi-file input specified from command line:

$ rtg index [OPTION]... FILE+

Multi-file input specified in a text file:

$ rtg index [OPTION]... -I FILE

Example:

$ rtg index -f sam alignments.sam.gz

Parameters:

File Input/Output

-f

--format=FORMAT

Format of input to index. Allowed values are [sam, bam, cram, sv, coveragetsv, bed, vcf, auto] (Default is auto)

-I

--input-list-file=FILE

File containing a list of block compressed files (1 per line) containing genome position data.

FILE+

Block compressed files containing data to be indexed. May be specified 0 or more times.

Utility

-h

--help

Print help on command-line flag usage.

Usage:

Use the index command to produce tabix indexes for block compressed genome position data files like SAM files, VCF files, BED files, and the TSV output from RTG commands such as coverage. The index command can also be used to produce BAM indexes for BAM files with no index.

See also

map, coverage, snp, extract, bgzip

extract

Synopsis:

Extract specified parts of an indexed block compressed genome position data file.

Syntax:

Extract whole file:

$ rtg extract [OPTION]... FILE

Extract specific regions:

$ rtg extract [OPTION]... FILE STRING+

Example:

$ rtg extract alignments.bam 'chr1:2500000~1000'

Parameters:

File Input/Output

FILE

The indexed block compressed genome position data file to extract.

Filtering

REGION+

The range to display. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>. May be specified 0 or more times.

Reporting

--header

Set to also display the file header.

--header-only

Set to only display the file header.

Utility

-h

--help

Prints help on command-line flag usage.

Usage:

Use the extract command to view specific parts of indexed block compressed genome position data files such as those in SAM/BAM/BED/VCF format.

See also

map, coverage, snp, index, bgzip

aview

Synopsis:

View read mapping and variants corresponding to a region of the genome, with output as ASCII to the terminal, or HTML.

Syntax:

$ rtg aview [OPTION]... --region STRING -t SDF FILE+

Example:

$ rtg aview -t hg19 -b omni.vcf -c calls.vcf map/alignments.bam \
  --region Chr10:100000+3 –padding 30

Parameters:

File Input/Output

-b

--baseline=FILE

VCF file containing baseline variants.

-B

--bed=FILE

BED file containing regions to overlay. May be specified 0 or more times.

-c

--calls=FILE

VCF file containing called variants. May be specified 0 or more times.

-I

--input-list-file=FILE

File containing a list of SAM/BAM format files (1 per line)

-r

--reads=SDF

Read SDF (only needed to indicate correctness of simulated read mappings). May be specified 0 or more times.

-t

--template=SDF

SDF containing the reference genome.

FILE+

Alignment SAM/BAM files. May be specified 0 or more times.

Filtering

-p

--padding=INT

Padding around region of interest (Default is to automatically determine padding to avoid read truncation)

--region=REGION

The region of interest to display. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>

--sample=STRING

Specify name of sample to select. May be specified 0 or more times, or as a comma separated list.

Reporting

--html

Output as HTML.

--no-base-colors

Do not use base-colors.

--no-color

Do not use colors.

--no-dots

Display nucleotide instead of dots.

--print-cigars

Print alignment cigars.

--print-mapq

Print alignment MAPQ values.

--print-mate-position

Print mate position.

--print-names

Print read names.

--print-readgroup

Print read group id for each alignment.

--print-reference-line=INT

Print reference line every N lines (Default is 0)

--print-sample

Print sample id for each alignment.

--print-soft-clipped-bases

Print soft clipped bases.

--project-track=INT

If set, project highlighting for the specified track down through reads (Default projects the union of tracks)

--sort-readgroup

Sort reads first on read group and then on start position.

--sort-reads

Sort reads on start position.

--sort-sample

Sort reads first on sample id and then on start position.

--unflatten

Display unflattened CGI reads when present.

Utility

-h

--help

Print help on command-line flag usage.

Usage:

Use the aview command to display a textual view of mappings and variants corresponding to a small region of the reference genome. This is useful when examining evidence for variant calls in a server environment where a graphical display application such as IGV is not available. The aview command is easy to script in order to output displays for multiple regions for later viewing (either as text or HTML).

See also

map, snp

sdfstats

Synopsis:

Print statistics that describe a directory of SDF formatted data.

Syntax:

$ rtg sdfstats [OPTION]... SDF+

Example:

$ rtg sdfstats human_READS_SDF

Location           : C:\human_READS_SDF
Parameters         : format -f solexa -o human_READS_SDF
                                c:\users\Elle\human\SRR005490.fastq.gz
SDF Version        : 6
Type               : DNA
Source             : SOLEXA
Paired arm         : UNKNOWN
Number of sequences: 4193903
Maximum length     : 48
Minimum length     : 48
N                  : 931268
A                  : 61100096
C                  : 41452181
G                  : 45262380
T                  : 52561419
Total residues     : 201307344
Quality scores available on this SDF

Parameters:

File Input/Output

SDF+

Specifies an SDF on which statistics are to be reported. May be specified 1 or more times.

Reporting

--lengths

Set to print out the name and length of each sequence. (Not recommended for read sets).

-p

--position

Set to include information about unknown bases (Ns) by read position.

-q

--quality

Set to display mean of quality.

--sex=SEX

Set to display the reference sequence list for the given sex. Allowed values are [male, female, either]. May be specified 0 or more times, or as a comma separated list.

--taxonomy

Set to display information about the taxonomy.

-n

--unknowns

Set to include information about unknown bases (Ns).

Utility

-h

--help

Prints help on command-line flag usage.

Usage:

Use the sdfstats command to get information about the contents of SDFs.

sdfsubset

Synopsis:

Extracts a specified subset of sequences from one SDF and outputs them to another SDF.

Syntax:

Individual specification of sequence ids:

$ rtg sdfsubset [OPTION]... -i SDF -o SDF STRING+

File list specification of sequence ids:

$ rtg sdfsubset [OPTION]... -i SDF -o SDF -I FILE

Example:

$ rtg sdfsubset -i reads -o subset_reads 10 20 30 40 50

Parameters:

File Input/Output

-i

--input=SDF

Specifies the input SDF.

-o

--output=SDF

The name of the output SDF.

Filtering

--end-id=INT

Only output sequences with sequence id less than the given number. (Sequence ids start at 0).

--start-id=INT

Only output sequences with sequence id greater than or equal to the given number. (Sequence ids start at 0).

-I

--id-file=FILE

Name of a file containing a list of sequences to extract, one per line.

--names

Interpret any specified sequence as names instead of numeric sequence ids.

STRING+

Specifies the sequence id, or sequence name if the names flag is set to extract from the input SDF. May be specified 0 or more times.

Utility

-h

--help

Prints help on command-line flag usage.

Usage:

Use this command to obtain a subset of sequences from an SDF. Either specify the subset on the command line as a list of space-separated sequence ids or using the --id-file parameter to specify a file containing a list of sequence ids, one per line. Sequence ids start from zero and are the same as the ids that map uses by default in the QNAME field of its BAM files.

For example:

$ rtg sdfsubset -i reads -o subset_reads 10 20 30 40 50

This will produce an SDF called subset_reads with sequences 10, 20, 30, 40 and 50 from the original SDF contained in it.

See also

sdfsubseq, sdfstats

sdfsubseq

Synopsis:

Prints a subsequence of a given sequence in an SDF.

Syntax:

Print sequences from sequence names:

$ rtg sdfsubseq [OPTION]... -i FILE STRING+

Print sequences from sequence ids:

$ rtg sdfsubseq [OPTION]... -i FILE -I STRING+

Example:

$ rtg sdfsubseq -i reads -I 0:1+100

Parameters:

File Input/Output

-i

--input=FILE

Specifies the input SDF.

Filtering

-I

--sequence-id

If set, use sequence id instead of sequence name in region (0-based)

REGION+

The range to display. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>. Must be specified 1 or more times.

Utility

-f

--fasta

Set to output in FASTA format.

-q

--fastq

Set to output in FASTQ format.

-h

--help

Prints help on command-line flag usage.

-r

--reverse-complement

Set to output in reverse complement.

Usage:

Prints out the nucleotides or amino acids of specified regions in a set of sequences.

For example:

$ rtg sdfsubseq --input reads --sequence-id 0:1+20
AGGCGTCTGCAGCCGACGCG

See also

sdfsubset, sdfstats

mendelian

Synopsis:

The mendelian command checks a multi-sample VCF file for variant calls which do not follow Mendelian inheritance, and compute aggregate sample concordance.

Syntax:

$ rtg mendelian [OPTION]... -i FILE -t SDF

Example:

$ rtg mendelian -i family.vcf.gz -t genome_ref

Parameters:

File Input/Output

-i

--input=FILE

VCF file containing multi-sample variant calls. Use ‘-‘ to read from standard input.

-o

--output=FILE

If set, output annotated calls to this VCF file. Use ‘-‘ to write to standard output.

--output-consistent=FILE

If set, output only consistent calls to this VCF file.

--output-inconsistent=FILE

If set, output only non-Mendelian calls to this VCF file.

-t

--template=SDF

SDF containing the reference genome.

Sensitivity Tuning

--all-records

Use all records, regardless of filters (Default is to only process records where FILTER is . or PASS)

-l

--lenient

Allow homozygous diploid calls in place of haploid calls and assume missing values are equal to the reference.

--min-concordance=FLOAT

Percentage concordance required for consistent parentage (Default is 99.0)

--pedigree=FILE

Genome relationships PED file (Default is to extract pedigree information from VCF header fields)

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

Usage:

Given a multi-sample VCF file for a nuclear family with a defined pedigree, the mendelian command examines the variant calls and outputs the number of violations of Mendelian inheritance. If the --output-inconsistent parameter is set, all detected violations are written into an output VCF file. As such, this command may be regarded as a VCF filter, outputting those variant calls needing a non-Mendelian explanation. Such calls may be the consequence of sequencing error, calling on low-coverage, or genuine novel variants in one or more individuals.

Pedigree information regarding the relationships between samples and the sex of each sample is extracted from the VCF headers automatically created by the RTG pedigree-aware variant calling commands. If this pedigree information is absent from the VCF header or is incorrect, a pedigree file can be explicitly supplied with the --pedigree flag.

To ensure correct behavior when dealing with sex chromosomes it is necessary to specify a sex-aware reference and ensure the sex of each sample is supplied as part of the pedigree information. While it is best to give the reference SDF used in the creation of the VCF, for checking third-party outputs any reference SDF containing the same chromosome names and an appropriate reference.txt file will work. For more information, see RTG reference file format. Variants calls where the call ploidy does not match what is expected are annotated in the output VCF with an MCP FORMAT annotation.

Particularly when evaluating VCF files that have been produced by third party tools or when the VCF is the result of combining independent per-sample calling, it is common to end up with situations where calls are not available for every member of the family. Under normal circumstances mendelian will attempt to determine Mendelian consistency on the basis of the values that have been provided. Records where the presence of missing values makes the Mendelian consistency undecidable contain MCU INFO annotations in the annotated output VCF. The following examples illustrate some consistent, undecidable, and inconsistent calls in the presence of missing values:

CHROM  FATHER_GT  MOTHER_GT  SON_GT  STATUS
chrX   .          0/1        1       OK
chr1   ./.        1/1        1/2     MCU
chr1   ./.        1/1        2/2     MCV

Since the number of calls where one sample is missing can be quite high, an alternative option is to treat missing values as equal to the reference by using the --lenient parameter. Note that while this approach will be correct in most cases, it will give inaccurate results where the calling between different samples has reported the variant in an equivalent but slightly different position or representation (e.g. positioning of indels within homopolymer regions, differences of representation such as splitting MNPs into multiple SNPs etc).

The mendelian command computes overall concordance between related samples to assist detecting cases where pedigree has been incorrectly recorded or samples have been mislabeled. For each child in the pedigree, pairwise concordance is computed with respect to each parent by identifying diploid calls where the parent does not contain either allele called in the child. Low pairwise concordance with a single parent may indicate that the parent is the source of the problem, whereas low pairwise concordance with both parents may indicate that the child is the source of the problem. A stricter three-way concordance is also recorded.

By default, only VCF records with the FILTER field set to PASS or missing are processed. All variant records can be examined by specifying the --all-records parameter.

vcfannotate

Synopsis:

Used to add annotations to a VCF file, either to the VCF ID field, as a VCF INFO sub-field, or as a VCF FORMAT sub-field.

Syntax:

$ rtg vcfannotate [OPTION]... -b FILE -i FILE -o FILE

Example:

$ rtg vcfannotate -b dbsnp.bed -i snps.vcf.gz -o snps-dbsnp.vcf.gz

Parameters:

File Input/Output

--bed-regions=FILE

If set, only read VCF records that overlap the ranges contained in the specified BED file.

-i

--input=FILE

VCF file containing variants to annotate. Use ‘-‘ to read from standard input.

-o

--output=FILE

Output VCF file name. Use ‘-‘ to write to standard output.

--region=REGION

If set, only read VCF records within the specified range. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>

Reporting

-A

--annotation=STRING

Add computed annotation to VCF records. Allowed values are [AC, AN, EP, GQD, IC, LAL, MEANQAD, NAA, PD, QA, QD, RA, SCONT, VAF, VAF1, ZY]. May be specified 0 or more times, or as a comma separated list.

--bed-ids=FILE

Add variant IDs from BED file. May be specified 0 or more times.

--bed-info=FILE

Add INFO annotations from BED file. May be specified 0 or more times.

--fill-an-ac

Add or update the AN and AC INFO fields.

--info-description=STRING

If the BED INFO field is not already declared, use this description in the header (Default is Annotation)

--info-id=STRING

The INFO ID for BED INFO annotations (Default is ANN)

--relabel=FILE

Relabel samples according to old-name new-name pairs in specified file.

--vcf-ids=FILE

Add variant IDs from VCF file. May be specified 0 or more times.

Utility

-a

--add-header=STRING|FILE

File containing VCF header lines to add, or a literal header line. May be specified 0 or more times.

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--no-header

Prevent VCF header from being written.

Usage:

Use vcfannotate to add text annotations to variants.

A common use case is to add annotations to only those variants that fall within ranges specified in a BED or VCF file, supplied via --bed-ids or --vcf-ids respectively. The annotations from the BED file are added as an INFO field in the output VCF file. It can also be used to compute or fill in certain additional annotations from the existing content. Note that this annotation method is solely based on the position and span of the variant, ignoring actual alleles and genotypes.

If the --bed-ids flag is used, instead of adding the annotation to the INFO fields, it is added to the ID column of the VCF file instead. If the --vcf-ids flag is used, the ID column of the input VCF file is used to update the ID column of the output VCF file instead.

If the --fill-an-ac flag is set, the output VCF will have the AN and AC info fields (as defined in the VCF 4.2 specification) created or updated.

It is also possible to use vcfannotate to insert additional VCF header lines into the VCF header. These are supplied using the --add-header flag which may either be a literal VCF header line (useful for adding one or two header lines), or from a file.

$ rtg vcfannotate -i in.vcf.gz -o out.vcf.gz \
--add-header "##SAMPLE=<ID=NA24385,Sex=MALE>" \
--add-header "##SAMPLE=<ID=NA24143,Sex=FEMALE>" \
--add-header "##SAMPLE=<ID=NA24149,Sex=MALE>" \
--add-header "##PEDIGREE=<Child=NA24385,Mother=NA24143,Father=NA24149>"

or alternatively:

$ rtg vcfannotate -i in.vcf.gz -o out.vcf.gz --add-header ped_vcf_headers.txt

Care should be taken that the lines being inserted are valid VCF header lines.

If the --annotation flag is set, vcfannotate attempts to compute the specified annotation(s) and add them as FORMAT fields in the corresponding records. Records for which particular annotations cannot be computed, due to a lack of pre-requisite fields, will not be modified.

For a description of the meaning of fields available for annotation, see Small-variant VCF output file description. The SCONT annotation is a convenience to annotate with all of the contrary evidence annotations: DCOC, DCOF, OCOC, OCOF.

vcfdecompose

Synopsis:

Decomposes complex variants within a VCF file into smaller components.

Syntax:

$ rtg vcfdecompose [OPTION]... -i FILE -o FILE

Parameters:

File Input/Output

-i

--input=FILE

VCF file containing variants to decompose. Use ‘-‘ to read from standard input.

-o

--output=FILE

Output VCF file name. Use ‘-‘ to write to standard output.

-t

--template=SDF

SDF of the reference genome the variants are called against.

Sensitivity Tuning

--break-indels

If set, peel as many SNPs off an indel as possible.

--break-mnps

If set, break MNPs into individual SNPs.

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--no-header

Prevent VCF header from being written.

Usage:

The vcfdecompose command decomposes and trims variants based on a multiple sequence alignment between the alleles in each VCF record. Only records where every ALT allele is an ordinary allele (i.e. consisting of nucleotides) will undergo decomposition. In addition, if there are redundant same-as-reference bases in the alleles, these will be trimmed off.

The default behavior is to break the variant at positions where there is at least one base aligned to the reference across all ALT alleles, so the output may contain MNPs or impure indels. If desired, MNPs can be split into individual SNPs via --break-mnps. Similarly, impure indels can be split into a combination of SNPs and pure indels via --break-indels.

Although decomposed variants carry through the original INFO and FORMAT annotations, the decomposition may mean that some annotations are no longer semantically correct. In particular, any VCF FORMAT fields declared to be of type A, G, or R will no longer be valid if the set of alleles has changed.

Note that the reference genome is an optional parameter. When variants are decomposed and trimmed, the resulting variant may require a padding base to be added, as required by the VCF specification. The VCF specification suggests that the padding base should be the base before the variant (i.e. padding on the left), but sometimes this requires knowledge of reference bases not present in the original record. When the reference genome is supplied, vcfdecompose will ensure that any padding bases are added on the left of the variant. If the reference genome is not supplied, padding bases may sometimes be on the right hand side of the variant. For example:

1  20  .  GCGCGCGCGCG  TTTGCGCGCTTGCGCGTTT  .  PASS  .              GT  1/0

will decompose without a reference genome as:

1  20  .  G            TTTG                 .  PASS  ORP=20;ORL=11  GT  1/0
1  25  .  C            CTT                  .  PASS  ORP=20;ORL=11  GT  1/0

and with a reference genome (where the reference base at position 19 can be determined to be a T) as:

1  19  .  T            TTTT                 .  PASS  ORP=20;ORL=11  GT  1/0
1  25  .  C            CTT                  .  PASS  ORP=20;ORL=11  GT  1/0

The variants that are left vs right-padded are equivalent and identified as such by haplotype-aware comparison tools such as vcfeval.

See also

vcffilter, vcfeval

vcfeval

Synopsis:

Evaluates called variants for agreement with a baseline variant set irrespective of representational differences. Outputs a weighted ROC file which can be viewed with rtg rocplot and VCF files containing false positives (called variants not matched in the baseline), false negatives (baseline variants not matched in the call set), and true positives (variants that match between the baseline and calls).

The baseline variants might be the variants that were used to generate a synthetic simulated sample (such as via popsim, samplesim, etc), a gold-standard VCF corresponding to a reference sample such as NA12878, or simply an alternative call-set being used as a basis for comparison.

Syntax:

$ rtg vcfeval [OPTION]... -b FILE -c FILE -o DIR -t SDF

Example:

$ rtg vcfeval -b goldstandard.vcf.gz -c snps.vcf.gz -t HUMAN_reference \
  --sample daughter -f AVR -o eval

Parameters:

File Input/Output

-b

--baseline=FILE

VCF file containing baseline variants.

--bed-regions=FILE

If set, only read VCF records that overlap the ranges contained in the specified BED file.

-c

--calls=FILE

VCF file containing called variants.

-e

--evaluation-regions=FILE

If set, evaluate within regions contained in the supplied BED file, allowing transborder matches. To be used for truth-set high-confidence regions or other regions of interest where region boundary effects should be minimized.

-o

--output=DIR

Directory for output.

--region=REGION

If set, only read VCF records within the specified range. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>

-t

--template=SDF

SDF of the reference genome the variants are called against.

Filtering

--all-records

Use all records regardless of FILTER status (Default is to only process records where FILTER is . or PASS)

--decompose

Decompose complex variants into smaller constituents to allow partial credit.

--ref-overlap

Allow alleles to overlap where bases of either allele are same-as-ref (Default is to only allow VCF anchor base overlap)

--sample=STRING

The name of the sample to select. Use <baseline_sample>,<calls_sample> to select different sample names for baseline and calls. (Required when using multi-sample VCF files)

--sample-ploidy=INT

Expected ploidy of samples (Default is 2)

--squash-ploidy

Treat heterozygous genotypes as homozygous ALT in both baseline and calls, to allow matches that ignore zygosity differences.

Reporting

--at-precision=FLOAT

Output summary statistics where precision >= supplied value (Default is to summarize at maximum F-measure)

--at-sensitivity=FLOAT

Output summary statistics where sensitivity >= supplied value (Default is to summarize at maximum F-measure)

--no-roc

Do not produce ROCs.

-m

--output-mode=STRING

Output reporting mode. Allowed values are [split, annotate, combine, ga4gh, roc-only] (Default is split)

--roc-expr=STRING

Output ROC file for variants matching custom JavaScript expression. Use the form <LABEL>=<EXPRESSION>. May be specified 0 or more times.

--roc-regions=STRING

Output ROC file for variants overlapping custom regions supplied in BED file. Use the form <LABEL>=<FILENAME>. May be specified 0 or more times.

--roc-subset=STRING

Output ROC file for preset variant subset. Allowed values are [hom, het, snp, non-snp, mnp, indel]. May be specified 0 or more times, or as a comma separated list.

-O

--sort-order=STRING

The order in which to sort the ROC scores so that good scores come before bad scores. Allowed values are [ascending, descending] (Default is descending)

-f

--vcf-score-field=STRING

The name of the VCF FORMAT field to use as the ROC score. Also valid are QUAL, INFO.<name> or FORMAT.<name> to select the named VCF FORMAT or INFO field (Default is GQ)

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

-T

--threads=INT

Number of threads (Default is the number of available cores)

Usage:

The vcfeval command can be used to generate VCF files containing called variants that were in the baseline VCF, called variants that were not in the baseline VCF and baseline variants that were not in the called variants. It also produces ROC curve data files based on a score contained in a VCF field which show the predictive power of that field for the quality of the variant calls.

When developing and validating sequencing pipelines and variant calling algorithms, the comparison of variant call sets is a common problem. The naïve way of computing these numbers is to look at the same reference locations in the baseline (ground truth) and called variant set, and see if genotype calls match at the same position. However, a complication arises due to possible differences in representation for indels between the baseline and the call sets within repeats or homopolymers, and in multiple-nucleotide polymorphisms (MNPs), which encompass several nearby nucleotides and are locally phased. The vcfeval command includes a novel dynamic-programming algorithm for comparing variant call sets that deals with complex call representation discrepancies, and minimizes false positives and negatives across the entire call sets for accurate performance evaluation. A primary advantage of vcfeval (compared to other tools) is that the evaluation does not depend on normalization or decomposition, and so the results of analysis can easily be used to relate to the original variant calls and their annotations.

Note that vcfeval operates at the level of local haplotypes for a sample, so for a diploid genotype, both alleles must match in order to be considered correct. Some of the vcfeval output modes (described below) automatically perform an additional haploid analysis phase to identify variants which may not have a diploid match but which share a common allele (for example, zygosity errors made during calling). If desired, this more lenient haploid comparison can be used at the outset by setting the --squash-ploidy flag (see below).

Note that variants selected for inclusion in a haplotype cannot be permitted to overlap each other (otherwise the question arises of which variant should have priority when determining the resulting haplotype), and any well-formed call-set should not contain these situations in order to avoid such ambiguity. When such cases are encountered by vcfeval, the best non-overlapping result is determined. A special case of overlapping variants is where calls are denoted as partially the same as the reference (for example, a typical heterozygous call). Strictly speaking such variants are an assertion that the relevant haplotype bases must not be altered from the reference and overlap should not be permitted (this is the interpretation that vcfeval employs by default). However, sometimes as a result of using non-haplotype-aware variant calling tools or when using naïve merging of multiple call sets, a more lenient comparison is desired. The --ref-overlap flag will permit such overlapping variants to both match, as long as any overlap only occurs where one variant or other has asserted haplotype bases as being the same as reference.

Common allele matching with --squash-ploidy

When --squash-ploidy is specified, a haploid match is attempted using each of the non-reference alleles used in the sample genotype. For example if the baseline and call VCFs each had a record with the same REF and ALT alleles declared, the following GT fields would be considered a match:

0/1, 1/1, 1/2   (genotypes match due to the 1 allele)
0/2, 1/2, 2/2   (genotypes match due to the 2 allele)

Thus --squash-ploidy matches any case where the baseline and calls share a common allele. This is most often used to run matching that does not penalize for genotyping errors. For example, it is recommended to use this option when matching somatic variant calls, as since somatic variation is usually associated with variable allelic fractions and heterogeneity that mean strict diploid genotype comparisons are not appropriate.

Comparing non-diploid genomes

By default, vcfeval assumes diploid organisms (that is, the expected ploidy of any GT call is 2). As a special case to ease the comparison of male calls on sex chromosomes (where callers often continue to use diploid representation), haploid calls are treated as homozygous diploid. Any calls made with unexpected ploidy are ignored and reported in the vcfeval log file.

To compare genomes with non-diploid ploidy, the expected sample ploidy can be overridden via --sample-ploidy – for example --sample-ploidy=4 would be used to compare tetraploid organisms.

Comparing with a VCF that has no sample column

A common scenario is to match a call set against a baseline which contains no sample column, where the objective is to identify which baseline alleles which have been called. One example of this is to identify whether calls match a database of known high-priority somatic variants such as COSMIC, or to find calls which have been previously seen in a population allele database such as ExAC. Ordinarily vcfeval requires the input VCFs to contain a sample column containing a genotype in the GT field, however, it is possible to specify a special sample name of ‘ALT’ in order to indicate that the the genotypes for comparison should be derived from the ALT alleles of the record. This can be specified independently for baseline and calls, for example:

$ rtg vcfeval -t build37.sdf -b cosmic.vcf.gz -c tumor-calls.vcf.gz \
--squash-ploidy --sample ALT,tumor -o tumor-vs-cosmic

Which would perform a haploid matching of the GT of the called sample ‘tumor’ against all possible haploid genotypes in the COSMIC VCF. The resulting true positives file contains all the calls containing an allele present in the COSMIC VCF.

Note

It is also possible to run a diploid comparison by omitting --squash-ploidy, but this is not usually required, and is computationally more intensive since there may be many more possible diploid genotypes to explore, particularly if the ALT VCF contains many multiallelic records.)

Evaluation with respect to regions

When evaluating exome variant calls, it may be useful to restrict analysis only to exome target regions. In this case, supply a BED file containing the list of regions to restrict analysis to via the --bed-regions flag. For a quick way to restrict analysis only to a single region, the --region flag is also accepted. Note that when restricting analysis to regions, there may be variants which can not be correctly evaluated near the borders of each analysis region, if determination of equivalence would require inclusion of variants outside of the region. For this reason, it is recommended that such regions be relatively inclusive.

When matching against gold standard truth sets which have an accompanying high-confidence regions BED file, the flag --evaluation-regions should be used instead of --bed-regions, as it has special matching semantics that aims to reduce comparison region boundary effects. When this comparison method is used, call variants which match a baseline variant are only considered a true positive if the baseline variant is inside the high confidence regions, and call variants are only considered false positive if they fall inside the high confidence regions.

vcfeval outputs

The primary outputs of vcfeval are VCF files indicating which variants matched between the baseline and the calls VCF, and data files containing information used to generate ROC curves with the rocplot command (or via spreadsheet). vcfeval supports different VCF output modes which can be selected with the --output-mode flag according to the type of analysis workflow desired. The following modes are available:

Split (--output-mode=split)

This output mode is the default, and produces separate VCF files for each of the match categories. The individual VCF records in these files are not altered in any way, preserving all annotations present in the input files.

  • tp.vcf – contains those variants from the calls VCF which agree with variants in the baseline VCF

  • tp-baseline.vcf – contains those variants from the baseline VCF which agree with variants in the calls VCF. Thus, the variants in tp.vcf and tp-baseline.vcf are equivalent. This file can be used to successively refine a highly sensitive baseline variant set to produce a consensus from several call sets.

  • fp.vcf – contains variants from the calls VCF which do not agree with baseline variants.

  • fn.vcf – contains variants from the baseline VCF which were not correctly called.

This mode performs a single pass comparison, either in diploid mode (the default), or haploid mode (if --squash-ploidy has been set). The separate output files produced by this mode allow the use of vcfeval as an advanced haplotype-aware VCF intersection tool.

Annotate (--output-mode=annotate)

This output mode does not split the input VCFs by match status, but instead adds INFO annotations containing the match status of each record:

  • calls.vcf – contains variants from the calls VCF, augmented with match status annotations.

  • baseline.vcf – contains variants from the baseline VCF, augmented with match status annotations.

This output mode automatically performs two comparison passes, the first finds diploid matches (assigned a match status of TP), and a second pass that applies a haploid mode to the false positives and false negatives in order to find calls (such as zygosity errors) that contain a common allele. This second category of match are annotated with status FN_CA or FP_CA in the output VCFs, and those calls which do not have any match are assigned status FN or FP. A status value of IGN indicates a VCF record which was ignored (for example, due to having a non-PASS filter status, representing a structural variant, or otherwise containing a non-variant genotype). A status of OUT indicates a VCF record which does not contain a match status due to falling outside the evaluation regions when --evaluation-regions is being used. The annotated VCF files produced in this mode may also be used with vcf2rocplot to produce additional post-evaluation ROC data files.

Combine (--output-mode=combine)

This output mode provides an easy way to view the baseline and call variants in a single two-sample VCF.

  • output.vcf – contains variants from both the baseline and calls VCFs, augmented with match status annotations. The sample under comparison from each of the input VCFs is extracted as a column in the output. As the VCF records from the baseline and calls typically have very different input annotations which can be difficult to merge, and to keep the output format simple, there is no attempt to preserve any of the original variant annotations.

As with the annotation output mode, this output mode automatically performs two comparison passes to find both diploid matches and haploid (lenient) matches.

ROC-only (--output-mode=roc-only)

This output mode provides a lightweight way to run performance benchmarking, as VCF file output is omitted, and only ROC data files are produced.

Note

In addition, vcfeval has an output mode (--output-mode=ga4gh) which produces the intermediate evaluation format defined by the GA4GH Benchmarking Team, without additional statistics files. This mode is not generally intended for end users, rather it is used when vcfeval is selected as the comparison engine inside the hap.py benchmarking tool see: https://github.com/ga4gh/benchmarking-tools and https://github.com/Illumina/hap.py

Additional ROC stratifications

All of the output modes produce the following ROC data files (unless disabled by --no-roc):

  • weighted_roc.tsv – contains ROC data derived from all analyzed call variants, regardless of their representation. Columns include the score field, and standard accuracy metrics such as true positives, false positives, false negatives, precision, sensitivity, and f-measure corresponding to each score threshold.

  • snp_roc.tsv – contains ROC data derived from only those variants which were represented as SNPs. Since the representation conventions can differ between the baseline and calls, there are some subtleties to be aware of when interpreting metrics such as precision, sensitivity, etc, described below.

  • non_snp_roc.tsv – contains ROC data derived from those variants which were not represented as SNPs. As above, not all metrics are computed for this file.

vcfeval also provides the ability to produce additional ROC data files corresponding to preset and customized variant stratifications with the following flags:

Preset stratifications

The --roc-subset flag allows selection from preset stratifications based on variant type (according to their representation in the relevant input VCF):

  • hom – homozygous variants only

  • het – heterozygous variants only

  • snp – SNP variants (enabled by default)

  • non-snp – non-SNP variants (enabled by default)

  • mnp – multi-nucleotide polymorphisms only

  • indel – length-changing variants only

Multiple presets can be enabled in a single run, e.g. --roc-subset hom,het,indel

Region-based stratifications

The --roc-regions flag produces a stratified ROC data file using only variants that overlap regions specified in a user-supplied BED file. The special syntax for this flag is: --roc-regions LABEL=FILE, where LABEL is a short tag used to determine ROC output file names and FILE is the path to the relevant BED file. For example, to produce additional stratifications based on BED files partitioning the genome based on GC content:

$ rtg vcfeval -t build37.sdf -b baseline.vcf.gz -c calls.vcf.gz \
  --roc-regions GC55TO60=/path/to/GCcontent/GRCh37_gc55to60_slop50.bed.gz \
  --roc-regions GC60TO65=/path/to/GCcontent/GRCh37_gc60to65_slop50.bed.gz
Custom JavaScript based stratifications

The above stratification flags will satisfy most common usages, but vcfeval also includes the ability to write custom stratifications using JavaScript expressions (similar to vcffilter --keep-expr). The special syntax for this flag is: --roc-expr LABEL=EXPRESSION, where LABEL is a short tag used to determine ROC output file names and EXPRESSION is the JavaScript expression that accepts a variant for inclusion in the stratification. This is most useful when the input VCFs contain annotations useful for the stratification. For example, to produce stratifications based on depth of coverage during variant calling:

$ rtg vcfeval -t build37.sdf -b baseline.vcf.gz -c calls.vcf.gz \
  --roc-expr "DP10TO20=has(INFO.DP) && INFO.DP>=10 && INFO.DP<20" \
  --roc-expr "DP20TO30=has(INFO.DP) && INFO.DP>=20 && INFO.DP<30" \
  --roc-expr "DP30TO40=has(INFO.DP) && INFO.DP>=30 && INFO.DP<40"

Tips:

  • Ensure the expression is valid to evaluate on all variants (for example, take care when referring to sample fields names if the sample names are different between baseline and calls files).

  • It may be useful to test or debug the expression (without the label) via vcffilter --keep-expr.

For more information on JavaScript expressions, see RTG JavaScript filtering API

Benchmarking comparisons using ROC and precision/sensitivity curves

Multiple ROC data files (from a single or several vcfeval runs) can be plotted with the rocplot command, which allows output to a PNG or SVG image or analysis in an interactive GUI that provides zooming and visualization of the effects of threshold adjustment. As these files are simple tab-separated-value format, they can also be loaded into a spreadsheet tool or processed with shell scripts.

While ROC curve analysis provides a much more thorough method for examining the performance of a call set with respect to a baseline truth set, for convenience, vcfeval also produces a summary.txt file which indicates match summary statistics that correspond to two key points on the ROC curve. The first point is where all called variants are included (i.e., no thresholding on a score value); and second point corresponding to a score threshold that maximises the F-measure of the curve. While this latter point is somewhat arbitrary, it represents a balanced tradeoff between precision and sensitivity which is likely to provide a fairer comparison when comparing call sets from different callers.

Sometimes it is useful to perform the summary statistic evaluation at some point other than maximized F-measure (for example when comparing a large number of results at a particular precision level). This can be accomplished by specifying a different point using either the --at-precision or --at-sensitivity flag with a value in the range [0, 1].

Note that vcfeval reports true positives both counted using the baseline variant representation as well as counted using the call variant representation. When these numbers differ greatly, it indicates a general difference in representational conventions used between the two call sets. Since false negatives can only be measured in terms of the baseline representation, sensitivity is defined as:

\text{Sensitivity} = \text{TP}_\text{baseline} / (\text{TP}_\text{baseline} + \text{FN}).

Conversely since false positives can only be measured in terms of the call representation, precision is defined as:

\text{Precision} = \text{TP}_\text{call} / (\text{TP}_\text{call} + \text{FP}).

Note

For definitions of the terminology used when evaluating caller accuracy, see: https://en.wikipedia.org/wiki/Receiver_operating_characteristic and https://en.wikipedia.org/wiki/Sensitivity_and_specificity

Benchmarking performance for SNPs versus indels

A common desire is to perform analysis separately for SNPs versus indels. However, it is important to note that due the representation ambiguity problem, it is not always trivial to decide in a global sense whether a variant is a SNP or an indel or other complex variant. A group of variants that may be represented as single SNPs in one call-set may be represented as a single complex variant in another call-set. Consider the following example reference and alternate haplotypes:

     12345678901234567
REF: ATCGTAAATAAAATGCA
ALT: ATCGTAAAATAAATGCA

One variant caller might represent the haplotypes as the following VCF records:

chr1 5 . T TA . . . GT 1/1
chr1 9 . TA T . . . GT 1/1

While another variant caller could represent the same haplotypes as:

chr1 9 . T A . . . GT 1/1
chr1 10 . A T . . . GT 1/1

The decision as to which representation to use is essentially arbitrary, yet one caller has used indels (and no SNPs), and the other has used SNPs (and no indels). For this reason it is certainly a poor idea to attempt to divide baseline and called variants into separate SNP and indel datasets up front and perform evaluation on each set separately, as any variants that use different representation categories will not be matched across the independent comparisons. Any variant-type specific metrics should be computed after matching is carried out on the full variant sets.

Note that when there are different representational conventions between the baseline and calls (or between calls from one variant caller and another), then at some level there is really a semantic difference between a “baseline indel” and a “call-set indel” (or “variant-caller-A indel” and “variant-caller-B indel”), so caution should be applied when making conclusions related to SNP versus indel accuracy.

In the snp_roc.tsv and non_snp_roc.tsv output files (and other preset stratifications available via --roc-subset), vcfeval notes the number of baseline and call variants of each variant type. When considering benchmarking metrics in the absence of any thresholding with respect to a score field, it is straight-forward to use the previous formulae (i.e. sensitivity is computed using the counts from baseline variants, and precision is computed using the counts from called variants). When computing threshold-specific metrics for ROC data points, the computation is more involved. Since only the call variants contain the score field used to rank variants, the number of (say) TP baseline indels that exceed threshold x is not defined. vcfeval computes a scaled count as:

\text{TP}_\text{baseline\_indel}(x) = \text{TP}_\text{call\_indel}(x) \times \text{TP}_\text{baseline\_indel} / \text{TP}_\text{call\_indel}

and thus threshold-specific sensitivity is computed as

\text{Sensitivity}_\text{indel}(x) = \text{TP}_\text{baseline\_indel}(x) / (\text{TP}_\text{baseline\_indel} + \text{FN}_\text{indel})

This scaling ensures that the end point of the variant type specific ROC or precision/sensitivity curve ends at the same point that is obtained when computing metrics without any threshold.

The scaling described above is applied to all of the preset stratifications available via --roc-subset, but is not applied to any custom stratifications produced via --roc-regions or --roc-expr.

Variant decomposition and benchmarking

In general, it is not necessary to run any variant decomposition and/or normalization on variant call sets prior to evaluation with vcfeval, as the haplotype aware matching process can account for representation differences. However, since matching is at the granularity of entire variants, a single long complex call will be categorized as either correct or incorrect, even if part of the call may match. If partial credit in the case of long calls is of interest, vcfeval includes an option to internally decompose variants prior to matching, using the --decompose flag. This decomposition is applied to both baseline and call variants, and any output VCFs will contain the decomposed representation. External VCF decomposition (with more control over decomposition options) is also available via rtg vcfdecompose.

vcffilter

Synopsis:

Filters VCF records based on various criteria. When filtering on multiple samples, if any of the specified samples fail the criteria, the record will be filtered. By default filtered records are removed, but see the –fail, –clear-failed-samples, and –fail-samples options for alternatives.

Syntax:

$ rtg vcffilter [OPTION]... -i FILE -o FILE

Examples:

Keep only records where the sample has depth of coverage at least 5:

$ rtg vcffilter -i snps.vcf.gz -o snps_cov5.vcf.gz -d 5

Keep only biallelic records:

$ rtg vcffilter -i snps.vcf.gz -o snps_biallelic.vcf.gz --max-alleles 2

Parameters:

File Input/Output

--all-samples

Apply sample-specific criteria to all samples contained in the input VCF.

--bed-regions=FILE

If set, only read VCF records that overlap the ranges contained in the specified BED file.

-i

--input=FILE

VCF file containing variants to be filtered. Use ‘-‘ to read from standard input.

-o

--output=FILE

Output VCF file. Use ‘-‘ to write to standard output. This option is required, unless --javascript is being used.

--region=REGION

If set, only read VCF records within the specified range. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>

--sample=STRING

Apply sample-specific criteria to the named sample contained in the input VCF. May be specified 0 or more times.

Filtering (Record based)

-w

--density-window=INT

Window within which multiple variants are discarded.

--exclude-bed=FILE

Discard all variants within the regions in this BED file.

--exclude-vcf=FILE

Discard all variants that overlap with the ones in this file.

--include-bed=FILE

Only keep variants within the regions in this BED file.

--include-vcf=FILE

Only keep variants that overlap with the ones in this file.

-j

--javascript=STRING

Javascript filtering functions for determining whether to keep record. May be either an expression or a file name. May be specified 0 or more times. See Examples

-e

--keep-expr=STRING

Records for which this expression evaluates to true will be retained. See Examples

-k

--keep-filter=STRING

Only keep variants with this FILTER tag. May be specified 0 or more times, or as a comma separated list.

-K

--keep-info=STRING

Only keep variants with this INFO tag. May be specified 0 or more times, or as a comma separated list.

--max-alleles=INT

Maximum number of alleles (REF included)

-C

--max-combined-read-depth=INT

Maximum allowed combined read depth.

-Q

--max-quality=FLOAT

Maximum allowed quality.

--min-alleles=INT

Minimum number of alleles (REF included)

-c

--min-combined-read-depth=INT

Minimum allowed combined read depth.

-q

--min-quality=FLOAT

Minimum allowed quality.

-r

--remove-filter=STRING

Remove variants with this FILTER tag. May be specified 0 or more times, or as a comma separated list.

-R

--remove-info=STRING

Remove variants with this INFO tag. May be specified 0 or more times, or as a comma separated list.

--remove-overlapping

Remove records that overlap with previous records.

Filtering (Sample based)

-A

--max-ambiguity-ratio=FLOAT

Maximum allowed ambiguity ratio.

--max-avr-score=FLOAT

Maximum allowed AVR score.

--max-denovo-score=FLOAT

Maximum de novo score threshold.

-G

--max-genotype-quality=FLOAT

Maximum allowed genotype quality.

-D

--max-read-depth=INT

Maximum allowed sample read depth.

--min-avr-score=FLOAT

Minimum allowed AVR score.

--min-denovo-score=FLOAT

Minimum de novo score threshold.

-g

--min-genotype-quality=FLOAT

Minimum allowed genotype quality.

-d

--min-read-depth=INT

Minimum allowed sample read depth.

--non-snps-only

Only keep where sample variant is MNP or INDEL.

--remove-all-same-as-ref

Remove where all samples are same as reference.

--remove-hom

Remove where sample is homozygous.

--remove-same-as-ref

Remove where sample is same as reference.

--snps-only

Only keep where sample variant is a simple SNP.

Reporting

--clear-failed-samples

Retain failed records, set the sample GT field to missing.

-f

--fail=STRING

Retain failed records, add the provided label to the FILTER field.

-F

--fail-samples=STRING

Retain failed records, add the provided label to the sample FT field.

Utility

-a

--add-header=STRING|FILE

File containing VCF header lines to add, or a literal header line. May be specified 0 or more times.

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--no-header

Prevent VCF header from being written.

Usage:

Use vcffilter to get a subset of the results from variant calling based on the filtering criteria supplied by the filter flags. Multiple criteria can be specified at once, and advanced processing can be specified via JavaScript scripting.

When filtering on multiple samples, if any of the specified samples fail the criteria, the record will be filtered. The default behavior is for filtered records to be excluded from output altogether, but alternatively the records can be retained but with an additional user-specified VCF FILTER status set via --fail option, or if sample-specific filtering criteria is being applied, only those samples can be filtered either by setting their GT field to missing by using the --clear-failed-samples option, or by setting the FORMAT FT field with a user-specified status via the --fail-samples option.

The --bed-regions option makes use of tabix indexes to avoid loading VCF records outside the supplied regions, which can give faster filtering performance. If the input VCF is not indexed or being read from standard input, or if records failing filters are to be annotated via the --fail option, use the --include-bed option instead.

The flags --min-denovo-score and --max-denovo-score can only be used on a single sample. Records will only be kept if the specified sample is flagged as a de novo variant and the score is within the range specified by the flags. It will also only be kept if none of the other samples for the record are also flagged as a de novo variant within the specified score range.

The --add-header option allows inserting arbitrary VCF header lines into the output VCF. For more information, see vcfannotate.

A powerful general-purpose filtering capability has been included that permits the specification of filter criteria as simple JavaScript expressions (--keep-expr) or more comprehensive JavaScript processing functions (--javascript). Both --keep-expr and --javascript can take JavaScript on the command line or if a filename is supplied then the script/expression will be read from that file. --keep-expr will be applied before --javascript, so the --javascript record function will not be called for records filtered out by --keep-expr.

See also

For full details of functions available in --keep-expr and --javascript see RTG JavaScript filtering API

Simple filtering by JavaScript expression with --keep-expr

The --keep-expr flag aims to provide a convenient way to apply some simple (typically one line) filtering expressions which are evaluated in the context of each record. The final expression of the fragment must evaluate to a boolean value. Records which evaluate to true will be retained, while false will be removed. The value must be of type boolean, simply being truthy/falsy (in the JavaScript sense) will raise an error.

--keep-expr examples:

The following expression keeps records where the NA12878 sample has GQ > 30 and the total depth is > 20. JavaScript will auto convert numerical strings when comparing a string with a number, so calls to parseInt can be omitted.

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz \
--keep-expr "'NA12878'.GQ > 30 && INFO.DP > 20"

If the field of interest may contain the missing value (‘.’) or may be entirely missing on a per-record basis, the has() function can be used to control whether such records are kept vs filtered. For example, to keep records with depth greater than 20, and remove any without a DP annotation:

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz \
--keep-expr "has(INFO.DP) && INFO.DP > 20"

Alternatively, to keep records with depth greater than 20, as well as those without a DP annotation:

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz \
--keep-expr "!has(INFO.DP) || INFO.DP > 20"

The next example keeps records where all samples have a depth > 10. The standard JavaScript array methods every and some can be used to apply a condition on every sample column.

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz \
  --keep-expr "SAMPLES.every(function(s) {return s.DP > 10})"

Similarly, the following example retains records where the FILTER field is unset, or if set must be either PASS or MED_QUAL:

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz \
  --keep-expr "FILTER.every(function(f) {return f == 'PASS' || f == 'MED_QUAL'})"

Note that multi-valued INFO and FORMAT fields are not split into sub-values, so in some cases correct filtering may require splitting the values first. For example, to select bi-allelic records with AF greater than 0.1, the following simple selection will work:

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz \
  --keep-expr "INFO.AF>=0.1"

However, in the presence of multi-allelic records, something like the following is required:

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz \
  --keep-expr "INFO.AF.split(',').some(function(af) {return af >= 0.1})"

Advanced JavaScript filtering with --javascript

The --javascript option aims to support more complicated processing than --keep-expr, permitting modification of the output VCF, or supporting use cases where the script is tasked to compute and output alternative information in addition to (or instead of) the output VCF. The scripts specified by the user are evaluated once at the start of processing. Two special functions may be defined in a --javascript script, which will then be executed in different contexts:

  • A function with the name record will be executed once for each VCF record. If the record function has a return value it must have type boolean. Records which evaluate to true will be retained, while false will be removed. If the record function has no return value then the record will be retained. The record function is applied after any --keep-expr expression.

  • A function with the name end will be called once at the end of processing. This allows reporting of summary statistics collected during the filter process.

This --javascript flag may be specified multiple times, they will be evaluated in order, in a shared JavaScript namespace, before VCF processing commences. This permits a use case where an initial JavaScript expression supplies parameter values which will be required by a subsequent JavaScript file.

Example --javascript scripts:

To find indels with length greater than 5, save the following to a file named find-indels.js:

// Finds indels with length > 5
function record() {
  var deltas = ALT.map(function (alt) {
    return Math.abs(alt.length - REF.length);
  });
  return deltas.some(function (delta) {return delta > 5});
}

Then perform the filtering via:

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz --javascript find-indels.js

The following example derives a new FORMAT column containing variant allelic fraction to two decimal places based on the values in the AD and DP FORMAT annotations, for every sample contained in the VCF. Save the following to a file named add-vaf.js:

// Derive new VAF FORMAT field for each sample
ensureFormatHeader('##FORMAT=<ID=VAF,Number=1,Type=Float,' +
  'Description="Variant Allelic Fraction">');

function record() {
  SAMPLES.forEach(function(sample) {
    // Take all but the first AD value as numerics
    var altDepths = sample.AD.split(",").slice(1);
    // Find the max
    var maxAltDepth = Math.max.apply(null, altDepths);
    if (maxAltDepth > 0) {
      sample.VAF = (maxAltDepth / sample.DP).toFixed(2);
    }
  });
}

Then run the filtering via:

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz --javascript add-vaf.js

The next example produces a table of binned indel lengths, save the following to a file named indel-lengths.js:

// bin breakpoints can be customised by defining your own bins[] in a
// previous -j flag
if (typeof bins == "undefined") {
  var bins = [-10, -5, -3, 0, 4, 6, 11];
}

var counts = [0];
bins.forEach(function () {counts.push(0)});
function record() {
  if (ALT.length == 0) {
    return false;
  }
  var deltas = ALT.map(function (alt) { return alt.length - REF.length; });
  var maxDel = Math.min.apply(null, deltas);
  var maxIns = Math.max.apply(null, deltas);
  var delta = Math.abs(maxDel) > maxIns ? maxDel : maxIns;

  if (delta == 0) {
    return false;
  }
  for (var i = 0; i < bins.length; i++) {
    if (delta < bins[i]) {
      counts[i]++;
      break;
    }
  }
  if (delta > bins[bins.length - 1]) {
    counts[counts.length - 1]++;
  }
  return false;
}

function end() {
  print("Delta\\tCount");
  for (var i = 0; i < bins.length; i++) {
    print("<" + bins[i] + "\\t" + counts[i]);
  }
  print(">" + bins[bins.length - 1] + "\\t" + counts[counts.length - 1]);
}

Then run the filtering via:

$ rtg vcffilter -i in.vcf.gz -o out.vcf.gz --javascript indel-lengths.js

We could use this same script with adjusted bins and omitting the output of the VCF via:

$ rtg vcffilter -i in.vcf.gz -j "var bins = [-20, -10, 0, 20, 20];" \
  -j indel-lengths.js

See also

For full details of functions available in --keep-expr and --javascript see RTG JavaScript filtering API

vcfmerge

Synopsis:

Combines the contents of two or more VCF files. The vcfmerge command can concatenate the outputs of per-chromosome variant detection runs to create a complete genome VCF file, and also merge VCF outputs from multiple samples to form a multi-sample VCF file.

Syntax:

$ rtg vcfmerge [OPTION]... -o FILE FILE+

Example:

$ rtg vcfmerge -o merged.vcf.gz snp1.vcf.gz snp2.vcf.gz

Parameters:

File Input/Output

--bed-regions=FILE

If set, only read VCF records that overlap the ranges contained in the specified BED file.

-I

--input-list-file=FILE

File containing a list of VCF format files (1 per line) to be merged.

-o

--output=FILE

Output VCF file. Use ‘-‘ to write to standard output.

--region=REGION

If set, only read VCF records within the specified range. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>

FILE+

Input VCF files to merge. May be specified 0 or more times.

Utility

-a

--add-header=STRING|FILE

File containing VCF header lines to add, or a literal header line. May be specified 0 or more times.

-f

--force-merge=STRING

Allow merging of specified header ID even when descriptions do not match. May be specified 0 or more times, or as a comma separated list.

-F

--force-merge-all

Attempt merging of all non-matching header declarations.

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--no-header

Prevent VCF header from being written.

--no-merge-alts

Do not merge multiple records if the ALTs are different.

--no-merge-records

Do not merge multiple records at the same position into one.

--preserve-formats

Do not merge multiple records containing unmergeable FORMAT fields (Default is to remove those FORMAT fields so the variants can be combined)

--stats

Output statistics for the merged VCF file.

Usage:

The vcfmerge command takes a list of VCF files and outputs to a single VCF file. Each VCF file must be block compressed and have a corresponding tabix index file, which is the default for outputs from RTG variant detection tools, but may also be created from an existing VCF file using the RTG bgzip and index commands.

There are two primary usage scenarios for the vcfmerge command. The first is to combine input VCFs corresponding to different genomic regions (for example, if variant calling was carried out for each chromosome independently on different nodes of a compute cluster). The second scenario is when combining VCFs containing variant calls for different samples (e.g. combining calls made for separate cohorts into a single VCF).

The input files must have consistent header lines, although specific similar header lines can be forced to merge using the --force-merge parameter, or inconsistent header checking can be entirely disabled with --force-merge-all.

The --add-header option allows inserting arbitrary VCF header lines into the output VCF. For more information, see vcfannotate.

Merging records at the same position

When multiple records occur with the same position and length on the reference, vcfmerge will attempt to combine the records into a single record. Combining multiple fully annotated records is non-trivial, and can lead to loss of information depending on the annotations present. The default behavior takes a pragmatic approach to merging, with options to adjust the merging behavior as required. Multi-record merging can be disabled entirely by setting --no-merge-records.

The first point to note is that the QUAL, FILTER, INFO fields are taken from the first record only (those values from other records at the position are ignored).

If the combined record would result in a change in the set of ALT alleles, any VCF INFO or FORMAT fields declared to be of type A, G, or R cannot meaningfully be retained. By default such fields will be removed or set to the missing value (.). (Other annotations may also become semantically meaningless, but it isn’t possible to tell in general.)

Similarly, if multiple input records with the same position and length on the reference contain information for the same sample, only the information from the first record will be retained.

These behaviors can be altered with additional flags: --no-merge-alts will simply prevent record merging if the ALTs are not the same across the records; --preserve-formats will attempt merging as long as the records do not contain problematic A, G, or R annotations, or contain the same sample.

vcfsplit

Synopsis:

Splits samples contained within a VCF into separate files, one per sample.

Syntax:

$ rtg vcfsplit [OPTION]... -i FILE -o DIR

Example:

$ rtg vcfsplit --keep-sample NA12878,NA12891,NA12892 -i population-ceph-calls.vcf.gz -o trio-vcfs

Parameters:

File Input/Output

--bed-regions=FILE

If set, only read VCF records that overlap the ranges contained in the specified BED file.

-i

--input=FILE

The input VCF, or ‘-‘ to read from standard input.

-o

--output=DIR

Directory for output.

--region=REGION

If set, only read VCF records within the specified range. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>

Filtering

--keep-ref

Keep records where the sample is reference.

--keep-sample=STRING|FILE

File containing sample IDs to select, or a literal sample name. May be specified 0 or more times, or as a comma separated list.

--remove-sample=STRING|FILE

File containing sample IDs to ignore, or a literal sample name. May be specified 0 or more times, or as a comma separated list.

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

Usage:

The vcfsplit command allows producing separate single-sample VCF files from a single multi-sample VCF, and is more efficient than running vcfsubset separately for each required sample, particularly when the input VCF contains many samples, as the input VCF only needs to be read once.

The output VCFs are all written into the specified output directory, as sample_name.vcf.gz. For any particular output VCF, only records where the sample has a non-reference genotype will be output, unless --keep-ref is used.

See also

vcfsubset

vcfstats

Synopsis:

Display simple statistics about the contents of a set of VCF files.

Syntax:

$ rtg vcfstats [OPTION]... FILE+

Example:

$ rtg vcfstats /data/human/wgs/NA19240/snp_chr5.vcf.gz

Location                      : /data/human/wgs/NA19240/snp_chr5.vcf.gz
Passed Filters                : 283144
Failed Filters                : 83568
SNPs                          : 241595
MNPs                          : 5654
Insertions                    : 15424
Deletions                     : 14667
Indels                        : 1477
Unchanged                     : 4327
SNP Transitions/Transversions : 1.93 (210572/108835)
Total Het/Hom ratio           : 2.13 (189645/89172)
SNP Het/Hom ratio             : 2.12 (164111/77484)
MNP Het/Hom ratio             : 3.72 (4457/1197)
Insertion Het/Hom ratio       : 1.69 (9695/5729)
Deletion Het/Hom ratio        : 2.33 (10263/4404)
Indel Het/Hom ratio           : 3.13 (1119/358)
Insertion/Deletion ratio      : 1.05 (15424/14667)
Indel/SNP+MNP ratio           : 0.13 (31568/247249)

Parameters:

File Input/Output

--known

Set to only calculate statistics for known variants.

--novel

Set to only calculate statistics for novel variants.

--sample=FILE

Set to only calculate statistics for the specified sample. (Default is to include all samples). May be specified 0 or more times.

FILE+

VCF file from which to derive statistics. Use ‘-‘ to read from standard input. Must be specified 1 or more times.

Reporting

--allele-lengths

Set to output variant length histogram.

Utility

-h

--help

Prints help on command-line flag usage.

Usage:

Use the vcfstats command to display summary statistics for a set of VCF files. If a VCF file contains multiple sample columns, the statistics for each sample are shown individually.

When determining the categorization of a REF to ALT transformation, some normalization is carried out to ignore same as reference bases at the start and end of the alleles. Thus the following REF to ALT transformations are categorized as SNPs:

A    -> G      (simple case)
ATGC -> ATGG   (leading bases match)
ATGC -> ACGC   (leading and trailing bases match)

Cases where multiple bases change, but the lengths of the two alleles do not are considered to be MNPs:

ATGC -> TTGG   (two bases change)
ATGC -> GTCT   (three bases change)

Cases where there is pure addition or removal of bases are classified as Insertions or Deletions respectively:

A     -> AT      (one base insertion)
ATT   -> ATTTT   (two base insertion)
AT    -> A       (one base deletion)
ATTTT -> ATT     (two base deletion)

The remaining case is there there is a length change between the REF and ALT, but it is not pure. These are called Indels:

ATT   -> CTTT    (one base changed, one base inserted)
CTTT  -> ATT     (one base changed, one base deleted)

In the per-sample summary output of vcfstats, each genotype is classified as a whole into one of the above categories, preferring the more complex of the transformations when ploidy is greater than one.

When computing the per-sample variant length histograms, note that the histograms are incremented for each called allele (thus a diploid homozygous call will increment the appropriate cell by two), and the length of an indel is taken as the change in length rather than the overall length.

vcfsubset

Synopsis:

Create a VCF file containing a subset of the original columns.

Syntax:

$ rtg vcfsubset [OPTION]... -i FILE -o FILE

Example:

$ rtg vcfsubset -i snps.vcf.gz -o frequency.vcf.gz --keep-info AF --remove-samples

Parameters:

File Input/Output

--bed-regions=FILE

If set, only read VCF records that overlap the ranges contained in the specified BED file.

-i

--input=FILE

VCF file containing variants to manipulate. Use ‘-‘ to read from standard input.

-o

--output=FILE

Output VCF file. Use ‘-‘ to write to standard output.

--region=REGION

If set, only read VCF records within the specified range. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>

Filtering

--keep-filter=STRING

Keep the specified FILTER tag. May be specified 0 or more times, or as a comma separated list.

--keep-format=STRING

Keep the specified FORMAT field. May be specified 0 or more times, or as a comma separated list.

--keep-info=STRING

Keep the specified INFO tag. May be specified 0 or more times, or as a comma separated list.

--keep-sample=STRING|FILE

File containing sample IDs to keep, or a literal sample name. May be specified 0 or more times, or as a comma separated list.

--remove-filter=STRING

Remove the specified FILTER tag. May be specified 0 or more times, or as a comma separated list.

--remove-filters

Remove all FILTER tags.

--remove-format=STRING

Remove the specified FORMAT field. May be specified 0 or more times, or as a comma separated list.

--remove-ids

Remove the contents of the ID field.

--remove-info=STRING

Remove the specified INFO tag. May be specified 0 or more times, or as a comma separated list.

--remove-infos

Remove all INFO tags.

--remove-qual

Remove the QUAL field.

--remove-sample=STRING|FILE

File containing sample IDs to remove, or a literal sample name. May be specified 0 or more times, or as a comma separated list.

--remove-samples

Remove all samples.

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--no-header

Prevent VCF header from being written.

Usage:

Use the vcfsubset command to produce a smaller copy of an original VCF file containing only the columns and information desired. For example, to produce a VCF containing only the information for one sample from a multiple sample VCF file use the --keep-sample flag to specify the sample to keep. The various --keep and --remove options can either be specified multiple times or with comma separated lists, for example, --keep-format GT --keep-format DP is equivalent to –keep-format GT,DP.

vcf2rocplot

Synopsis:

Produce rocplot compatible ROC data files from vcfeval annotated VCFs. The primary use cases for this command are to produce aggregate ROCs from several independent vcfeval evaluation runs, and to generate ROCs with respect to different criteria than were used during the initial vcfeval evaluation.

Syntax:

$ rtg vcf2rocplot [OPTION]... -o DIR FILE+

Create an aggregate ROC from evaluations of several independent samples:

$ rtg vcfeval -b goldstandard.vcf.gz -c snps.vcf.gz -t HUMAN_reference \
  --sample daughter1 -f AVR -o eval -m annotate
$ rtg vcfeval -b goldstandard.vcf.gz -c snps.vcf.gz -t HUMAN_reference \
  --sample daughter2 -f AVR -o eval -m annotate
$ rtg vcfeval -b goldstandard.vcf.gz -c snps.vcf.gz -t HUMAN_reference \
  --sample son2 -f AVR -o eval -m annotate
$ rtg vcf2rocplot -o eval_family -f AVR \
  eval_{daughter1,daughter2,son1}/{baseline,calls}.vcf.gz
$ rtg rocplot eval_family/weighted_roc.tsv.gz \
  --png eval_family_roc.png

Parameters:

File Input/Output

--bed-regions=FILE

If set, only read VCF records that overlap the ranges contained in the specified BED file.

-o

--output=DIR

Directory for output.

--region=REGION

If set, only read VCF records within the specified range. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>

FILE+

Input VCF files containing vcfeval annotations. Must be specified 1 or more times.

Reporting

--at-precision=FLOAT

Output summary statistics where precision >= supplied value (Default is to summarize at maximum F-measure)

--at-sensitivity=FLOAT

Output summary statistics where sensitivity >= supplied value (Default is to summarize at maximum F-measure)

--roc-expr=STRING

Output ROC file for variants matching custom JavaScript expression. Use the form <LABEL>=<EXPRESSION>. May be specified 0 or more times.

--roc-regions=STRING

Output ROC file for variants overlapping custom regions supplied in BED file. Use the form <LABEL>=<FILENAME>. May be specified 0 or more times.

--roc-subset=STRING

Output ROC file for preset variant subset. Allowed values are [hom, het, snp, non-snp, mnp, indel]. May be specified 0 or more times, or as a comma separated list.

-O

--sort-order=STRING

The order in which to sort the ROC scores so that good scores come before bad scores. Allowed values are [ascending, descending] (Default is descending)

-f

--vcf-score-field=STRING

The name of the VCF FORMAT field to use as the ROC score. Also valid are QUAL, INFO.<name> or FORMAT.<name> to select the named VCF FORMAT or INFO field (Default is GQ)

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

-T

--threads=INT

Number of threads (Default is the number of available cores)

Usage:

While the ROC outputs generated by vcfeval are sufficient for many scenarios, there are some situations where it is useful to regenerate ROC files from one or more existing vcfeval outputs.

One such situation is when evaluating caller accuracy across a cohort of samples where the number of variants per individual sample is low. Individual sample ROC curves are fairly uninformative with regard to overall accuracy or where to place suitable filter thresholds. An ROC curve from the combined evaluations provides a much better indication of overall caller accuracy.

Another use case for generating ROC files from an existing vcfeval run is to look at alternative scoring methods or stratifications without having to execute the time-consuming variant matching stage implied by a new vcfeval run.

The prerequisite for using vcf2rocplot is that vcfeval must have been run in “annotation” mode (i.e. via --output-mode=annotate) so that the output VCF files contain sufficient annotations regarding variant classification status. It is important that vcf2rocplot be provided with both annotated baseline and call VCFs in order to produce correct outputs.

The operation of the various score field selection and ROC stratification flags works the same as when running vcfeval.

See also

rocplot, vcfeval

svdecompose

Synopsis:

Split composite structural variants into a breakend representation.

Syntax:

$ rtg svdecompose [OPTION]... -i FILE -o FILE

Parameters:

File Input/Output

-i

--input=FILE

VCF file containing variants to filter. Use ‘-‘ to read from standard input.

-o

--output=FILE

Output VCF file name. Use ‘-‘ to write to standard output.

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

--no-header

Prevent VCF header from being written.

Usage:

The svdecompose command is applied to a VCF containing structural variants and converts deletion, insertion, inversion, and tandem duplications with SVTYPE of DEL, INS, INV, and DUP, respectively, into corresponding breakend events with SVTYPE=BND. svdecompose will also decompose sequence-resolved insertions and deletions greater than --min-indel-length into breakend representation. Records of others types are passed through without modification.

This operation can be useful for the purposes of reducing output from various structural variant callers to a common representation to better facilitate comparison with the bndeval command.

For insertions, svdecompose will represent the insertion as breakends between the reference and a “virtual haplotype”, where for example, contig “<INS_A>” represents the destination of all insertions made on chromosome A. So if another caller produced a similar insertion event (in position and/or length), the break end versions will also be nearby on the virtual contig. For the following insertions:

1  54712   .     T  TTTTTTTTTCTTTCTTTCTTTCTTTCTTTCTTTCTTTCTTTCTTTCTTTCTTTC  .  .  .
1  934144  I_7   C  CGGAGGGGAGGGCGCGGAGCGGAGG                               .  .  .
1  934144  I_22  C  CGGAGGGGAGGGCGCGGAGCGGAGGGGAGGGCGCGGAGCGGAGG            .  .  .

each insertion gets two breakends like this:

1  54712   .     T  T[<INS_1>:54712[   .  .  SVTYPE=BND;CIPOS=0,0
1  54712   .     T  ]<INS_1>:54765]C   .  .  SVTYPE=BND;CIPOS=0,0
1  934144  I_7   C  C[<INS_1>:934144[  .  .  SVTYPE=BND;CIPOS=0,0
1  934144  I_22  C  C[<INS_1>:934144[  .  .  SVTYPE=BND;CIPOS=0,0
1  934144  I_7   C  ]<INS_1>:934168]G  .  .  SVTYPE=BND;CIPOS=0,0
1  934144  I_22  C  ]<INS_1>:934187]G  .  .  SVTYPE=BND;CIPOS=0,0

See also

bndeval

bndeval

Synopsis:

Evaluate called breakends for agreement with a baseline breakend set. Outputs a weighted ROC file which can be viewed with rtg rocplot and VCF files containing false positives (called breakends not matched in the baseline), false negatives (baseline breakends not matched in the call set), and true positives (breakends that match between the baseline and calls).

Syntax:

$ rtg bndeval [OPTION]... -b FILE -c FILE -o DIR

Parameters:

File Input/Output

-b

--baseline=FILE

VCF file containing baseline variants.

--bed-regions=FILE

If set, only read VCF records that overlap the ranges contained in the specified BED file.

-c

--calls=FILE

VCF file containing called variants.

-o

--output=DIR

Directory for output.

--region=REGION

If set, only read VCF records within the specified range. The format is one of <sequence_name>, <sequence_name>:<start>-<end>, <sequence_name>:<pos>+<length> or <sequence_name>:<pos>~<padding>

Filtering

--all-records

Use all records regardless of FILTER status (Default is to only process records where FILTER is “.” or “PASS”)

--bidirectional

If set, allow matches between flipped breakends.

--tolerance=INT

Positional tolerance for breakend matching (Default is 100)

Reporting

--no-roc

Do not produce ROCs.

-m

--output-mode=STRING

Output reporting mode. Allowed values are [split, annotate] (Default is split)

-O

--sort-order=STRING

The order in which to sort the ROC scores so that good scores come before bad scores. Allowed values are [ascending, descending] (Default is descending)

-f

--vcf-score-field=STRING

The name of the VCF FORMAT field to use as the ROC score. Also valid are QUAL, INFO.<name> or FORMAT.<name> to select the named VCF FORMAT or INFO field (Default is INFO.DP)

Utility

-h

--help

Print help on command-line flag usage.

-Z

--no-gzip

Do not gzip the output.

Usage:

The bndeval command operates on VCF files containing breakends such as those produced by the discord command. In particular, it considers records having the breakend structural variant type (SVTYPE=BND) as defined in the VCF specification. Other types of record are ignored, but the svdecompose command can be applied beforehand to split certain other structural variants (e.g., INV and DEL) or sequence-resolved insertions and deletions into constituent breakend events.

The input and output requirements of bndeval are broadly similar to the vcfeval command. The primary inputs to bndeval are a truth/baseline VCF containing expected breakends, and a query/call VCF containing the called breakends. Evaluation can be restricted to particular regions by specifying a BED file.

The regions contained in the evaluation regions BED file are intersected with the breakend records contained in the truth VCF in order to obtain a list of truth breakend regions. An evaluation region is included if there is any overlapping truth VCF record (no attempt is made to look at the degree of overlap). Thus by supplying either evaluation regions corresponding to targeted regions or larger gene-level regions bndeval can be used to evaluate at different levels of granularity.

Similarly, the evaluation regions are intersected with the breakend records records contained in the calls VCF to obtain called breakend regions.

The truth breakend regions are then intersected with the called breakend regions to obtain TP/FP/FN metrics. The intersection supports a user-selectable tolerance in position. Further, be default, a breakend must occur in the same orientation to be considered a match, but this constraint can be relaxed by supplying the --bidirectional command line option.

bndeval outputs

Once complete, bndeval command produces summary statistics and the following primary result files in the output directory:

  • weighted_roc.tsv.gz - contains ROC data that can be plotted with rocplot

  • baseline.bed.gz contains the truth breakend regions, where each BED record contains the region status as TP or FN, the SVTYPE, and the span of the original truth VCF record.

  • calls.bed.gz contains the called breakend regions, where each BED record contains the region status as TP or FP, the SVTYPE, the span of the original calls VCF record, and the score value used for ranking in the ROC plot.

  • summary.txt contains the same summary statistics printed to standard output.

pedfilter

Synopsis:

Filter and convert a pedigree file.

Syntax:

$ rtg pedfilter [OPTION]... FILE

Example:

$ rtg pedfilter --remove-parentage mypedigree.ped

Parameters:

File Input/Output

FILE

The pedigree file to process, may be PED or VCF, use ‘-‘ to read from stdin.

Filtering

--keep-family=STRING

Keep only individuals with the specified family ID. May be specified 0 or more times, or as a comma separated list.

--keep-ids=STRING

Keep only individuals with the specified ID. May be specified 0 or more times, or as a comma separated list.

--keep-primary

Keep only primary individuals (those with a PED individual line / VCF sample column)

--remove-parentage

Remove all parent-child relationship information.

Reporting

--vcf

Output pedigree in in the form of a VCF header rather than PED.

Utility

-h

--help

Print help on command-line flag usage.

Usage:

The pedfilter command can be used to perform manipulations on pedigree information and convert pedigree information between PED and VCF header format. For more information about the PED file format see Pedigree PED input file format.

The VCF files output by the family and population commands contain full pedigree information represented as VCF header lines, and the pedfilter command allows this information to be extracted in PED format.

This command produces the pedigree output on standard output, which can be redirected to a file or another pipeline command as required.

pedstats

Synopsis:

Output information from pedigree files of various formats.

Syntax:

$ rtg pedstats [OPTION]... FILE

Example:

For a summary of pedigree information:

$ rtg pedstats ceph_pedigree.ped

Pedigree file: /data/ceph/ceph_pedigree.ped

Total samples:               17
Primary samples:             17
Male samples:                 9
Female samples:               8
Afflicted samples:            0
Founder samples:              4
Parent-child relationships:  26
Other relationships:          0
Families:                     3

To output a list of all founders:

$ rtg pedstats --founder-ids ceph_pedigree.ped
NA12889
NA12890
NA12891
NA12892

For quick pedigree visualization using Graphviz and ImageMagick, use a command-line such as:

$ dot -Tpng <(rtg pedstats --dot "A Title" mypedigree.ped) | display -

Parameters:

File Input/Output

FILE

The pedigree file to process, may be PED or VCF, use ‘-‘ to read from stdin.

Reporting

-d

--delimiter=STRING

Output id lists using this separator (Default is \n)

--dot=STRING

Output pedigree in Graphviz format, using the supplied text as a title.

--families

Output information about family structures.

--female-ids

Output ids of all females.

--founder-ids

Output ids of all founders.

--male-ids

Output ids of all males.

--maternal-ids

Output ids of maternal individuals.

--paternal-ids

Output ids of paternal individuals.

--primary-ids

Output ids of all primary individuals.

--simple-dot

When outputting Graphviz format, use a layout that looks less like a traditional pedigree diagram but works better with large complex pedigrees.

Utility

-h

--help

Print help on command-line flag usage.

Usage:

This command is used to show pedigree summary statistics or select groups of individual IDs.

When using pedstats to output a list of sample IDs, the default is to print one ID per line. Depending on subsequent use, it may be convenient to use a different separator between output IDs. For example, with comma separated output it is possible to directly use the results as an argument to vcfsubset:

$ rtg vcfsubset -i pedigree-calls.vcf.gz -o family1.vcf.gz \
    --keep-samples <(rtg pedstats -d , --founder-ids ceph_pedigree.ped)

In addition, pedstats can be used to generate a simple pedigree visualization, using the well-known Graphviz graphics drawing package, which can be saved to PNG or PDF. For example, with the following chinese-trio.ped:

#PED format pedigree
#
#fam-id/ind-id/pat-id/mat-id: 0=unknown
#sex: 1=male; 2=female; 0=unknown
#phenotype: -9=missing, 0=missing; 1=unaffected; 2=affected
#
#fam-id ind-id  pat-id  mat-id  sex     phen
0       NA24631 NA24694 NA24695 1       0
0       NA24694 0       0       1       0
0       NA24695 0       0       2       0

We can visualize the pedigree with:

$ dot -Tpng <(rtg pedstats --dot "Chinese Trio" chinese-trio.ped) -o chinese-trio.png

This will create a PNG image that can be displayed in any image viewing tool and contains the pedigree structure as shown below.

_images/chinese-trio.png

For more information about the PED file format see Pedigree PED input file format.

The VCF files output by the RTG pedigree-aware variant calling commands contain full pedigree information represented as VCF header lines, and the pedstats command can also take these VCFs as input. For example, given a VCF produced by the population command after calling the CEPH-1463 pedigree:

$ dot -Tpng <(rtg pedstats --dot "CEPH 1463" population-ceph-calls.vcf.gz) -o ceph-1463.png

Would produce the following pedigree directly from the VCF:

_images/ceph-1463.png

Note

Graphviz is provided directly via many operating system package managers, and can also be downloaded from their web site: https://www.graphviz.org/

rocplot

Synopsis:

Plot ROC curves from readsimeval and vcfeval ROC data files, either to an image, or using an interactive GUI.

Syntax:

$ rtg rocplot [OPTION]... FILE+
$ rtg rocplot [OPTION]... --curve STRING

Example:

$ rtg rocplot eval/weighted_roc.tsv.gz

Parameters:

File Input/Output

--curve=STRING

ROC data file with title optionally specified (path[=title]). May be specified 0 or more times.

--png=FILE

If set, output a PNG image to the given file.

--svg=FILE

If set, output a SVG image to the given file.

--zoom=STRING

Show a zoomed view with the given coordinates, supplied in the form <xmax>,<ymax> or <xmin>,<ymin>,<xmax>,<ymax>

FILE+

ROC data file. May be specified 0 or more times.

Reporting

--cmd=FILE

If set, print rocplot command used in previously saved image.

--hide-sidepane

If set, hide the side pane from the GUI on startup.

--interpolate

If set, interpolate curves at regular intervals.

--line-width=INT

Sets the plot line width (Default is 2)

--palette=STRING

Name of color palette to use. Allowed values are [blind_13, blind_15, blind_8, brewer_accent, brewer_dark2, brewer_paired, brewer_pastel1, brewer_pastel2, brewer_set1, brewer_set2, brewer_set3, classic] (Default is classic)

--plain

If set, use a plain plot style.

-P

--precision-sensitivity

If set, plot precision vs sensitivity rather than ROC.

--scores

If set, show scores on the plot.

-t

--title=STRING

Title for the plot.

Utility

-h

--help

Print help on command-line flag usage.

Usage:

Used to produce ROC plots from the ROC files produced by readsimeval, bndeval and vcfeval. By default this opens the ROC plots in an interactive viewer. On a system with only console access the plot can be saved directly to an image file using the either the --png or --svg parameter.

ROC data files may be specified either as direct file arguments to the command, or via the --curve flag. The former method is useful when selecting files using shell wild card globbing, and the latter method allows specifying a custom title for each curve, so use whichever method is most convenient.

Strictly speaking, a true ROC curve should use rates rather than absolute numbers on the X and Y axes (e.g. True Positive / Total Positives rather than True Positives on the Y, and False Positive / Total Negatives on the X axis). However, there are a couple of difficulties involved with computing these rates with variant calling datasets. Firstly, the truth sets do not include any indication of the set of negatives (the closest we may get is in the cases of truth sets which contain a set of confidence regions, where it can be assumed that no other variants may be present inside the specified regions); secondly even with knowledge of negative regions, how do you count the set of possible negative calls, when a call could occupy multiple reference bases, or even (in the case of insertions) zero reference bases. It is conceptually even possible to have a call-set contain more false positives than there are reference bases. For this reason the ROC curves are plotted using the absolute counts.

Precision/sensitivity (also known as precision/recall) curves are another popular means of visualizing call-set accuracy, and these metrics also do not require a count of Total Negatives and so cause no particular difficulty to plot. Precision/sensitivity graphs can be selected from the command line via the --precision-sensitivity flag, or may be interactively selected in the GUI.

An interesting result of ROC analysis is that although there may be few data points on an ROC graph, it is possible to construct a filtered dataset corresponding to any point that lies on a straight line between two points on the graph. (For example, using threshold A for 25% of the variants and threshold B for 75% of the variants would result in accuracy that is 75% of the way between the points corresponding to thresholds A and B on the ROC plot). So in a sense it is meaningful to connect points on an ROC graph with straight lines. However, for precision/sensitivity graphs, it’s incorrect to connect adjacent points with a straight line, as this does not correspond to achievable accuracy on the ROC convex hull and can over-estimate the accuracy. Instead, we can plot appropriately interpolated values with the --interpolate option.

The default ROC graphs include some flourishes such as background gradients and axes drop shadows, these can be disabled via the --plain parameter. Alternative preset color palettes for the curve colors may be selected with the --palette parameter, and in particular some palettes are more color-blind friendly than the default palette. In addition, PNG images saved by rocplot include metadata indicating the graph configuration that may be useful when recreating graphs. This metadata can be displayed (when present) via the --cmd parameter.

Interactive GUI

The following image shows the rocplot GUI with an example ROC plot :

_images/rocplot_roc.png

Similarly, here is an example precision/sensitivity plot:

_images/rocplot_ps.png

Some quick tips for the interactive GUI:

  • Select regions within the graph to zoom in. Right click within the graph area to bring up a context menu that allows undoing the zoom one level at a time, or resetting the zoom to the default.

  • The graph right click menu also allows exporting the image as PNG or SVG. (The saved image does not include the RTG banner).

  • Click on a spot in the graph to show the equivalent accuracy metrics for that location in the status bar. Clicking to the left or below the axes will remove the cross-hair. Note that sensitivity depends on the baseline total number of variants being correct. If for example the ROC curve corresponds to evaluating an exome call-set against a whole-genome baseline, this number will be inaccurate.

  • A secondary cross-hair is also available by holding down shift when placing (or removing) the cross-hair. When two cross-hairs are placed or moved, metrics in the status bar indicate the difference between the two positions.

  • Additional ROC data files can be loaded by clicking on the “Open…” button, and multiple ROC data files within a directory can be loaded at once using multi-select. Alternatively, you may use Drag and Drop from your file browser to drop ROC data files into either the graph area or the right hand ROC curve widget area.

  • The “Cmd” button will open a message window that contains a command-line equivalent to the currently displayed set of curves. This command-line may be copy-pasted, providing an easy way to replicate the current set of curves in another session, generate a curve in a script, or share with a colleague.

  • There is a drop down that allows for switching between ROC and precision/sensitivity graph types.

Each curve in the GUI has a customization widget on the right hand side of the window that allows several operations:

  • Rename the title used for the curve via the editable text.

  • Temporarily hide/show the curve via selection checkbox.

  • Reorder curves via drag and drop using the colored handle on the left.

  • Right click within the ROC widget area to bring up a context menu that allows permanently removing that curve, or customizing the color used for the curve

  • Each curve has a slider to simulate the effect of applying a threshold on the scoring attribute. If the “show scores” option is set, this provides an easy way to select appropriate filter threshold values, which you might apply to variant sets using rtg vcffilter or similar VCF filtering tools.

Note

For definitions of the terminology used when evaluating caller accuracy, see: https://en.wikipedia.org/wiki/Receiver_operating_characteristic and https://en.wikipedia.org/wiki/Sensitivity_and_specificity

Note

For a description of the precision/sensitivity interpolation, see: “The relationship between Precision-Recall and ROC curves”, Davis, J., (2006), https://dx.doi.org/10.1145/1143844.1143874

version

Synopsis:

The RTG version display utility.

Syntax:

$ rtg version

Example:

$ rtg version

Product: RTG Core 3.9
Core Version: 718f8317b7 (2018-05-29)
RAM: 25.0GB of 31.3GB RAM can be used by rtg (79%)
CPU: Defaulting to 4 of 4 available processors (100%)
JVM: Java HotSpot(TM) 64-Bit Server VM 1.8.0_161
License: Expires on 2019-05-20
Contact: support@realtimegenomics.com

Patents / Patents pending:
US: 7,640,256, 9,165,253, 13/129,329, 13/681,046, 13/681,215, 13/848,653, 13/925,704, 14/015,295, 13/971,654, 13/971,630, 14/564,810
UK: 1222923.3, 1222921.7, 1304502.6, 1311209.9, 1314888.7, 1314908.3
New Zealand: 626777, 626783, 615491, 614897, 614560
Australia: 2005255348, Singapore: 128254

Citation (variant calling):
John G. Cleary, Ross Braithwaite, Kurt Gaastra, Brian S. Hilbush, Stuart Inglis, Sean A. Irvine, Alan Jackson, Richard Littin, Sahar Nohzadeh-Malakshah, Mehul Rathod, David Ware, Len Trigg, and Francisco M. De La Vega. "Joint Variant and De Novo Mutation Identification on Pedigrees from High-Throughput Sequencing Data." Journal of Computational Biology. June 2014, 21(6): 405-419. doi:10.1089/cmb.2014.0029.

Citation (vcfeval):
John G. Cleary, Ross Braithwaite, Kurt Gaastra, Brian S. Hilbush, Stuart Inglis, Sean A. Irvine, Alan Jackson, Richard Littin, Mehul Rathod, David Ware, Justin M. Zook, Len Trigg, and Francisco M. De La Vega. "Comparing Variant Call Files for Performance Benchmarking of Next-Generation Sequencing Variant Calling Pipelines." bioRxiv, 2015. doi:10.1101/023754.

(c) Real Time Genomics, 2017

Parameters:

There are no options associated with the version command.

Usage:

Use the version command to display release and version information.

See also

help, license

license

Synopsis:

The RTG license display utility.

Syntax:

$ rtg license

Example:

$ rtg license

Parameters:

There are no options associated with the license command.

Usage:

Use the license command to display license information and expiration date. Output at the command line (standard output) shows command name, licensed status, and command release level.

See also

help, version

help

Synopsis:

The RTG help command provides online help for all RTG commands.

Syntax:

List all commands:

$ rtg help

Show usage syntax and flags for one command:

$ rtg help COMMAND

Example:

$ rtg help format

Parameters:

There are no options associated with the help command.

Usage:

Use the help command to view syntax and usage information for the main rtg command as well as individual RTG commands.

See also

license, version