There are 17 comments on this article. To see them or add your own, read this post on the forum

IMPORTANT NOTE: Our testing has shown that not all combinations of snpEff/database versions produce high-quality results. Be sure to read this document completely to familiarize yourself with our recommended best practices BEFORE running snpEff.

Introduction

Until recently we were using an in-house annotation tool for genomic annotation, but the burden of keeping the database current and our lack of ability to annotate indels has led us to employ the use of a third-party tool instead. After reviewing many external tools (including annoVar, VAT, and Oncotator), we decided that SnpEff best meets our needs as it accepts VCF files as input, can annotate a full exome callset (including indels) in seconds, and provides continually-updated transcript databases. We have implemented support in the GATK for parsing the output from the SnpEff tool and annotating VCFs with the information provided in it.

SnpEff Setup and Usage

Download the SnpEff core program. If you want to be able to run VariantAnnotator on the SnpEff output, you'll need to download a version of SnpEff that VariantAnnotator supports from this page (currently supported versions are listed below). If you just want the most recent version of SnpEff and don't plan to run VariantAnnotator on its output, you can get it from here.

After unzipping the core program, open the file snpEff.config in a text editor, and change the "database_repository" line to the following:

database_repository = http://sourceforge.net/projects/snpeff/files/databases/

Then, download one or more databases using SnpEff's built-in download command:

java -jar snpEff.jar download GRCh37.64

You can find a list of available databases here. The human genome databases have GRCh or hg in their names. You can also download the databases directly from the SnpEff website, if you prefer.

The download command by default puts the databases into a subdirectory called data within the directory containing the SnpEff jar file. If you want the databases in a different directory, you'll need to edit the data_dir entry in the file snpEff.config to point to the correct directory.

Run SnpEff on the file containing your variants, and redirect its output to a file. SnpEff supports many input file formats including VCF 4.1, BED, and SAM pileup. Full details and command-line options can be found on the SnpEff home page.

Supported SnpEff Versions

If you want to take advantage of SnpEff integration in the GATK, you'll need to run SnpEff version **2.0.5*. Note: newer versions are currently unsupported by the GATK, as we haven't yet had the reources to test it.

Current Recommended Best Practices When Running SnpEff

These best practices are based on our analysis of various snpEff/database versions as described in detail in the Analysis of SnpEff Annotations Across Versions section below.

• We recommend using only the GRCh37.64 database with SnpEff 2.0.5. The more recent GRCh37.65 database produces many false-positive Missense annotations due to a regression in the ENSEMBL Release 65 GTF file used to build the database. This regression has been acknowledged by ENSEMBL and is supposedly fixed as of 1-30-2012; however as we have not yet tested the fixed version of the database we continue to recommend using only GRCh37.64 for now.

• We recommend always running with -onlyCoding true with human databases (eg., the GRCh37.* databases). Setting -onlyCoding false causes snpEff to report all transcripts as if they were coding (even if they're not), which can lead to nonsensical results. The -onlyCoding false option should only be used with databases that lack protein coding information.

• Do not trust annotations from versions of snpEff prior to 2.0.4. Older versions of snpEff (such as 2.0.2) produced many incorrect annotations due to the presence of a certain number of nonsensical transcripts in the underlying ENSEMBL databases. Newer versions of snpEff filter out such transcripts.

Analyses of SnpEff Annotations Across Versions

See our analysis of the SNP annotations produced by snpEff across various snpEff/database versions here.

• Both snpEff 2.0.2 + GRCh37.63 and snpEff 2.0.5 + GRCh37.65 produce an abnormally high Missense:Silent ratio, with elevated levels of Missense mutations across the entire spectrum of allele counts. They also have a relatively low (~70%) level of concordance with the 1000G Gencode annotations when it comes to Silent mutations. This suggests that these combinations of snpEff/database versions incorrectly annotate many Silent mutations as Missense.

• snpEff 2.0.4 RC3 + GRCh37.64 and snpEff 2.0.5 + GRCh37.64 produce a Missense:Silent ratio in line with expectations, and have a very high (~97%-99%) level of concordance with the 1000G Gencode annotations across all categories.

See our comparison of SNP annotations produced using the GRCh37.64 and GRCh37.65 databases with snpEff 2.0.5 here

• The GRCh37.64 database gives good results on the condition that you run snpEff with the -onlyCoding true option. The -onlyCoding false option causes snpEff to mark all transcripts as coding, and so produces many false-positive Missense annotations.

• The GRCh37.65 database gives results that are as poor as those you get with the -onlyCoding false option on the GRCh37.64 database. This is due to a regression in the ENSEMBL release 65 GTF file used to build snpEff's GRCh37.65 database. The regression has been acknowledged by ENSEMBL and is due to be fixed shortly.

See our analysis of the INDEL annotations produced by snpEff across snpEff/database versions here

• snpEff's indel annotations are highly concordant with those of a high-quality set of genomic annotations from the 1000 Genomes project. This is true across all snpEff/database versions tested.

Example SnpEff Usage with a VCF Input File

Below is an example of how to run SnpEff version 2.0.5 with a VCF input file and have it write its output in VCF format as well. Notice that you need to explicitly specify the database you want to use (in this case, GRCh37.64). This database must be present in a directory of the same name within the data_dir as defined in snpEff.config.

java -Xmx4G -jar snpEff.jar eff -v -onlyCoding true -i vcf -o vcf GRCh37.64 1000G.exomes.vcf > snpEff_output.vcf

In this mode, SnpEff aggregates all effects associated with each variant record together into a single INFO field annotation with the key EFF. The general format is:

EFF=Effect1(Information about Effect1),Effect2(Information about Effect2),etc.

And here is the precise layout with all the subfields:

EFF=Effect1(Effect_Impact|Effect_Functional_Class|Codon_Change|Amino_Acid_Change|Gene_Name|Gene_BioType|Coding|Transcript_ID|Exon_ID),Effect2(etc...

It's also possible to get SnpEff to output in a (non-VCF) text format with one Effect per line. See the SnpEff home page for full details.

Adding SnpEff Annotations using VariantAnnotator

Once you have a SnpEff output VCF file, you can use the VariantAnnotator walker to add SnpEff annotations based on that output to the input file you ran SnpEff on.

There are two different options for doing this:

Option 1: Annotate with only the highest-impact effect for each variant

NOTE: This option works only with supported SnpEff versions as explained above. VariantAnnotator run as described below will refuse to parse SnpEff output files produced by other versions of the tool, or which lack a SnpEff version number in their header.

The default behavior when you run VariantAnnotator on a SnpEff output file is to parse the complete set of effects resulting from the current variant, select the most biologically-significant effect, and add annotations for just that effect to the INFO field of the VCF record for the current variant. This is the mode we plan to use in our Production Data-Processing Pipeline.

When selecting the most biologically-significant effect associated with the current variant, VariantAnnotator does the following:

• Prioritizes the effects according to the categories (in order of decreasing precedence) "High-Impact", "Moderate-Impact", "Low-Impact", and "Modifier", and always selects one of the effects from the highest-priority category. For example, if there are three moderate-impact effects and two high-impact effects resulting from the current variant, the annotator will choose one of the high-impact effects and add annotations based on it. See below for a full list of the effects arranged by category.

• Within each category, ties are broken using the functional class of each effect (in order of precedence: NONSENSE, MISSENSE, SILENT, or NONE). For example, if there is both a NON_SYNONYMOUS_CODING (MODERATE-impact, MISSENSE) and a CODON_CHANGE (MODERATE-impact, NONE) effect associated with the current variant, the annotator will select the NON_SYNONYMOUS_CODING effect. This is to allow for more accurate counts of the total number of sites with NONSENSE/MISSENSE/SILENT mutations. See below for a description of the functional classes SnpEff associates with the various effects.

• Effects that are within a non-coding region are always considered lower-impact than effects that are within a coding region.

Example Usage:

java -jar dist/GenomeAnalysisTK.jar \
     -T VariantAnnotator \
     -R /humgen/1kg/reference/human_g1k_v37.fasta \
     -A SnpEff \       
     --variant 1000G.exomes.vcf \        (file to annotate)
     --snpEffFile snpEff_output.vcf \    (SnpEff VCF output file generated by running SnpEff on the file to annotate)
     -L 1000G.exomes.vcf \
     -o out.vcf

VariantAnnotator adds some or all of the following INFO field annotations to each variant record:

• SNPEFF_EFFECT - The highest-impact effect resulting from the current variant (or one of the highest-impact effects, if there is a tie)
• SNPEFF_IMPACT - Impact of the highest-impact effect resulting from the current variant (HIGH, MODERATE, LOW, or MODIFIER)
• SNPEFF_FUNCTIONAL_CLASS - Functional class of the highest-impact effect resulting from the current variant (NONE, SILENT, MISSENSE, or NONSENSE)
• SNPEFF_CODON_CHANGE - Old/New codon for the highest-impact effect resulting from the current variant
• SNPEFF_AMINO_ACID_CHANGE - Old/New amino acid for the highest-impact effect resulting from the current variant
• SNPEFF_GENE_NAME - Gene name for the highest-impact effect resulting from the current variant
• SNPEFF_GENE_BIOTYPE - Gene biotype for the highest-impact effect resulting from the current variant
• SNPEFF_TRANSCRIPT_ID - Transcript ID for the highest-impact effect resulting from the current variant
• SNPEFF_EXON_ID - Exon ID for the highest-impact effect resulting from the current variant

Example VCF records annotated using SnpEff and VariantAnnotator:

1   874779  .   C   T   279.94  . AC=1;AF=0.0032;AN=310;BaseQRankSum=-1.800;DP=3371;Dels=0.00;FS=0.000;HRun=0;HaplotypeScore=1.4493;InbreedingCoeff=-0.0045;
MQ=54.49;MQ0=10;MQRankSum=0.982;QD=13.33;ReadPosRankSum=-0.060;SB=-120.09;SNPEFF_AMINO_ACID_CHANGE=G215;SNPEFF_CODON_CHANGE=ggC/ggT;
SNPEFF_EFFECT=SYNONYMOUS_CODING;SNPEFF_EXON_ID=exon_1_874655_874840;SNPEFF_FUNCTIONAL_CLASS=SILENT;SNPEFF_GENE_BIOTYPE=protein_coding;SNPEFF_GENE_NAME=SAMD11;
SNPEFF_IMPACT=LOW;SNPEFF_TRANSCRIPT_ID=ENST00000342066

1   874816  .   C   CT  2527.52 .   AC=15;AF=0.0484;AN=310;BaseQRankSum=-11.876;DP=4718;FS=48.575;HRun=1;HaplotypeScore=91.9147;InbreedingCoeff=-0.0520;
MQ=53.37;MQ0=6;MQRankSum=-1.388;QD=5.92;ReadPosRankSum=-1.932;SB=-741.06;SNPEFF_EFFECT=FRAME_SHIFT;SNPEFF_EXON_ID=exon_1_874655_874840;
SNPEFF_FUNCTIONAL_CLASS=NONE;SNPEFF_GENE_BIOTYPE=protein_coding;SNPEFF_GENE_NAME=SAMD11;SNPEFF_IMPACT=HIGH;SNPEFF_TRANSCRIPT_ID=ENST00000342066

Option 2: Annotate with all effects for each variant

VariantAnnotator also has the ability to take the EFF field from the SnpEff VCF output file containing all the effects aggregated together and copy it verbatim into the VCF to annotate.

Here's an example of how to do this:

java -jar dist/GenomeAnalysisTK.jar \
     -T VariantAnnotator \
     -R /humgen/1kg/reference/human_g1k_v37.fasta \      
     -E resource.EFF \
     --variant 1000G.exomes.vcf \      (file to annotate)
     --resource snpEff_output.vcf \    (SnpEff VCF output file generated by running SnpEff on the file to annotate)
     -L 1000G.exomes.vcf \
     -o out.vcf

Of course, in this case you can also use the VCF output by SnpEff directly, but if you are using VariantAnnotator for other purposes anyway the above might be useful.

List of Genomic Effects

Below are the possible genomic effects recognized by SnpEff, grouped by biological impact. Full descriptions of each effect are available on this page.

High-Impact Effects

• SPLICE_SITE_ACCEPTOR
• SPLICE_SITE_DONOR
• START_LOST
• EXON_DELETED
• FRAME_SHIFT
• STOP_GAINED
• STOP_LOST

Moderate-Impact Effects

• NON_SYNONYMOUS_CODING
• CODON_CHANGE (note: this effect is used by SnpEff only for MNPs, not SNPs)
• CODON_INSERTION
• CODON_CHANGE_PLUS_CODON_INSERTION
• CODON_DELETION
• CODON_CHANGE_PLUS_CODON_DELETION
• UTR_5_DELETED
• UTR_3_DELETED

Low-Impact Effects

• SYNONYMOUS_START
• NON_SYNONYMOUS_START
• START_GAINED
• SYNONYMOUS_CODING
• SYNONYMOUS_STOP
• NON_SYNONYMOUS_STOP

Modifiers

• NONE
• CHROMOSOME
• CUSTOM
• CDS
• GENE
• TRANSCRIPT
• EXON
• INTRON_CONSERVED
• UTR_5_PRIME
• UTR_3_PRIME
• DOWNSTREAM
• INTRAGENIC
• INTERGENIC
• INTERGENIC_CONSERVED
• UPSTREAM
• REGULATION
• INTRON

Functional Classes

SnpEff assigns a functional class to certain effects, in addition to an impact:

• NONSENSE: assigned to point mutations that result in the creation of a new stop codon
• MISSENSE: assigned to point mutations that result in an amino acid change, but not a new stop codon
• SILENT: assigned to point mutations that result in a codon change, but not an amino acid change or new stop codon
• NONE: assigned to all effects that don't fall into any of the above categories (including all events larger than a point mutation)

The GATK prioritizes effects with functional classes over effects of equal impact that lack a functional class when selecting the most significant effect in VariantAnnotator. This is to enable accurate counts of NONSENSE/MISSENSE/SILENT sites.

There are 17 comments on this article. To see them or add your own, read this post on the forum

There are 6 comments on this article. To see them or add your own, read this post on the forum

Introduction

BEAGLE is a state of the art software package for analysis of large-scale genetic data sets with hundreds of thousands of markers genotyped on thousands of samples. BEAGLE can

• phase genotype data (i.e. infer haplotypes) for unrelated individuals, parent-offspring pairs, and parent-offspring trios.
• infer sporadic missing genotype data.
• impute ungenotyped markers that have been genotyped in a reference panel.
• perform single marker and haplotypic association analysis.
• detect genetic regions that are homozygous-by-descent in an individual or identical-by-descent in pairs of individuals.

The GATK provides and experimental interface to BEAGLE. Currently, the only use cases supported by this interface are a) inferring missing genotype data from call sets (e.g. for lack of coverage in low-pass data), b) Genotype inference for unrelated individuals.

The basic workflow for this interface is as follows:

After variants are called and possibly filtered, the GATK walker ProduceBeagleInput will take the resulting VCF as input, and will produce a likelihood file in BEAGLE format.

• User needs to run BEAGLE with this likelihood file specified as input.
• After Beagle runs, user must unzip resulting output files (.gprobs, .phased) containing posterior genotype probabilities and phased haplotypes.
• User can then run GATK walker BeagleOutputToVCF to produce a new VCF with updated data. The new VCF will contain updated genotypes as well as updated annotations.

Producing Beagle input likelihoods file

Before running BEAGLE, we need to first take an input VCF file with genotype likelihoods and produce the BEAGLE likelihoods file using walker ProduceBealgeInput, as described in detail in its documentation page.

For each variant in inputvcf.vcf, ProduceBeagleInput will extract the genotype likelihoods, convert from log to linear space, and produce a BEAGLE input file in Genotype likelihoods file format (see BEAGLE documentation for more details). Essentially, this file is a text file in tabular format, a snippet of which is pasted below:

marker    alleleA alleleB NA07056 NA07056 NA07056 NA11892 NA11892 NA11892 
20:60251    T        C     10.00   1.26    0.00     9.77   2.45    0.00 
20:60321    G        T     10.00   5.01    0.01    10.00   0.31    0.00 
20:60467    G        C      9.55   2.40    0.00     9.55   1.20    0.00 

Note that BEAGLE only supports biallelic sites. Markers can have an arbitrary label, but they need to be in chromosomal order. Sites that are not genotyped in the input VCF (i.e. which are annotated with a "./." string and have no Genotype Likelihood annotation) are assigned a likelihood value of (0.33, 0.33, 0.33).

IMPORTANT: Due to BEAGLE memory restrictions, it's strongly recommended that BEAGLE be run on a separate chromosome-by-chromosome basis. In the current use case, BEAGLE uses RAM in a manner approximately proportional to the number of input markers. After BEAGLE is run and an output VCF is produced as described below, CombineVariants can be used to combine resulting VCF's, using the "-variantMergeOptions UNION" argument.

Running Beagle

We currently only support a subset of BEAGLE functionality - only unphased, unrelated input likelihood data is supported. To run imputation analysis, run for example

java -Xmx4000m -jar path_to_beagle/beagle.jar like=path_to_beagle_output/beagle_output out=myrun

Extra BEAGLE arguments can be added as required.

About Beagle memory usage

Empirically, Beagle can run up to about ~800,000 markers with 4 GB of RAM. Larger chromosomes require additional memory.

Processing BEAGLE output files

BEAGLE will produce several output files. The following shell commands unzip the output files in preparation for their being processed, and put them all in the same place:

# unzip gzip'd files, force overwrite if existing
gunzip -f path_to_beagle_output/myrun.beagle_output.gprobs.gz
gunzip -f path_to_beagle_output/myrun.beagle_output.phased.gz
#rename also Beagle likelihood file to mantain consistency
mv path_to_beagle_output/beagle_output path_to_beagle_output/myrun.beagle_output.like 

Creating a new VCF from BEAGLE data with BeagleOutputToVCF

Once BEAGLE files are produced, we can update our original VCF with BEAGLE's data. This is done using the BeagleOutputToVCF tool.

The walker looks for the files specified with the -B(type,BEAGLE,file) triplets as above for the output posterior genotype probabilities, the output r^2 values and the output phased genotypes. The order in which these are given in the command line is arbitrary, but all three must be present for correct operation.

The output VCF has the new genotypes that Beagle produced, and several annotations are also updated. By default, the walker will update the per-genotype annotations GQ (Genotype Quality), the genotypes themselves, as well as the per-site annotations AF (Allele Frequency), AC (Allele Count) and AN (Allele Number).

The resulting VCF can now be used for further downstream analysis.

Merging VCFs broken up by chromosome into a single genome-wide file

Assuming you have broken up your calls into Beagle by chromosome (as recommended above), you can use the CombineVariants tool to merge the resulting VCFs into a single callset.

java -jar /path/to/dist/GenomeAnalysisTK.jar \
  -T CombineVariants \
  -R reffile.fasta \
  --out genome_wide_output.vcf \
  -V:input1 beagle_output_chr1.vcf \
  -V:input2 beagle_output_chr2.vcf \
  .
  .
  .
  -V:inputX beagle_output_chrX.vcf \
  -type UNION -priority input1,input2,...,inputX

There are 6 comments on this article. To see them or add your own, read this post on the forum

There are 25 comments on this article. To see them or add your own, read this post on the forum

liftOverVCF.pl

Introduction

This script converts a VCF file from one reference build to another. It runs 3 modules within our toolkit that are necessary for lifting over a VCF.
1. LiftoverVariants walker
2. sortByRef.pl to sort the lifted-over file
3. Filter out records whose ref field no longer matches the new reference

Obtaining the Script

The liftOverVCF.pl script is available in our public source repository under the 'perl' directory. Instructions for pulling down our source are available here.

Example

./liftOverVCF.pl -vcf calls.b36.vcf \
  -chain b36ToHg19.broad.over.chain \
  -out calls.hg19.vcf \
  -gatk /humgen/gsa-scr1/ebanks/Sting_dev
  -newRef /seq/references/Homo_sapiens_assembly19/v0/Homo_sapiens_assembly19
  -oldRef /humgen/1kg/reference/human_b36_both
  -tmp /broad/shptmp [defaults to /tmp]

Usage

Running the script with no arguments will show the usage:

Usage: liftOverVCF.pl
    -vcf        <input vcf>
    -gatk       <path to gatk trunk>
    -chain      <chain file>
    -newRef     <path to new reference prefix; we will need newRef.dict, .fasta, and .fasta.fai>
    -oldRef     <path to old reference prefix; we will need oldRef.fasta>
    -out        <output vcf>
    -tmp            <temp file location; defaults to /tmp>

• The 'tmp' argument is optional. It specifies the location to write the temporary file from step 1 of the process.

Chain files

Chain files from b36/hg18 to hg19 are located here within the Broad:

   /humgen/gsa-hpprojects/GATK/data/Liftover_Chain_Files/

External users can get them off our ftp site:

   location: ftp.broadinstitute.org
   username: gsapubftp-anonymous
   path:     Liftover_Chain_Files

There are 25 comments on this article. To see them or add your own, read this post on the forum

There are no comments on this article yet. To add one, read this post on the forum

Introduction

Three-stage procedure:

• Create a master set of sites from your N batch VCFs that you want to genotype in all samples. At this stage you need to determine how you want to resolve disagreements among the VCFs. This is your master sites VCF.

• Take the master sites VCF and genotype each sample BAM file at these sites

• (Optionally) Merge the single sample VCFs into a master VCF file

Creating the master set of sites: SNPs and Indels

The first step of batch merging is to create a master set of sites that you want to genotype in all samples. To make this problem concrete, suppose I have two VCF files:

Batch 1:

##fileformat=VCFv4.0
#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  NA12891 
20      9999996     .       A       ATC     .       PASS    .       GT:GQ   0/1:30
20      10000000        .       T       G       .       PASS    .       GT:GQ   0/1:30
20      10000117        .       C       T       .       FAIL    .       GT:GQ   0/1:30
20      10000211        .       C       T       .       PASS    .       GT:GQ   0/1:30
20      10001436        .       A       AGG     .       PASS    .       GT:GQ   1/1:30

Batch 2:

##fileformat=VCFv4.0
#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  NA12878
20      9999996     .       A       ATC     .       PASS    .       GT:GQ   0/1:30
20      10000117        .       C       T       .       FAIL    .       GT:GQ   0/1:30
20      10000211        .       C       T       .       FAIL    .       GT:GQ   0/1:30
20      10000598        .       T       A       .       PASS    .       GT:GQ   1/1:30
20      10001436        .       A       AGGCT   .       PASS    .       GT:GQ   1/1:30

In order to merge these batches, I need to make a variety of bookkeeping and filtering decisions, as outlined in the merged VCF below:

Master VCF:

20      9999996     .       A       ATC     .       PASS    .       GT:GQ   0/1:30  [pass in both]
20      10000000        .       T       G       .       PASS    .       GT:GQ   0/1:30  [only in batch 1]
20      10000117        .       C       T       .       FAIL    .       GT:GQ   0/1:30  [fail in both]
20      10000211        .       C       T       .       FAIL    .       GT:GQ   0/1:30  [pass in 1, fail in 2, choice in unclear]
20      10000598        .       T       A       .       PASS    .       GT:GQ   1/1:30  [only in batch 2]
20      10001436        .       A       AGGCT   .       PASS    .       GT:GQ   1/1:30  [A/AGG in batch 1, A/AGGCT in batch 2, including this site may be problematic]

These issues fall into the following categories:

• For sites present in all VCFs (20:9999996 above), the alleles agree, and each site PASS is pass, this site can obviously be considered "PASS" in the master VCF
• Some sites may be PASS in one batch, but absent in others (20:10000000 and 20:10000598), which occurs when the site is polymorphic in one batch but all samples are reference or no-called in the other batch
• Similarly, sites that are fail in all batches in which they occur can be safely filtered out, or included as failing filters in the master VCF (20:10000117)

There are two difficult situations that must be addressed by the needs of the project merging batches:

• Some sites may be PASS in some batches but FAIL in others. This might indicate that either:
• The site is actually truly polymorphic, but due to limited coverage, poor sequencing, or other issues it is flag as unreliable in some batches. In these cases, it makes sense to include the site
• The site is actually a common machine artifact, but just happened to escape standard filtering in a few batches. In these cases, you would obviously like to filter out the site
• Even more complicated, it is possible that in the PASS batches you have found a reliable allele (C/T, for example) while in others there's no alt allele but actually a low-frequency error, which is flagged as failing. Ideally, here you could filter out the failing allele from the FAIL batches, and keep the pass ones
• Some sites may have multiple segregating alleles in each batch. Such sites are often errors, but in some cases may be actual multi-allelic sites, in particular for indels.

Unfortunately, we cannot determine which is actually the correct choice, especially given the goals of the project. We leave it up the project bioinformatician to handle these cases when creating the master VCF. We are hopeful that at some point in the future we'll have a consensus approach to handle such merging, but until then this will be a manual process.

The GATK tool CombineVariants can be used to merge multiple VCF files, and parameter choices will allow you to handle some of the above issues. With tools like SelectVariants one can slice-and-dice the merged VCFs to handle these complexities as appropriate for your project's needs. For example, the above master merge can be produced with the following CombineVariants:

java -jar dist/GenomeAnalysisTK.jar \
-T CombineVariants \
-R human_g1k_v37.fasta \
-V:one,VCF combine.1.vcf -V:two,VCF combine.2.vcf \
--sites_only \
-minimalVCF \
-o master.vcf

producing the following VCF:

##fileformat=VCFv4.0
#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO
20      9999996     .       A       ACT         .       PASS    set=Intersection
20      10000000        .       T       G           .   PASS    set=one
20      10000117        .       C       T           .       FAIL    set=FilteredInAll
20      10000211        .       C       T           .       PASS    set=filterIntwo-one
20      10000598        .       T       A           .       PASS    set=two
20      10001436        .       A       AGG,AGGCT       .       PASS    set=Intersection

Genotyping your samples at these sites

Having created the master set of sites to genotype, along with their alleles, as in the previous section, you now use the UnifiedGenotyper to genotype each sample independently at the master set of sites. This GENOTYPE_GIVEN_ALLELES mode of the UnifiedGenotyper will jump into the sample BAM file, and calculate the genotype and genotype likelihoods of the sample at the site for each of the genotypes available for the REF and ALT alleles. For example, for site 10000211, the UnifiedGenotyper would evaluate the likelihoods of the CC, CT, and TT genotypes for the sample at this site, choose the most likely configuration, and generate a VCF record containing the genotype call and the likelihoods for the three genotype configurations.

As a concrete example command line, you can genotype the master.vcf file using in the bundle sample NA12878 with the following command:

java -Xmx2g -jar dist/GenomeAnalysisTK.jar \
-T UnifiedGenotyper \
-R bundle/b37/human_g1k_v37.fasta \
-I bundle/b37/NA12878.HiSeq.WGS.bwa.cleaned.recal.hg19.20.bam \
-alleles master.vcf \
-L master.vcf \
-gt_mode GENOTYPE_GIVEN_ALLELES \
-out_mode EMIT_ALL_SITES \
-stand_call_conf 0.0 \
-glm BOTH \
-G none \

The -L master.vcf argument tells the UG to only genotype the sites in the master file. If you don't specify this, the UG will genotype the master sites in GGA mode, but it will also genotype all other sites in the genome in regular mode.

The last item,-G ` prevents the UG from computing annotations you don't need. This command produces something like the following output:

##fileformat=VCFv4.0
#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  NA12878
20      9999996     .       A       ACT         4576.19 .       .   GT:DP:GQ:PL     1/1:76:99:4576,229,0
20      10000000        .       T       G           0       .       .       GT:DP:GQ:PL     0/0:79:99:0,238,3093
20      10000211        .       C       T       857.79  .       .   GT:AD:DP:GQ:PL  0/1:28,27:55:99:888,0,870
20      10000598        .       T       A           1800.57 .       .   GT:AD:DP:GQ:PL  1/1:0,48:48:99:1834,144,0
20      10001436        .       A       AGG,AGGCT       1921.12 .       .   GT:DP:GQ:PL     0/2:49:84.06:1960,2065,0,2695,222,84

Several things should be noted here:

• The genotype likelihoods calculation evolves, especially for indels, the exact results of this command will change.
• The command will emit sites that are hom-ref in the sample at the site, but the -stand_call_conf 0.0 argument should be provided so that they aren't tagged as "LowQual" by the UnifiedGenotyper.
• The filtered site 10000117 in the master.vcf is not genotyped by the UG, as it doesn't pass filters and so is considered bad by the GATK UG. If you want to determine the genotypes for all sites, independent on filtering, you must unfilter all of your records in master.vcf, and if desired, restore the filter string for these records later.

This genotyping command can be performed independently per sample, and so can be parallelized easily on a farm with one job per sample, as in the following:

foreach sample in samples:
  run UnifiedGenotyper command above with -I $sample.bam -o $sample.vcf
end

(Optional) Merging the sample VCFs together

You can use a similar command for CombineVariants above to merge back together all of your single sample genotyping runs. Suppose all of my UnifiedGenotyper jobs have completed, and I have VCF files named sample1.vcf, sample2.vcf, to sampleN.vcf. The single command:

java -jar dist/GenomeAnalysisTK.jar -T CombineVariants -R human_g1k_v37.fasta -V:sample1 sample1.vcf -V:sample2 sample2.vcf [repeat until] -V:sampleN sampleN.vcf -o combined.vcf

General notes

• Because the GATK uses dynamic downsampling of reads, it is possible for truly marginal calls to change likelihoods from discovery (processing the BAM incrementally) vs. genotyping (jumping into the BAM). Consequently, do not be surprised to see minor differences in the genotypes for samples from discovery and genotyping.
• More advanced users may want to consider group several samples together for genotyping. For example, 100 samples could be genotyped in 10 groups of 10 samples, resulting in only 10 VCF files. Merging the 10 VCF files may be faster (or just easier to manage) than 1000 individual VCFs.
• Sometimes, using this method, a monomorphic site within a batch will be identified as polymorphic in one or more samples within that same batch. This is because the UnifiedGenotyper applies a frequency prior to determine whether a site is likely to be monomorphic. If the site is monomorphic, it is either not output, or if EMIT_ALL_SITES is thrown, reference genotypes are output. If the site is determined to be polymorphic, genotypes are assigned greedily (as of GATK-v1.4). Calling single-sample reduces the effect of the prior, so sites which were considered monomorphic within a batch could be considered polymorphic within a sub-batch.

There are no comments on this article yet. To add one, read this post on the forum

There are 6 comments on this article. To see them or add your own, read this post on the forum

1. About CombineVariants

This tool combines VCF records from different sources. Any (unique) name can be used to bind your rod data and any number of sources can be input. This tool currently supports two different combination types for each of variants (the first 8 fields of the VCF) and genotypes (the rest)

For a complete, detailed argument reference, refer to the GATK document page here.

2. Logic for merging records across VCFs

CombineVariants will include a record at every site in all of your input VCF files, and annotate which input ROD bindings the record is present, pass, or filtered in in the set attribute in the INFO field (see below). In effect, CombineVariants always produces a union of the input VCFs. However, any part of the Venn of the N merged VCFs can be exacted using JEXL expressions on the set attribute using SelectVariants. If you want to extract just the records in common between two VCFs, you would first CombineVariants the two files into a single VCF, and then run SelectVariants to extract the common records with -select 'set == "Intersection"', as worked out in the detailed example below.

3. Handling PASS/FAIL records at the same site in multiple input files

The -filteredRecordsMergeType argument determines how CombineVariants handles sites where a record is present in multiple VCFs, but it is filtered in some and unfiltered in others, as described in the Tech Doc page for the tool.

4. Understanding the set attribute

The set INFO field indicates which call set the variant was found in. It can take on a variety of values indicating the exact nature of the overlap between the call sets. Note that the values are generalized for multi-way combinations, but here we describe only the values for 2 call sets being combined.

• set=Intersection : occurred in both call sets, not filtered out

• set=NAME : occurred in the call set NAME only

• set=NAME1-filteredInNAME : occurred in both call sets, but was not filtered in NAME1 but was filtered in NAME2

• set=filteredInAll : occurred in both call sets, but was filtered out of both

For three or more call sets combinations, you can see records like NAME1-NAME2 indicating a variant occurred in both NAME1 and NAME2 but not all sets.

5. Changing the set key

You can use -setKey foo to change the set=XXX tag to foo=XXX in your output. Additionally, -setKey null stops the set tag=value pair from being emitted at all.

6. Minimal VCF output

Add the -minimalVCF argument to CombineVariants if you want to eliminate unnecessary information from the INFO field and genotypes. The only fields emitted will be GT:GQ for genotypes and the keySet for INFO

An even more extreme output format is -sites_only, a general engine capability, where the genotypes for all samples are completely stripped away from the output format. Enabling this option results in a significant performance speedup as well.

7. Combining Variant Calls with a minimum set of input sites

Add the -minN (or --minimumN) command, followed by an integer if you want to only output records present in at least N input files. Useful, for example in combining several data sets where we only want to keep sites present in for example at least 2 of them (in which case -minN 2 should be added to the command line).

8. Example: intersecting two VCFs

In the following example, we use CombineVariants and SelectVariants to obtain only the sites in common between the OMNI 2.5M and HapMap3 sites in the GSA bundle.

java -Xmx2g -jar dist/GenomeAnalysisTK.jar -T CombineVariants -R bundle/b37/human_g1k_v37.fasta -L 1:1-1,000,000 -V:omni bundle/b37/1000G_omni2.5.b37.sites.vcf -V:hm3 bundle/b37/hapmap_3.3.b37.sites.vcf -o union.vcf
java -Xmx2g -jar dist/GenomeAnalysisTK.jar -T SelectVariants -R ~/Desktop/broadLocal/localData/human_g1k_v37.fasta -L 1:1-1,000,000 -V:variant union.vcf -select 'set == ";Intersection";' -o intersect.vcf

This results in two vcf files, which look like:

==> union.vcf <==
1       990839  SNP1-980702     C       T       .       PASS    AC=150;AF=0.05384;AN=2786;CR=100.0;GentrainScore=0.7267;HW=0.0027632264;set=Intersection
1       990882  SNP1-980745     C       T       .       PASS    CR=99.79873;GentrainScore=0.7403;HW=0.005225421;set=omni
1       990984  SNP1-980847     G       A       .       PASS    CR=99.76005;GentrainScore=0.8406;HW=0.26163524;set=omni
1       992265  SNP1-982128     C       T       .       PASS    CR=100.0;GentrainScore=0.7412;HW=0.0025895447;set=omni
1       992819  SNP1-982682     G       A       .       id50    CR=99.72961;GentrainScore=0.8505;HW=4.811053E-17;set=FilteredInAll
1       993987  SNP1-983850     T       C       .       PASS    CR=99.85935;GentrainScore=0.8336;HW=9.959717E-28;set=omni
1       994391  rs2488991       G       T       .       PASS    AC=1936;AF=0.69341;AN=2792;CR=99.89378;GentrainScore=0.7330;HW=1.1741E-41;set=filterInomni-hm3
1       996184  SNP1-986047     G       A       .       PASS    CR=99.932205;GentrainScore=0.8216;HW=3.8830226E-6;set=omni
1       998395  rs7526076       A       G       .       PASS    AC=2234;AF=0.80187;AN=2786;CR=100.0;GentrainScore=0.8758;HW=0.67373306;set=Intersection
1       999649  SNP1-989512     G       A       .       PASS    CR=99.93262;GentrainScore=0.7965;HW=4.9767335E-4;set=omni

==> intersect.vcf <==
1       950243  SNP1-940106     A       C       .       PASS    AC=826;AF=0.29993;AN=2754;CR=97.341675;GentrainScore=0.7311;HW=0.15148845;set=Intersection
1       957640  rs6657048       C       T       .       PASS    AC=127;AF=0.04552;AN=2790;CR=99.86667;GentrainScore=0.6806;HW=2.286109E-4;set=Intersection
1       959842  rs2710888       C       T       .       PASS    AC=654;AF=0.23559;AN=2776;CR=99.849;GentrainScore=0.8072;HW=0.17526293;set=Intersection
1       977780  rs2710875       C       T       .       PASS    AC=1989;AF=0.71341;AN=2788;CR=99.89077;GentrainScore=0.7875;HW=2.9912625E-32;set=Intersection
1       985900  SNP1-975763     C       T       .       PASS    AC=182;AF=0.06528;AN=2788;CR=99.79926;GentrainScore=0.8374;HW=0.017794203;set=Intersection
1       987200  SNP1-977063     C       T       .       PASS    AC=1956;AF=0.70007;AN=2794;CR=99.45917;GentrainScore=0.7914;HW=1.413E-42;set=Intersection
1       987670  SNP1-977533     T       G       .       PASS    AC=2485;AF=0.89196;AN=2786;CR=99.51427;GentrainScore=0.7005;HW=0.24214932;set=Intersection
1       990417  rs2465136       T       C       .       PASS    AC=1113;AF=0.40007;AN=2782;CR=99.7599;GentrainScore=0.8750;HW=8.595538E-5;set=Intersection
1       990839  SNP1-980702     C       T       .       PASS    AC=150;AF=0.05384;AN=2786;CR=100.0;GentrainScore=0.7267;HW=0.0027632264;set=Intersection
1       998395  rs7526076       A       G       .       PASS    AC=2234;AF=0.80187;AN=2786;CR=100.0;GentrainScore=0.8758;HW=0.67373306;set=Intersection

There are 6 comments on this article. To see them or add your own, read this post on the forum

There are 6 comments on this article. To see them or add your own, read this post on the forum

Introduction

SelectVariants is a GATK tool used to subset a VCF file by many arbitrary criteria listed in the command line options below. The output VCF wiil have the AN (number of alleles), AC (allele count), AF (allele frequency), and DP (depth of coverage) annotations updated as necessary to accurately reflect the file's new contents.

Select Variants operates on VCF files (ROD Tracks) provided in the command line using the GATK's built in --variant option. You can provide multiple tracks for Select Variants but at least one must be named 'variant' and this will be the file all your analysis will be based of. Other tracks can be named as you please. Options requiring a reference to a ROD track name will use the track name provided in the -B option to refer to the correct VCF file (e.g. --discordance / --concordance ). All other analysis will be done in the 'variant' track.

Often, a VCF containing many samples and/or variants will need to be subset in order to facilitate certain analyses (e.g. comparing and contrasting cases vs. controls; extracting variant or non-variant loci that meet certain requirements, displaying just a few samples in a browser like IGV, etc.). SelectVariants can be used for this purpose. Given a single VCF file, one or more samples can be extracted from the file (based on a complete sample name or a pattern match). Variants can be further selected by specifying criteria for inclusion, i.e. "DP > 1000" (depth of coverage greater than 1000x), "AF < 0.25" (sites with allele frequency less than 0.25). These JEXL expressions are documented here in the FAQ article on JEXL expressions; it is particularly important to note the section on working with complex expressions.

Command-line arguments

For a complete, detailed argument reference, refer to the GATK document page here.

How do the AC, AF, AN, and DP fields change?

Let's say you have a file with three samples. The numbers before the ":" will be the genotype (0/0 is hom-ref, 0/1 is het, and 1/1 is hom-var), and the number after will be the depth of coverage.

BOB        MARY        LINDA
1/0:20     0/0:30      1/1:50

In this case, the INFO field will say AN=6, AC=3, AF=0.5, and DP=100 (in practice, I think these numbers won't necessarily add up perfectly because of some read filters we apply when calling, but it's approximately right).

Now imagine I only want a file with the samples "BOB" and "MARY". The new file would look like:

BOB        MARY
1/0:20     0/0:30

The INFO field will now have to change to reflect the state of the new data. It will be AN=4, AC=1, AF=0.25, DP=50.

Let's pretend that MARY's genotype wasn't 0/0, but was instead "./." (no genotype could be ascertained). This would look like

BOB        MARY
1/0:20     ./.:.

with AN=2, AC=1, AF=0.5, and DP=20.

Subsetting by sample and ALT alleles

SelectVariants now keeps (r5832) the alt allele, even if a record is AC=0 after subsetting the site down to selected samples. For example, when selecting down to just sample NA12878 from the OMNI VCF in 1000G (1525 samples), the resulting VCF will look like:

1       82154   rs4477212       A       G       .       PASS    AC=0;AF=0.00;AN=2;CR=100.0;DP=0;GentrainScore=0.7826;HW=1.0     GT:GC   0/0:0.7205
1       534247  SNP1-524110     C       T       .       PASS    AC=0;AF=0.00;AN=2;CR=99.93414;DP=0;GentrainScore=0.7423;HW=1.0  GT:GC   0/0:0.6491
1       565286  SNP1-555149     C       T       .       PASS    AC=2;AF=1.00;AN=2;CR=98.8266;DP=0;GentrainScore=0.7029;HW=1.0   GT:GC   1/1:0.3471
1       569624  SNP1-559487     T       C       .       PASS    AC=2;AF=1.00;AN=2;CR=97.8022;DP=0;GentrainScore=0.8070;HW=1.0   GT:GC   1/1:0.3942

Although NA12878 is 0/0 at the first sites, ALT allele is preserved in the VCF record. This is the correct behavior, as reducing samples down shouldn't change the character of the site, only the AC in the subpopulation. This is related to the tricky issue of isPolymorphic() vs. isVariant().

• isVariant => is there an ALT allele?

• isPolymorphic => is some sample non-ref in the samples?

In part this is complicated as the semantics of sites-only VCFs, where ALT = . is used to mean not-polymorphic. Unfortunately, I just don't think there's a consistent convention right now, but it might be worth at some point to adopt a single approach to handling this.

For clarity, in previous versions of SelectVariants, the first two monomorphic sites lose the ALT allele, because NA12878 is hom-ref at this site, resulting in VCF that looks like:

1       82154   rs4477212       A       .       .       PASS    AC=0;AF=0.00;AN=2;CR=100.0;DP=0;GentrainScore=0.7826;HW=1.0     GT:GC   0/0:0.7205
1       534247  SNP1-524110     C       .       .       PASS    AC=0;AF=0.00;AN=2;CR=99.93414;DP=0;GentrainScore=0.7423;HW=1.0  GT:GC   0/0:0.6491
1       565286  SNP1-555149     C       T       .       PASS    AC=2;AF=1.00;AN=2;CR=98.8266;DP=0;GentrainScore=0.7029;HW=1.0   GT:GC   1/1:0.3471
1       569624  SNP1-559487     T       C       .       PASS    AC=2;AF=1.00;AN=2;CR=97.8022;DP=0;GentrainScore=0.8070;HW=1.0   GT:GC   1/1:0.3942

If you really want a VCF without monomorphic sites, use the option to drop monomorphic sites after subsetting.

Known issues

Some VCFs may have repeated header entries with the same key name, for instance:

##fileformat=VCFv3.3
##FILTER=ABFilter,"AB > 0.75"
##FILTER=HRunFilter,"HRun > 3.0"
##FILTER=QDFilter,"QD < 5.0"
##UG_bam_file_used=file1.bam
##UG_bam_file_used=file2.bam
##UG_bam_file_used=file3.bam
##UG_bam_file_used=file4.bam
##UG_bam_file_used=file5.bam
##source=UnifiedGenotyper
##source=VariantFiltration
##source=AnnotateVCFwithMAF
...

Here, the "UG_bam_file_used" and "source" header lines appear multiple times. When SelectVariants is run on such a file, the program will emit warnings that these repeated header lines are being discarded, resulting in only the first instance of such a line being written to the resulting VCF. This behavior is not ideal, but expected under the current architecture.

Additional information

For information on how to construct regular expressions for use with this tool, see the "Summary of regular-expression constructs" section here.

There are 6 comments on this article. To see them or add your own, read this post on the forum

There are 9 comments on this article. To see them or add your own, read this post on the forum

2 SNPs with significant strand bias

Several SNPs with excessive coverage

For a complete, detailed argument reference, refer to the GATK document page here.

Introduction

In addition to true variation, variant callers emit a number of false-positives. Some of these false-positives can be detected and rejected by various statistical tests. VariantAnnotator provides a way of annotating variant calls as preparation for executing these tests.

Description of the haplotype score annotation

Examples of Available Annotations

The list below is not comprehensive. Please use the --list argument to get a list of all possible annotations available. Also, see the FAQ article on understanding the Unified Genotyper's VCF files for a description of some of the more standard annotations.

• BaseQualityRankSumTest (BaseQRankSum)
• DepthOfCoverage (DP)
• FisherStrand (FS)
• HaplotypeScore (HaplotypeScore)
• MappingQualityRankSumTest (MQRankSum)
• MappingQualityZero (MQ0)
• QualByDepth (QD)
• ReadPositionRankSumTest (ReadPosRankSum)
• RMSMappingQuality (MQ)
• SnpEff: Add genomic annotations using the third-party tool SnpEff with VariantAnnotator

Note that technically the VariantAnnotator does not require reads (from a BAM file) to run; if no reads are provided, only those Annotations which don't use reads (e.g. Chromosome Counts) will be added. But most Annotations do require reads. When running the tool we recommend that you add the -L argument with the variant rod to your command line for efficiency and speed.

There are 9 comments on this article. To see them or add your own, read this post on the forum

There are no comments on this article yet. To add one, read this post on the forum

VariantFiltration

For a complete, detailed argument reference, refer to the GATK document page here.

The documentation for Using JEXL expressions within the GATK contains very important information about limitations of the filtering that can be done; in particular please note the section on working with complex expressions.

Filtering Individual Genotypes

One can now filter individual samples/genotypes in a VCF based on information from the FORMAT field: Variant Filtration will add the sample-level FT tag to the FORMAT field of filtered samples (this does not affect the record's FILTER tag). This is still a work in progress and isn't quite as flexible and powerful yet as we'd like it to be. For now, one can filter based on most fields as normal (e.g. GQ < 5.0), but the GT (genotype) field is an exception. We have put in convenience methods so that one can now filter out hets (isHet == 1), refs (isHomRef == 1), or homs (isHomVar == 1).

There are no comments on this article yet. To add one, read this post on the forum

There are 5 comments on this article. To see them or add your own, read this post on the forum

For a complete, detailed argument reference, refer to the technical documentation page.

Modules

You can find detailed information about the various modules here.

Stratification modules

• AlleleFrequency
• AlleleCount
• CompRod
• Contig
• CpG
• Degeneracy
• EvalRod
• Filter
• FunctionalClass
• JexlExpression
• Novelty
• Sample

Evaluation modules

• CompOverlap
• CountVariants

Note that the GenotypeConcordance module has been rewritten as a separate walker tool (see its Technical Documentation page).

A useful analysis using VariantEval

We in GSA often find ourselves performing an analysis of 2 different call sets. For SNPs, we often show the overlap of the sets (their "venn") and the relative dbSNP rates and/or transition-transversion ratios. The picture provided is an example of such a slide and is easy to create using VariantEval. Assuming you have 2 filtered VCF callsets named 'foo.vcf' and 'bar.vcf', there are 2 quick steps.

Combine the VCFs

java -jar GenomeAnalysisTK.jar \
    -R ref.fasta \
    -T CombineVariants \
    -V:FOO foo.vcf \
    -V:BAR bar.vcf \
    -priority FOO,BAR \
    -o merged.vcf

Run VariantEval

java -jar GenomeAnalysisTK.jar \
     -T VariantEval \
     -R ref.fasta \
     -D dbsnp.vcf \
     -select 'set=="Intersection"' -selectName Intersection \
     -select 'set=="FOO"' -selectName FOO \
     -select 'set=="FOO-filterInBAR"' -selectName InFOO-FilteredInBAR \
     -select 'set=="BAR"' -selectName BAR \
     -select 'set=="filterInFOO-BAR"' -selectName InBAR-FilteredInFOO \
     -select 'set=="FilteredInAll"' -selectName FilteredInAll \
     -o merged.eval.gatkreport \
     -eval merged.vcf \
     -l INFO

Checking the possible values of 'set'

It is wise to check the actual values for the set names present in your file before writing complex VariantEval commands. An easy way to do this is to extract the value of the set fields and then reduce that to the unique entries, like so:

java -jar GenomeAnalysisTK.jar -T VariantsToTable -R ref.fasta -V merged.vcf -F set -o fields.txt
grep -v 'set' fields.txt | sort | uniq -c

This will provide you with a list of all of the possible values for 'set' in your VCF so that you can be sure to supply the correct select statements to VariantEval.

Reading the VariantEval output file

The VariantEval output is formatted as a GATKReport.

Understanding Genotype Concordance values from Variant Eval

The VariantEval genotype concordance module emits information the relationship between the eval calls and genotypes and the comp calls and genotypes. The following three slides provide some insight into three key metrics to assess call sensitivity and concordance between genotypes.

##:GATKReport.v0.1 GenotypeConcordance.sampleSummaryStats : the concordance statistics summary for each sample
GenotypeConcordance.sampleSummaryStats  CompRod   CpG      EvalRod  JexlExpression  Novelty  percent_comp_ref_called_var  percent_comp_het_called_het  percent_comp_het_called_var  percent_comp_hom_called_hom  percent_comp_hom_called_var  percent_non-reference_sensitivity  percent_overall_genotype_concordance  percent_non-reference_discrepancy_rate
GenotypeConcordance.sampleSummaryStats  compOMNI  all      eval     none            all      0.78                         97.65                        98.39                        99.13                        99.44                        98.80                              99.09                                 3.60

The key outputs:

• percent_overall_genotype_concordance
• percent_non_ref_sensitivity_rate
• percent_non_ref_discrepancy_rate

All defined below.

There are 5 comments on this article. To see them or add your own, read this post on the forum

There are 46 comments on this article. To see them or add your own, read this post on the forum

Slides that explain the VQSR methodology as well as the individual component variant annotations can be found here in the GSA Public Drop Box.

Detailed information about command line options for VariantRecalibrator can be found here.

Detailed information about command line options for ApplyRecalibration can be found here.

Introduction

The purpose of the variant recalibrator is to assign a well-calibrated probability to each variant call in a call set. One can then create highly accurate call sets by filtering based on this single estimate for the accuracy of each call.

The approach taken by variant quality score recalibration is to develop a continuous, covarying estimate of the relationship between SNP call annotations (QD, SB, HaplotypeScore, HRun, for example) and the the probability that a SNP is a true genetic variant versus a sequencing or data processing artifact. This model is determined adaptively based on "true sites" provided as input, typically HapMap 3 sites and those sites found to be polymorphic on the Omni 2.5M SNP chip array. This adaptive error model can then be applied to both known and novel variation discovered in the call set of interest to evaluate the probability that each call is real. The score that gets added to the INFO field of each variant is called the VQSLOD. It is the log odds ratio of being a true variant versus being false under the trained Gaussian mixture model. The variant recalibrator contrastively evaluates variants in a two step process:

• VariantRecalibration - Create a Gaussian mixture model by looking at the annotations values over a high quality subset of the input call set and then evaluate all input variants.
• ApplyRecalibration - Apply the model parameters to each variant in input VCF files producing a recalibrated VCF file in which each variant is annotated with its VQSLOD value. In addition, this step will filter the calls based on this new lod score by adding lines to the FILTER column for variants that don't meet the lod threshold as provided by the user (with the ts_filter_level parameter).

Recalibration tutorial with example HiSeq, single sample, deep coverage, whole genome call set

By way of explaining how one uses the variant quality score recalibrator and evaluating its performance we have put together this tutorial which uses example sequencing data produced at the Broad Institute. All of the data used in this tutorial is available in VCF format from our GATK resource bundle.

Input call set

input: NA12878.HiSeq.WGS.bwa.cleaned.raw.b37.subset.vcf

• These calls were generated with the UnifiedGenotyper from a 30X coverage modern, single sample run of HiSeq. They were randomly downsampled to keep the file size small but in general one would want to use the full set of variants available genome-wide for this procedure. No other pre-filtering steps were applied to the raw output.

Training sets

HapMap 3.3: hapmap_3.3.b37.sites.vcf

• These high quality sites are used both to train the Gaussian mixture model and then again when choosing a LOD threshold based on sensitivity to truth sites.
• The parameters for these sites will be: known = false, training = true, truth = true, prior = Q15 (96.84%)

Omni 2.5M chip: 1000G_omni2.5.b37.sites.vcf

• These polymorphic sites from the Omni genotyping array are used when training the model.
• The parameters for these sites will be: known = false, training = true, truth = false, prior = Q12 (93.69%)

dbSNP build 132: dbsnp_132.b37.vcf

• The dbsnp sites are generally considered to be not of high enough quality to be used in training but here we stratify output metrics such as ti/tv ratio by presence in dbsnp (known sites) or not (novel sites).
• The parameters for these sites will be: known = true, training = false, truth = false, prior = Q8 (84.15%)

The default prior for all other variants is Q2 (36.90%). This low value reflects the fact that the philosophy of the UnifiedGenotyper is to produce a large, highly sensitive callset that needs to be heavily refined through additional filtering.

VariantRecalibrator

Detailed information about command line options for VariantRecalibrator can be found here.

Build a Gaussian mixture model using a high quality subset of the input variants and evaluate those model parameters over the full call set. The following notes describe the appropriate inputs to use for this tool.

• Note that this walker expects call sets in which each record has been appropriately annotated (see e.g. VariantAnnotator). Input call set rod bindings must start with "input". See the command line below.
• When constructing an initial call set (see e.g. Unified Genotyper or Haplotype Caller) for use with the Recalibrator, it's generally best to turn down the confidence threshold to allow more borderline calls (trusting the Recalibrator to keep the real ones while filtering out the false positives). For example, we often use a Q20 threshold on our deep coverage calls with the Recalibrator (whereas the default threshold in the UnifiedGenotyper is Q30).
• No pre-filtering is necessary when using the Recalibrator. See below for the advanced options which allow the user to selectively ignore certain filters if they have already been applied to your call set.
• The tool accepts any ROD bindings when specifying the set of truth sites to be used during modeling. Information about how to download VCF files which we routinely use for training is in the FAQ section at the bottom of the page.
• Each training set ROD binding is specified with key-value tags to qualify whether the set should be considered as known sites, training sites, and/or truth sites. Additionally, the prior probability of being true for those sites is also specified via these tags in Phred scale. See the command line below for an example. An explanation for how each of the training sets is used by the algorithm:
• Training sites: Input variants which are found to overlap with these training sites are used to build the Gaussian mixture model.
• Truth sites: When deciding where to set the cutoff in VQSLOD sensitivity to these truth sites is used. Typically one might want to say I dropped my threshold until I got back 99% of HapMap sites, for example.
• Known sites: The known / novel status of a variant isn't used by the algorithm itself and is only used for reporting / display purposes. The output metrics are stratified by known status in order to aid in comparisons with other call sets.

Interpretation of the Gaussian mixture model plots

The variant recalibration step fits a Gaussian mixture model to the contextual annotations given to each variant. By fitting this probability model to the training variants (variants considered to be true-positives), a probability can be assigned to the putative novel variants (some of which will be true-positives, some of which will be false-positives). It is useful for users to see how the probability model was fit to their data. Therefore a modeling report is automatically generated each time VariantRecalibrator is run (in the above command line the report will appear as path/to/output.plots.R.pdf). For every pair-wise combination of annotations used in modeling, a 2D projection of the Gaussian mixture model is shown.

Gaussian mixture model report that is automatically generated by the VQSR from the example HiSeq call set. This page shows the 2D projection of mapping quality rank sum test versus Haplotype score by marginalizing over the other annotation dimensions in the model.

In each page there are four panels which show different ways of looking at the 2D projection of the model. The upper left panel shows the probability density function that was fit to the data. The 2D projection was created by marginalizing over the other annotation dimensions in the model via random sampling. Green areas show locations in the space that are indicative of being high quality while red areas show the lowest probability areas. In general putative SNPs that fall in the red regions will be filtered out of the recalibrated call set.

The remaining three panels give scatter plots in which each SNP is plotted in the two annotation dimensions as points in a point cloud. The scale for each dimension is in normalized units. The data for the three panels is the same but the points are colored in different ways to highlight different aspects of the data. In the upper right panel SNPs are colored black and red to show which SNPs are retained and filtered, respectively, by applying the VQSR procedure. The red SNPs didn't meet the given truth sensitivity threshold and so are filtered out of the call set. The lower left panel colors SNPs green, grey, and purple to give a sense of the distribution of the variants used to train the model. The green SNPs are those which were found in the training sets passed into the VariantRecalibrator step, while the purple SNPs are those which were found to be furthest away from the learned Gaussians and thus given the lowest probability of being true. Finally, the lower right panel colors each SNP by their known/novel status with blue being the known SNPs and red being the novel SNPs. Here the idea is to see if the annotation dimensions provide a clear separation between the known SNPs (most of which are true) and the novel SNPs (most of which are false).

An example of good clustering for SNP calls from the tutorial dataset is shown to the right. The plot shows that the training data forms a distinct cluster at low values for each of the two statistics shown (haplotype score and mapping quality bias). As the SNPs fall off the distribution in either one or both of the dimensions they are assigned a lower probability (that is, move into the red region of the model's PDF) and are filtered out. This makes sense as not only do higher values of HaplotypeScore indicate a lower chance of the data being explained by only two haplotypes but also higher values for mapping quality bias indicate more evidence of bias between the reference bases and the alternative bases. The model has captured our intuition that this area of the distribution is highly enriched for machine artifacts and putative variants here should be filtered out!

Tranches and the tranche plot

The recalibrated variant quality score provides a continuous estimate of the probability that each variant is true, allowing one to partition the call sets into quality tranches. The first tranche is exceedingly specific but less sensitive, and each subsequent tranche in turn introduces additional true positive calls along with a growing number of false positive calls. Downstream applications can select in a principled way more specific or more sensitive call sets or incorporate directly the recalibrated quality scores to avoid entirely the need to analyze only a fixed subset of calls but rather weight individual variant calls by their probability of being real. An example tranche plot, automatically generated by the VariantRecalibator walker, is shown on the right.

Tranches plot for example HiSeq call set. The x-axis gives the number of novel variants called while the y-axis shows two quality metrics -- novel transition to transversion ratio and the overall truth sensitivity.

Ti/Tv-free recalibration

We use a Ti/Tv-free approach to variant quality score recalibration. This approach requires an additional truth data set, and cuts the VQSLOD at given sensitivities to the truth set. It has several advantages over the Ti/Tv-targeted approach:

• The truth sensitivity (TS) approach gives you back the novel Ti/Tv as a QC metric [YES!]
• The truth sensitivity (TS) approach is conceptual cleaner than deciding on a novel Ti/Tv target for your dataset
• The TS approach is easier to explain and defend, as saying "I took called variants until I found 99% of my known variable sites" is easier than "I took variants until I dropped my novel Ti/Tv ratio to 2.07"

We have used hapmap 3.3 sites as the truth set (genotypes_r27_nr.b37_fwd.vcf), but other sets of high-quality (~99% truly variable in the population) sets of sites should work just as well. In our experience, with HapMap, 99% is a good threshold, as the remaining 1% of sites often exhibit unusual features like being close to indels or are actually MNPs, and so receive a low VQSLOD score.
Note that the expected Ti/Tv is still an available argument but it is only used for display purposes.

ApplyRecalibration

Detailed information about command line options for ApplyRecalibration can be found here.

Using the tranche file generated by the previous step the ApplyRecalibration walker looks at each variant's VQSLOD value and decides which tranche it falls in. Variants in tranches that fall below the specified truth sensitivity filter level have their filter field annotated with its tranche level. This will result in a call set that simultaneously is filtered to the desired level but also has the information necessary to pull out more variants at a slightly lower quality level.

Frequently Asked Questions

How do I know which annotations to use for my data?

The five annotation values provided in the command lines above (QD, HaplotypeScore, MQRankSum, ReadPosRankSum, and HRun) have been show to give good results for a variety of data types. However this shouldn't be taken to mean these annotations give the absolute best modeling for every source of sequencing data. Better results could possibly be achieved through experimentation with which SNP annotations are used in the algorithm. The goal is to find annotation values with are approximately Gaussianly distributed and also serve to separate the probably true (known) SNPs from the probably false (novel) SNPs.

How do I know which -tranche arguments to pass into the VariantRecalibrator step?

The -tranche arguments main purpose is to create the tranche plot (as shown above). They are meant to convey the idea that with real, calibrated variant quality scores one can create call sets in which each variant doesn't have to have a hard answer as to whether it is in or out of the set. If a very high accuracy call set is desired then one can use the highest tranche, but if a larger, more complete call set is a higher priority than one can dip down into lower and lower tranches. These tranches are applied to the output VCF file using the FILTER field. In this way an end user can choose to use some of the filtered records or only use the PASSing records. For new users to the variant quality score recalibrator perhaps the easiest thing to do in the beginning is simply select the single desired false discovery rate and pass that value in as a single -tranche argument to make sure that the desired rate can be achieved given the other parameters to the algorithm.

What should I use as training data?

The VariantRecalibrator step accept lists of truth and training sites in several formats (dbsnp ROD, VCF, and BED, for example). Any list can be used but it is best to use only those sets which are of the best quality. The truth sets are passed into the algorithm using any rod binding name and their truth or training status is specified with rod tags (see VariantRecalibrator section above). We routinely use the HapMap v3.3 VCF file and the Omni2.5M SNP chip array in training the model. In general the false positive rate of dbsnp sites is too high to be used reliably for training the model.

HapMap v3.3 as well as the Omni validation array VCF files are available in our GATK resource bundle.

Does the VQSR work with non-human variant calls?

Absolutely! The VQSR accepts any list of sites to use as training / truth data, not just HapMap.

Don't have any truth data for your organism? No problem. There are several things one might experiment with. One idea is to first do an initial round of SNP calling and only use those SNPs which have the highest quality scores. These sites which have the most confidence are probably real and could be used as truth data to help disambiguate the rest of the variants in the call set. Another idea is to try using several SNP caller, of which the GATK is one, and use those sites which are concordant between the different methods as truth data. There are many fruitful avenues of research here. Hopefully the model reporting plots help facilitate this experimentation. Perhaps the best place to begin is to use a line like the following when specifying the truth set:

-resource:concordantSet,known=true,training=true,truth=true,prior=10.0 path/to/concordantSet.vcf

Can I use the variant quality score recalibrator with my small sequencing experiment?

This tool is expecting thousands of variant sites in order to achieve decent modeling with the Gaussian mixture model. Whole exome call sets work well, but anything smaller than that scale might run into difficulties. One piece of advice is to turn down the number of Gaussians used during training and to turn up the number of variants that are used to train the negative model. This can be accomplished by adding --maxGaussians 4 --percentBad 0.05 to your command line.

percentBadVariants is the proportion of variants that the program will use to build the negative model. If you have a small callset, you need to use a larger proportion in order to have enough "bad" variants to build the negative model. maxGaussians is the maximum number of different "clusters" (=Gaussians) of variants the program is "allowed" to try to identify. Lowering this number forces the program to group variants into a smaller number of clusters, which means there will be more variants in each cluster -- hopefully enough to satisfy the statistical requirements. Of course, this decreases the level of discrimination that you can achieve between variant profiles/error modes. It's all about trade-offs; and unfortunately if you don't have a lot of variants you can't afford to be very demanding in terms of resolution.

Why don't all the plots get generated for me?

The most common problem related to this is not having Rscript accessible in your environment path. Rscript is the command line version of R that gets installed right alongside. We also make use of the ggplot2 library so please be sure to install that package as well.

There are 46 comments on this article. To see them or add your own, read this post on the forum