Genotype concordance between two callsets
This tool takes in two callsets (vcfs) and tabulates the number of sites which overlap and share alleles, and for each sample, the genotype-by-genotype counts (e.g. the number of sites at which a sample was called homozygous-reference in the EVAL callset, but homozygous-variant in the COMP callset). It outputs these counts as well as convenient proportions (such as the proportion of het calls in the EVAL which were called REF in the COMP) and metrics (such as NRD and NRS).
Genotype concordance requires two callsets (as it does a comparison): an EVAL and a COMP callset, specified via the -eval and -comp arguments. Typically, the EVAL callset is an experimental set you want to evaluate, while the COMP callset is a previously existing set used as a standard for comparison (taken to represent "truth").
(Optional) Jexl expressions for genotype-level filtering of EVAL or COMP genotypes, specified via the -gfe and -cfe arguments, respectively.
Genotype Concordance writes a GATK report to the specified file (via -o), consisting of multiple tables of counts and proportions. These tables are constructed on a per-sample basis, and include counts of EVAL vs COMP genotype states.
Headers for the (non-moltenized -- see below) GenotypeConcordance counts and proportions tables give the genotype of the EVAL callset followed by the genotype of the COMP callset. For example the value corresponding to HOM_REF_HET reflects variants called HOM_REF in the EVAL callset and HET in the COMP callset. Variants for which the alternate alleles between the EVAL and COMP sample did not match are excluded from genotype comparisons and given in the "Mismatching_Alleles" field.
It may be informative to reshape rows of the GenotypeConcordance counts and proportions tables into separate row-major tables where the columns indicate the COMP genotype and the rows indicate the EVAL genotype for easy comparison between the two callsets. This can be done with the gsa.reshape.concordance.table function in the gsalib R library. In Excel this can be accomplished using the OFFSET function.
For strictly bi-allelic VCFs, only the ALLELES_MATCH, EVAL_ONLY, TRUTH_ONLY fields will be populated, but where multi-allelic sites are involved counts for EVAL_SUBSET_TRUTH and EVAL_SUPERSET_TRUTH will be generated.
For example, in the following situation
eval: ref - A alt - C comp: ref - A alt - C,Tthen the site is tabulated as EVAL_SUBSET_TRUTH. Were the situation reversed, it would be EVAL_SUPERSET_TRUTH. However, in the case where EVAL has both C and T alternate alleles, both must be observed in the genotypes (that is, there must be at least one of (0/1,1/1) and at least one of (0/2,1/2,2/2) in the genotype field). If one of the alleles has no observations in the genotype fields of the EVAL, the site-level concordance is tabulated as though that allele were not present in the record.
A site which has an alternate allele, but which is monomorphic in samples, is treated as not having been discovered, and will be recorded in the TRUTH_ONLY column (if a record exists in the COMP set), or not at all (if no record exists in the COMP set).
That is, in the situation
eval: ref - A alt - C genotypes - 0/0 0/0 0/0 ... 0/0 comp: ref - A alt - C ... 0/0 0/0 ...is equivalent to
eval: ref - A alt - . genotypes - 0/0 0/0 0/0 ... 0/0 comp: ref - A alt - C ... 0/0 0/0 ...
When a record is present in the COMP set the *genotypes* for the monomorphic site will still be used to evaluate per-sample genotype concordance counts.
These tables may be optionally moltenized via the -moltenize argument. That is, the standard table
Sample NO_CALL_HOM_REF NO_CALL_HET NO_CALL_HOM_VAR (...) NA12878 0.003 0.001 0.000 (...) NA12891 0.005 0.000 0.000 (...)would instead be displayed
NA12878 NO_CALL_HOM_REF 0.003 NA12878 NO_CALL_HET 0.001 NA12878 NO_CALL_HOM_VAR 0.000 NA12891 NO_CALL_HOM_REF 0.005 NA12891 NO_CALL_HET 0.000 NA12891 NO_CALL_HOM_VAR 0.000 (...)
java -jar GenomeAnalysisTK.jar \ -T GenotypeConcordance \ -R reference.fasta \ -eval test_set.vcf \ -comp truth_set.vcf \ -o output.grp
These Read Filters are automatically applied to the data by the Engine before processing by GenotypeConcordance.
All tools inherit arguments from the GATK Engine' "CommandLineGATK" argument collection, which can be used to modify various aspects of the tool's function. For example, the -L argument directs the GATK engine to restrict processing to specific genomic intervals; or the -rf argument allows you to apply certain read filters to exclude some of the data from the analysis.
This table summarizes the command-line arguments that are specific to this tool. For more details on each argument, see the list further down below the table or click on an argument name to jump directly to that entry in the list.
Argument name(s) | Default value | Summary | |
---|---|---|---|
Required Inputs | |||
--comp |
NA | The variants and genotypes to compare against | |
--eval |
NA | The variants and genotypes to evaluate | |
Required Flags | |||
--moltenize |
false | Molten rather than tabular output | |
Optional Outputs | |||
--out -o |
stdout | An output file created by the walker. Will overwrite contents if file exists | |
Optional Parameters | |||
--genotypeFilterExpressionComp -gfc |
[] | One or more criteria to use to set COMP genotypes to no-call. These genotype-level filters are only applied to the COMP rod. | |
--genotypeFilterExpressionEval -gfe |
[] | One or more criteria to use to set EVAL genotypes to no-call. These genotype-level filters are only applied to the EVAL rod. | |
--printInterestingSites -sites |
NA | File to output the discordant sites and genotypes. | |
Optional Flags | |||
--ignoreFilters |
false | Filters will be ignored |
Arguments in this list are specific to this tool. Keep in mind that other arguments are available that are shared with other tools (e.g. command-line GATK arguments); see Inherited arguments above.
The variants and genotypes to compare against
The callset you want to treat as 'truth'. Can also be of unknown quality for the sake of callset comparisons.
This argument supports reference-ordered data (ROD) files in the following formats: BCF2, VCF, VCF3
R RodBinding[VariantContext] NA
The variants and genotypes to evaluate
The callset you want to evaluate, typically this is where you'd put 'unassessed' callsets.
This argument supports reference-ordered data (ROD) files in the following formats: BCF2, VCF, VCF3
R RodBinding[VariantContext] NA
One or more criteria to use to set COMP genotypes to no-call. These genotype-level filters are only applied to the COMP rod.
Identical to -gfe except the filter is applied to genotypes in the comp rod.
ArrayList[String] []
One or more criteria to use to set EVAL genotypes to no-call. These genotype-level filters are only applied to the EVAL rod.
A genotype level JEXL expression to apply to eval genotypes. Genotypes filtered in this way will be replaced by NO_CALL.
For instance: -gfe 'GQ<20' will set to no-call any genotype with genotype quality less than 20.
ArrayList[String] []
Filters will be ignored
The FILTER field of the eval and comp VCFs will be ignored. If this flag is not included, all FILTER sites will
be treated as not being present in the VCF. (That is, the genotypes will be assigned UNAVAILABLE, as distinct
from NO_CALL).
boolean false
Molten rather than tabular output
Moltenize the count and proportion tables. Rather than moltenizing per-sample data into a 2x2 table, it is fully
moltenized into elements. That is, WITHOUT this argument, each row of the table begins with the sample name and
proceeds directly with counts/proportions of eval/comp counts (for instance HOM_REF/HOM_REF, HOM_REF/NO_CALL).
If the Moltenize argument is given, the output will begin with a sample name, followed by the contrastive genotype
type (such as HOM_REF/HOM_REF), followed by the count or proportion. This will significantly increase the number of
rows.
boolean false
An output file created by the walker. Will overwrite contents if file exists
PrintStream stdout
File to output the discordant sites and genotypes.
Print sites where genotypes are mismatched between callsets along with annotations giving the genotype of each callset
to the given filename
PrintStream NA