The calculate rule of genotypeconcordance, sensitivity and specificity
Answered
REQUIRED for all errors and issues:
a) GATK 4.1.7.0:
b) $gatk --java-options -Xmx10G GenotypeConcordance \
--CALL_VCF $cvcf \
--O $sample.concordance \
--CALL_SAMPLE $sample \
--TRUTH_SAMPLE $sample \
--TRUTH_VCF $tvcf
I can't understand the calculate rule of TP TN in *detail_metrics and the genotype concordance in *summary_metrics. I found some infomation about the fomular in https://www.n-genetics.com/products/1104/1023/18263.pdf, but I cannot replicated the result.
See forum topic details at forum guidelines page: https://gatk.broadinstitute.org/hc/en-us/articles/360053845952-Forum-Guidelines
-
Hi panhong liu,
Our best explanation for why you would be seeing this is on the tool documentation page for GenotypeConcordance: https://gatk.broadinstitute.org/hc/en-us/articles/4418051360155-GenotypeConcordance-Picard-
Please take a look and let me know if you have any questions about that page. The PDF you linked was not written by our team so I'm not familiar with the contents.
It's also helpful to provide examples of your confusing results.
Best,
Genevieve
-
Thank you for your reply.
I have calculated the genotype called from chip array and whole genome sequencing from the same sample using the GATK genotypeconcordance. Then I got the *detail_metrics and *summary_metrics file. I want to know how the GENOTYPE_CONCORDANCE in *summary_metrics file was calculated from the information of the *detail_metrics file. Thanks a lot.
the *detail_metrics info:
VARIANT_TYPE TRUTH_SAMPLE CALL_SAMPLE TRUTH_STATE CALL_STATE COUNT CONTINGENCY_VALUES
SNP S1 S1 MISSING HET_REF_VAR1 881 FP,TN
SNP S1 S1 MISSING HOM_VAR1 91 FP
SNP S1 S1 HOM_REF HET_REF_VAR1 4614 FP,TN
SNP S1 S1 HOM_REF HOM_VAR1 83 FP
SNP S1 S1 HET_REF_VAR1 MISSING 11229 TN,FN
SNP S1 S1 HET_REF_VAR1 HOM_REF 3892 TN,FN
SNP S1 S1 HET_REF_VAR1 HET_REF_VAR1 123752 TP,TN
SNP S1 S1 HET_REF_VAR1 HOM_VAR1 1788 TP,FP
SNP S1 S1 HET_REF_VAR1 HET_REF_VAR2 10 FP,TN,FN
SNP S1 S1 HET_REF_VAR1 HOM_VAR2 3 FP,FN
SNP S1 S1 HET_VAR1_VAR2 MISSING 28 FN
SNP S1 S1 HET_VAR1_VAR2 HOM_REF 6 FN
SNP S1 S1 HET_VAR1_VAR2 HET_REF_VAR1 51 TP,FN
SNP S1 S1 HET_VAR1_VAR2 HOM_VAR1 3 TP,FN
SNP S1 S1 HOM_VAR1 MISSING 6311 FN
SNP S1 S1 HOM_VAR1 HOM_REF 316 FN
SNP S1 S1 HOM_VAR1 HET_REF_VAR1 1529 TP,FN
SNP S1 S1 HOM_VAR1 HOM_VAR1 101027 TP
SNP S1 S1 HOM_VAR1 HET_REF_VAR2 5 FP,FN
SNP S1 S1 HOM_VAR1 HOM_VAR2 7 FP,FN
SNP S1 S1 NO_CALL HET_REF_VAR1 418 EMPTY
SNP S1 S1 NO_CALL HOM_VAR1 409 EMPTY
SNP S1 S1 VC_FILTERED MISSING 5806 EMPTY
SNP S1 S1 VC_FILTERED HOM_REF 370 EMPTY
SNP S1 S1 VC_FILTERED HET_REF_VAR1 533 EMPTY
SNP S1 S1 VC_FILTERED HOM_VAR1 647 EMPTY
SNP S1 S1 IS_MIXED HET_REF_VAR1 39 EMPTY
SNP S1 S1 IS_MIXED HOM_VAR1 8 EMPTY
the *summary_metrics file:VARIANT_TYPE TRUTH_SAMPLE CALL_SAMPLE HET_SENSITIVITY HET_PPV HET_SPECIFICITY HOMVAR_SENSITIVITY HOMVAR_PPV HOMVAR_SPECIFICITY VAR_SENSITIVITY VAR_PPV VAR_SPECIFICITY GENOTYPE_CONCORDANCE NON_REF_GENOTYPE_CONCORDANCE
SNP S1 S1 0.891901 0.957888 ? 0.926231 0.981181 ? 0.907013 0.968247 0.950731 0.938495 0.939947 -
Hi panhong liu,
Have you seen these tables on the Picard site? We are currently working on trying to move this information over to the GATK site, but this is what we have for now:
In the code, the summary metrics are where the calculations are happening, whereas the detail metrics are just counts. Check out these files for more details about the calculations, there are helpful details in the comments too:
- https://github.com/broadinstitute/picard/blob/master/src/main/java/picard/vcf/GenotypeConcordanceSummaryMetrics.java
- https://github.com/broadinstitute/picard/blob/master/src/main/java/picard/vcf/GenotypeConcordanceCounts.java
Let me know if there is a specific metric you want more details on.
Best,
Genevieve
-
I have two vcf files of the same sample including only variants. Why do the results of genotype-concordance and non-ref genotype concordance vary drastically? where concord is 77% and non-ref concord in 22%. I would have expected them to be the same as there is no ref-ref match
Please sign in to leave a comment.
4 comments