Variant Missing in VCF but Present in BAM Files: Discrepancy Identified via IGV Report in Join-Call Project

Question

[lunping98]

While following GATK's best practices for joint variant calling in a population, I noticed that some variants were missing in certain samples. However, when I checked the BAM files with IGV report, these samples actually contained the variants. For more details, please refer to the B.html file.

[Pierre Lindenbaum]

How many reads carrying tthe REF allèle do you have at the very same location ?

[lunping98]

Here are the reads carring the ref allele

chr1:3,044,701

Total Count 57

A 0

C 8 (14%, 6+, 2- )

G 49 (86%, 27+, 22- )

T 0

N 0

[Pierre Lindenbaum]

So there is only 14% of reads carrying the alt allele. May be your caller is too stringent.

Answer 1

[GokalpC]

It could be due to an assembly failure at the very same location. Can you bring up bamout files from those locations and check if the assembly has ever been there?

If not you can try different options to recover

--recover-all-dangling-branches true
--max-reads-per-alignment-start 0
--linked-debruijin-graph
--adaptive-pruning true

and finally

--use-pdhmm true
--pileup-detection true
--pileup-detection-enable-indel-pileup-calling true

Last 3 parameters are not compatible with GVCF output if that is what you are interested for joint calling. But you may enable them to diagnose your case.

Also check if those missing sites have call confidence of at least 30. GenotypeGVCFs default parameters do not emit sites that have GQ less than 30. You may wish to reduce that threshold to 10 to see if those sites get emitted.

Use version 4.6.1.0.

Regards.