Variant callers reporting different read depth on same alignment

Question

[Lauren]

I generated two VCFs, one with freebayes and one with GATK on the same BAM file.

At one position, freebayes called an MNV. At the same position, GATK called an SNV. What confuses me is that the depths are different in the VCF file. for example,

freebayes called:

chr 9 35906602 C A,T 0/1 6,7,6:19

and GATK called:

chr 9 35906602 C A 0/1 8,7:21

This might be obvious, but I'm new to all this. How can two callers report different total read depths like this? And how would one caller report a higher reference than the other?

[Devon Ryan]

It's quite likely that they have slightly different defaults for filtering.

[WouterDeCoster]

And how does it look like in e.g. igv?

Answer 1

[Pierre Lindenbaum]

What confuses me is that the depths are different in the VCF file. f

different tools use different treshold; for example GATK will not consider

• reads having a poor mapping quality ( HCMappingQualityFilter )
• bases having a poor --min_base_quality_score ( see https://software.broadinstitute.org/gatk/documentation/tooldocs/current/org_broadinstitute_gatk_tools_walkers_haplotypecaller_HaplotypeCaller.php )
• reads with bad properties see HapCaller page : HCMappingQualityFilter MalformedReadFilter BadCigarFilter UnmappedReadFilter NotPrimaryAlignmentFilter FailsVendorQualityCheckFilter DuplicateReadFilter MappingQualityUnavailableFilter