Variant position showed up in mpileup but not in VCF call

Question

[leemhmark]

Hello,

I'm doing consensus calling using bcftools. The command I'm using is:

bcftools mpileup --max-depth 0 --ff DUP -f [fa file] [deduped bam file] | bcftools call -mv -Ov -o [variants.vcf]

I looked at the intermediate mpileup output created using bcftools mpileup --max-depth 0 --ff DUP -f [fa file] [deduped bam file] > [mpileup.txt] and there's a variant found at position 1703:

gb-cds|KM069571.1|Influenza_virus|HA|A|H3|A/Singapore/C2012.669/2012|   1703    .   T   C,A,<*> 0   .   DP=40;I16=13,21,1,1,1043,39031,72,2592,1220,45432,86,3700,802,19468,30,650;
QS=0.927565,0.0362173,0.0362173,0;
VDB=0.02;
SGB=-0.453602;
RPB=0.132353;
MQB=0.220588;
MQSB=0.993571;
BQB=0.558824;
MQ0F=0  PL  0,69,255,69,255,255,102,255,255,255

But in the resulting .VCF file I don't see this variant showing up anywhere. Is it because the C and A variants are too infrequent or low quality? If so how do I interpret all the fields in the mpileup entry?

Thanks!

Answer 1

[Kevin Blighe]

Well, there are 3 variants found at the same position. What are the individual allelic depths? If needed, view the region in IGV, as this can usually make it easier to understand why a call is not made

[leemhmark]

So I checked the region in IGV and the allelic depths are: Total count: 1263 A: 5 reads (0%, 1+, 4-) C: 19 reads (2%, 4+, 15-) T: 1239 reads (98%, 148+, 1091-)

Is 2% considered too low a depth for bcftools to consider it as a variant? I tried to look for this cutoff depth threshold used by bcftools and didn't have any luck finding this information.

[Kevin Blighe]

Yes, that is very low - BCFtools is best at calling germline variants, i.e., at frequencies of 50% or 100%. You can still call them - take a look at the varFilter executable that can be used with BCFtools and that I use here: https://github.com/kevinblighe/ClinicalGradeDNAseq/blob/master/AnalysisMasterVersion1.sh#L354-L375

If these are cancer samples, then you should obviously use a somatic variant caller.