Variant Allele Frequency from Ion Torrent data

Question

[stamakro]

I have a vcf file with SNPs called using the Torrent Variant Caller and I want to calculate their Variant Allele Frequencies.

The issue is that there is both a DP (depth) and a FPD (flow depth) fields as well as an AO (Alternate Allele Observations) and a FAO (Flow Alternate Allele Observations) fields and the two often disagree by a lot.

Which pair should I trust? DP & AO or FDP & FAO?

Or should I just ignore the cases where they disagree?

Thanks

Answer 1

[Katie D'Aco]

The fields FPD and FAO are after correcting for flow order of nucleotides, and so should be more accurate.

From https://ionreporter.thermofisher.com/ionreporter/help/GUID-A6D3AE1A-84C6-4B66-AA0D-69244576D2FE.html it looks like they don't use FPD in their allele freq calculations:

"AF=FAO/(FAO+FRO) and not FAO/FDP. This is because FDP may include reads that don't fit the flow space profile of any hypothesis; in such cases, FDP>=FAO+FRO and this is not used in allele frequency calculation"