I've been attempting to find a CNV caller for WGS data that utilises both BAF and read depth but there seems to be a dearth of them. I've only managed to find PennCNV-Seq (appear to lack a corresponding publication), Weaver (cancer only, wasted a day failing to get it to run due to severe lack of documentation), and HadoopCNV (requires hadoop cluster).
I'm wondering what I'm missing. Are there really almost no WGS BAF+read depth CNV callers? Are people using purely read depth based callers for CNV calling, or are they using CNV callers designed for SNP arrays and converting the read depth signal to a logR signal to do CNV calling?
Sequenza uses BAF. I think cnMops does too.
I had to look up BAF. Where did that come from? What's wrong with "AF"?
I got:
So... is there a difference between BAF and AF? They seem the same to me. Perhaps that definition is wrong?
BAF tends to refer to inherited (germline) SNPs, whereas AF (or VAF) are often associated with somatic SNVs. The B allele could also be considered the allele that is more rare in the population.
Thanks for the reply. That still seems very vague, though. According to what you said, the definition I posted is wrong as there is an implicit "B is not the reference allele". Does anyone have a better definition of BAF?
It's a historical term from SNP arrays in which the "B" allele probe was the non-reference allele. You are correct in that the B is redundant as pretty much the first thing CNV callers do to allele frequency is convert to minor allelic fraction regardless of whether it was the reference allele or not (i.e af = min(af, 1-af))