Allele Distributions

[jkozubek]

Hello:

I found 320 mutations among 16 mouse clones using GATK. However, I noticed something strange. In only 40 of those 0/1 calls are there more reads with the ALT call (1) than reads with the REF call (0). In the vast remainder, 280 mutations I found, the reads that carry the ALT calls are a smaller fraction about 10-40 percent.

If all my 320 mutations were real, I would expect the REF/ALT distributions to be about equal. If it was true and normally distributed, I'd expect at least 50% of these mutaitons to show up with more ALT calls and 50% to show up with more REF calls. Is it natural to find fewer reads with the ALT call in genuine mutations?

I was thinking about using the bionomial distribution in Excel to remove these, something like -BINOMDIST(20,100,.50,FALSE) where I found 20 ALT calls with a Depth of 100 and expected to see 50% ALT calls.

Jim

[vivek_]

I can't point to a proper source right away but I remember reading in a publication that the alternate allele fraction for heterozygous calls could be anywhere between 20 - 80% at variation sites.

[jkozubek]

I think so too. Still it seems odd that the distribution is lopsided so that many more times its the ALT that is seen in only 20-30% in the reads.

[swbarnes2]

Keep in mind that there is some bias towards the reference allele. If your aligner only accepts 3 discrepancies between read and reference, reads matching the reference allele with 1,2 or 3 mistakes will align there, but only reads with 1 or 2 errors will align if they also have the alternate allele.

You could try aligning to a reference that contains both alleles, see if that fixes things.