SNP calling

[elondin]

Hi all,

I have a question regarding determing if an identified SNP would be a heterozygote. When looking at the total number of reads for a specific base, what percentage of those reads need to differ from the reference to be considered a heterozygote. For example, if a base has 50 a coverage of 50 reads, but 25% of them differ from the reference would this be called as a heterozygote? Essentially, at what proportion of calls that differ from the reference is a homozygote vs. heterozygote?

Any thoughts?
thanks

[flxlex]

The difference from the reference is in a way irrelevant here, I feel. What matters is the amount of reads that show a difference for the organism you sampled. If around 50% shows one base, and the rest shows the other, then the organism you sequenced is a heterozygote at that position, right? How much deviation you can have from 50/50, I don't know, but there will be some spread...

[husamia]

you stated "When looking at the total number of reads for a specific base, what percentage of those reads need to differ from the reference to be considered a heterozygote." I say "When looking at the total number of reads for a specific base, what percentage of those reads need to differ from the reference to be considered a variant?" I have seen cutoff of 15% and 20%. We're only talking about diploid organisms of course. Its a good question, since you have to take into account only true reads.

[krobison]

A number of the SNP callers have attempted to answer this question (with the additional complication that your bases may not all be called to the same confidence) using Bayesian principles; if you really want to understand it, you should study those papers.