Standard percentage of reads to call heterozygous

[michael-steffen]

Hello,

I was wondering what other people are doing themselves or if they can direct me to some literature, where people decide how many hits they get to call a position heterozygous with Illumina paired end read data.

I'm going to do a lit search myself, but thought posting could be a faster way to figure out my question in the mean time.

PS. I am predominantly using CLC genomics workbench.

Thanks,
Mike

[michael-steffen]

Hmmmm, 20% seems to be what many are doing. Is the general concensus?

[Brian Bushnell]

Depends on the data, mutation type, and number of reads... For example, I would expect an indel in highly-amplified, non-deduplicated exon-captured data with low coverage to be much farther from 50/50 compared to a SNP in whole-genome, unamplified data with high coverage.

The best approach is to calibrate your data, rather than just using a rule of thumb.

[michael-steffen]

Hmm, alright, well the data I am working with is transciptomic data which was extracted using a TRUseq RNA kit. So, the DNA was amplified a bit, but I also removed duplicated reads prior to mapping.

Any recommendations?

Thanks,
Mike

[Brian Bushnell]

I am not an expert on this, so perhaps someone will correct me, but my understanding is that you cannot reliably determine the ploidy of variations using RNA-seq data, because there may be highly differential expression of a gene from one chromosome copy compared to the other. Some genes are even completely shut off on one copy. Therefore, you should obtain DNA data if you want reliable ploidy data.

[Wallysb01]

Wouldn't this then mean that you could call hets, and called hets are very likely hets, but homozygous calls could be hets? So that with het calling, its more the false negative rate that causes the problem?

[Brian Bushnell]

Yes, that's true. As long as you don't care about either false positives or false negatives, calibration is trivial; so in this case, if false negative het calls are not a problem, just set the ratio at something conservative like 25% minor allele / minimum depth 16, and it will be fine. But most tests are employed specifically because both false positives and false negatives are a problem.