What allele frequency threshold do you recommend?

Question

[agata88]

Hi all!

I have a result from Varscan mpileup2snp and mpilepu2indel with default parameters including variant allele frequency =0.01 (1%). (--min-var-freq)

I am wondering, what is the recommendation for allele frequency threshold? This 1% looks very low and creates results with unwanted records.

Thanks in advance!

Best,

Agata

[WouterDeCoster]

Is this for somatic or germline variants in a diploid organism?

[agata88]

Its germline, human.

[WouterDeCoster]

So theoretically the variant frequency for a heterozygous variant would be ~50%, right? Then indeed 1% is very unlikely. Obviously it will rarely be exactly 50%, but with sufficient coverage it should be quite close to that. It's a trade-off between sensitivity and specificity, but I don't think it's unreasonable to expect at least 25%.

[agata88]

Thanks, I agree, especially when we have probe, WES and WGS sequencing where the de-duplication should be performed. I thought that maybe even higher in this situation, 35%?. In primer targeted sequencing (multipleks) I think it should stay 1%, do you agree?

[WouterDeCoster]

I thought that maybe even higher in this situation, 35%?

You can indeed be more stringent and as such reduce false positives but you'll miss variants which are by chance only represented in 34% of the reads. It's a trade-off.

In primer targeted sequencing (multipleks) I think it should stay 1%, do you agree?

Multiplex PCR-based target enrichment prior to NGS? I think you'll commonly get high coverage there and can be more stringent. Unless you have SNPs in your primer I don't think you should expect a divergent frequency of variant alleles.