Confidence of calling variants in "homopolymer and low complexity repeat regions"

[satish1482]

Hi Friends,

We are using NGS data from TSACP illumina 48 gene amplicon panel for diagnostic purpose. The data analysis results into lot of variants lying in the "homopolymer and repeat regions" with very good number of supporting reads. The Base Quality, Mapping Quality, Strand Bias and other parameters are all behaving like normal variants. What is the confidence level for caling such variants a "Significant Variant". Some are also known SNPs in SNP database.

Any comments regarding this will be grateful.

Cheers Satish

[Brian Bushnell]

If the mapping quality is high and trustworthy, that indicates that the reads are mapped with high confidence in those regions. So as long as the regions are assembled correctly, I don't see a reason to not trust those calls unless you believe the mapq scores are incorrect, in which case all of the data is compromised, not just that in homopolymer/repeat areas.

[nucacidhunter]

If variants are InDel types, they could be result of polymerase slippage during PCR but if they are base changes (SNPs) they are most likely real as Brian has mentioned.

[satish1482]

I am more concerned about the InDels. I am aware that, it can be the result of polymerase slippage during PCR. But, there are known variations at that location many times. I am looking at each InDels one by one and most of them results into a bad call based on Positional Bias, neighborhood noise. The most trustworthy are those with homozygous calls.