Ti/Tv ratio for known and novel SNP call sets deviates

[seqpig100]

Hi everyone,

In our group we are using samtools mpileup for variant calling on human whole genome deep sequencing data.

Basically our pipeline follows instructions on http://samtools.sourceforge.net/mpileup.shtml. We used bcftools to call snps and indels, and tried different filters on strand bias, tail distance bias, maximum read depth and minimum read depth.

The overall Ti/Tv ratio is quite smooth around 2.3. Then the snps sets were annotated. Known SNPs are defined as those found in dbSNP131, and the rest are novel SNPs. Then the Ti/Tv ratio deviates for the two sets, as shown in the attached.

This figure is the Ti/Tv ratio versus different alternative allele count for both known SNPS(red) and novel SNPS(blue). Ti/Tv ratio for known SNPs is still quite smooth. At low allele count, the Ti/Tv ratio for both known and novel SNPs are close; however, at higher alternative allele count, the Ti/Tv ratio for novel SNPs deviates away and differs great from that of known SNPs.

One possible reason could be that the number of SNPs at high alternative count is very low(only acound 150), and the Ti/Tv ratio calculated is not stable. Other than that, could anyone kindly point out what is going wrong? Is there any filtering SNPs or QC steps needed to have stable and smooth Ti/Tv ratio for novel SNPs set?

Any suggestion would be highly appreciated. Thanks!

[lh3]

You did not attached the figure. A ts/tv 2.3 is too high for whole-genome sequencing and too low for exome sequencing. How many samples?

High-frequency novel SNPs are more likely to be errors. Nonetheless, I would not worry about 150 SNPs. You have much more errors than those. In addition, I think the right way to do this kind of plot is to sort SNPs by frequency, divide SNPs into ~50 bins and plot ts/tv vs. median in each bin.