Entering edit mode
10.7 years ago
k.sivan
▴
10
We are trying to benchmark Samtools results against GATK to fine tune our SNP calling pipeline for exomes. While the specificity is extremely good (>98%), we are having low sensitivity (45% - 65%). We found that reducing the SNP Q-score threshold (from 70 - 30) improved our sensitivity. Decreasing the depth threshold (D value) from 5 to 3 also improved it marginally. What else could we try to improve the sensitivity? Thanks.
If you remove all filtering, you can maximize your sensitivity. However, that will come at the expense of decreasing specificity. The amount of tradeoff is, of course, up to you.