MAPQ algorithms of Giraffe

Question

[gulin]

Hello! I am comparing my giraffe mapping result with bwa-mem result (aligned to linear reference). The MAPQ algorithms in two different mapping tools is different or not? Can I use same threshold(such as MAPQ>30) to filter the reads? Or they have different meanings in different mapping tools?

Answer 1

[anovak]

vg giraffe does not use the exact same MAPQ algorithm as bwa-mem. However, its mapping qualities are still meant to be Phred-scale probabilities of a read being mapped to the wrong place. (They might not always be well-calibrated probability estimates that truly reflect the probability of a read being mismapped, depending on how good or bad Giraffe is at its job.) So MAPQ > 30 means the same thing for Giraffe and for bwa-mem, but the resulting populations of reads you select with it might still be importantly different.

If you want to compare Giraffe and BWA MAPQ calibration, Supplementary Figure 5 in the Giraffe paper has an evaluation of how well-calibrated Giraffe and BWA MEM mapping qualities are on simulated reads.