When I look at bwa mem alignments for some samples I have vs. a genomic reference, the highest mapping quality scores for unique, perfect matches I'm seeing are 60. But when I look at the same alignment from bowtie2 the scores assigned to those perfect matches seems to be in the 40s (43 I think). I can understand that different aligners may calculate the mapping quality in different ways, but shouldn't the score for a perfect alignment be the same?
Also, I'm not 100% sure that bowtie2 produces a real mapping score. I think I've read that it doesn't use quality values, but when I look at the full range of mapping qualities I'm getting from bowtie2 its a small, fixed set of values (0, 1, 3, 43). If it wasn't trying to use quality values I don't understand why it would output different values (as opposed to just dropping a fixed value in there).
I started looking into this because we have some bowtie2 alignments that we want to feed to the unified genotyper and for the final filtering step we need to choose an RMSMappingQuality cutoff. For bwa mem we've always set that value to 40, but since these bowtie results seem to produce lower mapping qualities for perfect hits, it made us wonder if 40 was still good to use.
Also, I'm not 100% sure that bowtie2 produces a real mapping score. I think I've read that it doesn't use quality values, but when I look at the full range of mapping qualities I'm getting from bowtie2 its a small, fixed set of values (0, 1, 3, 43). If it wasn't trying to use quality values I don't understand why it would output different values (as opposed to just dropping a fixed value in there).
I started looking into this because we have some bowtie2 alignments that we want to feed to the unified genotyper and for the final filtering step we need to choose an RMSMappingQuality cutoff. For bwa mem we've always set that value to 40, but since these bowtie results seem to produce lower mapping qualities for perfect hits, it made us wonder if 40 was still good to use.