I would like to know more about how BWA treats the mapping score of paired end reads.
I am familiar with the process used to assign mapping scores by BWA (thanks to this post). I see that the mapping score is affected by the correct mapping of a read's pair. But to what extent?
For instance, if one read in a pair is mapped and the other is not, does the mapped read receive a reduction in its own mapping score.
Further, does anyone know the file and location of the BWA which assigns mapping score?
Thanks in advance.
2 answers
I think that the pairing quality affects the mapping score of an individual read. If I remember correctly, BWA paper mentions about use of insert size information in case of paired-end data to assign MAPQ. I think, read-pairs with only one read-mapped uniquely should have a lower score than read-pair where both the reads mapped uniquely. This is based on my assumption. i can be wrong. The best way would be to test this:
1) Get all the uniquely aligned reads from the BAM files. 2) Compare MAPQ of read pairs where only one read was aligned uniquely with those where both reads were aligned uniquely.
Use http://picard.sourceforge.net/explain-flags.html this resource to figure out the flags.
Mapping scores are about how likely it is that the read is placed in the wrong place. This basically happens for one of two reasons: your read has a lot of differences to the reference, or because the region it did align to is repeated throughout the genome.
But both of those are less likely if the mate is somewhat securely aligned to a position that makes sense (i.e. is in the right orientation at a sane distance).
But if the mate's alignment is also questionable, like if it falls in the same large repeat, or it has a lot of differences to the reference, it might not be all that helpful.
Log in to answer this question.