I am using BWA version 0.7.8-r455.
I am performing different mappings with sequences from genomic DNA (targeted by exome capture), typical 2x101 paired-end reads from HiSeq machines.
I repeated the alignment for one specific sample 3 times, just changing the -T parameter of BWA MEM:
-T Don't output alignment with score lower than INT. This option only affects output. [30]
using the values 30, 60 and 90. When I look inside the data generated, in SAM format, I can find MAPQ values from 0 to 60, as well as AS and XS scores equal to 30. However, the SAM flags show me that the results are changing, increasing the "correctly mapped" pairs, as well as the "both unmapped", for example.
Could anybody explain exactly how the -T option affects the output generated? Which alignments are filtered? Why does say "This option only affects output" when I have an increase in the number of correctly mapped reads?
Edit: Moreover, it seems that when I increase the value of -T the secondary alignments (with FLAG > 200) are the ones that are filtered. When I reach -T 90 no secondary alignments are reported, and I have just one alignment per read in the output. I will output again the documented sentence:
-T Don't output alignment with score lower than INT. This option only affects output. [30]
Edit: Is it finally just a way to filter out secondary alignments? Is it so awfully documented?
Heng says "-T sets a threshold on the alignment score, not mapping quality. Alignment score is shown in the AS:i: SAM tag. The alignment score is determined by the scoring scheme, which is specified by -A, -B, -O and -E." https://sourceforge.net/p/bio-bwa/mailman/message/30624738/