When the read maps to several places equally well and one of these mappings span exon/intron junction only one alignment is reported (alignment spanning exon/intron junction). I will provide test case:
file genome.fa:
but there should be two alignments. Note also that alignment is reported as unique (score 50 and NH:i:1), and when the user will filter uniquely mapped reads it will get many reads spanning exon/intron junctions that are actually not unique.
I tracked the problem to tophat/src/align_status.cpp:138 where score for alignments spanning exon/intron junctions is increased by 2, subsequently all other alignments will be considered as secondary since they have less score (by 2). I don't understand the reason for re-weighting scores of these alignments, but commenting out this line solves the problem.
When the read maps to several places equally well and one of these mappings span exon/intron junction only one alignment is reported (alignment spanning exon/intron junction). I will provide test case: file genome.fa:
file test.gtf:
file test.fq:
commands to run:
this will produce file tophat_out/accepted_hits.bam containing single alignment:
but there should be two alignments. Note also that alignment is reported as unique (score 50 and NH:i:1), and when the user will filter uniquely mapped reads it will get many reads spanning exon/intron junctions that are actually not unique.
I tracked the problem to tophat/src/align_status.cpp:138 where score for alignments spanning exon/intron junctions is increased by 2, subsequently all other alignments will be considered as secondary since they have less score (by 2). I don't understand the reason for re-weighting scores of these alignments, but commenting out this line solves the problem.