Closed serine closed 5 years ago
this has been implemented as per commit about, but turns out that this wasn't necessary as featureCounts
filters multi-mappers via NH
field in the bam files. I think I new that but have forgotten. So this wasn't a bug after all, but I'm going to leave -Q 10
filtering in. For STAR and tophat2/hisat2 aligners this shouldn't do anything.
relevant bit of code from featureCounts
source
if( NH_value > 1 ) {
if(global_context -> is_multi_mapping_allowed == 0) {
// now it is a NH>1 read!
// not allow multimapping -> discard!
if(RG_ptr){
void ** tab4s = get_RG_tables(global_context, thread_context, RG_ptr);
fc_read_counters * sumtab = tab4s[1];
sumtab -> unassigned_multimapping++;
}else thread_context->read_counters.unassigned_multimapping ++;
if(global_context -> read_details_out_FP)
write_read_details_FP(global_context, thread_context, "Unassigned_MultiMapping", -1, NULL, bin1, bin2);
return;
}
}
note for myself; think about
-Q
field infeatureCounts
Is it right to remove multi-mapping reads from counts table?It feels reasonable to remove them, otherwise you exaggerate you counts?
Bear in mind that at least for STAR
MAPQ
is a discrete numbers more information hereAny comments on this issue will be much appreciated guys !
Cheers