multi-mappers (non-ribosomal)

[Joseph Mudd]

Hi Alex,

In one of my runs I'm getting a near ~50% of multi-mappers (attached log.final.out).

I thought these may be ribosomal.  However I've aligned one of the fastq files to human 18S rRNA sequences with BBduk and I get surprisingly few contaminants:

Input is being processed as unpaired

Started output streams: 0.078 seconds.

Processing time:    6.266 seconds.

Input:                  6336220 reads 921191628 bases.

Contaminants:            164762 reads (2.60%) 22565483 bases (2.45%)

Total Removed:          164762 reads (2.60%) 22565483 bases (2.45%)

Result:                  6171458 reads (97.40%) 898626145 bases (97.55%)

If multi-mappers are not ribosomal contamination, could you guess what they may be?  This is assuming everything is right with my command.  I have attached the Log.out.

thanks

jc

[Vasily A.]

(while waiting for Alex reply), you can just check in your output sam/bam file: filter the reads by `NH` tag. Multimapped reads will have the value >1 (in SAM file it will look like `NH:i:2` etc.)

[Alexander Dobin]

Hi Joseph,

you are using --outFilterScoreMinOverLread    0  --outFilterMatchNminOverLread   0 
which removes any limitation of the minimum alignment length. Because of that, you are getting a lot of short reads mapped as multimappers. With default paramters they would not be mapped at all.  I would recommend mapping with default parameters and 

1. Map just one pair of read1/read2 files first, to check that there is no problem with the list.

2. If the mapped % is still small, map read1 and read2 separately.

3. If the mapped % is still small, BLAST the unmapped reads from 1 to check for contamination.

Cheers

Alex