Samtools filter reads based on CIGAR values

[krapulaxdoctor]

Dear all,

I would like to ask for some help with samtools:

I found many threads about using samtools to filter reads based on some criteria, however, I did not find any clue how to filter reads based on the CIGAR value.

How can I filter reads (from a BAM) from RNA-seq files that are non-splitted / non-gapped?

From one BAM, I would like to have two BAMs ne with gapped reads and one with non-gapped reads.

Could someone write an example?
Thank you for the help in advance,

[dariober]

Gapped reads are those containing the N operator in the cigar string right? You could do this:

samtools view -h in.bam \
| awk '{if($0 ~ /^@/ || $6 ~ /N/) {print $0}}' \
| samtools view -Sb - > gapped.bam

And for ungapped reads:

samtools view -h in.bam \
| awk '{if($0 ~ /^@/ || $6 !~ /N/) {print $0}}' \
| samtools view -Sb - > ungapped.bam

[HESmith]

Be aware that deletions (CIGAR string D) also give rise to gapped alignments, and the representation as N vs. D depends on the gap length and the aligner.

[dariober]

Be aware that deletions (CIGAR string D) also give rise to gapped alignments, and the representation as N vs. D depends on the gap length and the aligner.

True, but I surmise the OP wants to select reads spanning different exons as opposed those only assigned to one exon. If this is the case, I think tophat uses N to mark gaps between exons (don't know other aligners).

[dpryan]

True, but I surmise the OP wants to select reads spanning different exons as opposed those only assigned to one exon. If this is the case, I think tophat uses N to mark gaps between exons (don't know other aligners).

Realistically, any aligner that advertises the ability to handle spliced reads will do this too. If not, it shouldn't be used.

[krapulaxdoctor]

Thank you for all the quick responses.
It is a great help for me.

dariober:

True, but I surmise the OP wants to select reads spanning different exons as opposed those only assigned to one exon. If this is the case...

Yes, that would be the main goal. Thanks.

[HESmith]

Realistically, any aligner that advertises the ability to handle spliced reads will do this too. If not, it shouldn't be used.

I agree 100%, but it would not be the first time that someone used an inappropriate tool for the job. The alert was directed at the OP, in case the awk filter did not produce the expected results.