Bowtie2 - Finding reads with multiple mapping positions

[all_your_base]

Hi, I have a question dealing with PE RNAseq data (100x2) from Illumina.

It is my understanding that Bowtie2 checks a read for possible alignments across the reference genome, and reports the best alignment, discarding the rest. However, I am interested in finding just the reads that align to exactly 2 places, and determining where both locations are on the reference. I want to be able to extract this information from the BAM file if possible. I first tried setting the -k parameter to 5, and then looking for reads that could only find two alignments, but am not sure where in the BAM to find this information.

Does anyone have experience looking for such reads in a SAM/BAM file?
Thanks!

[qiongyi]

Hi, I think you should also consider the flag column (the second column) in the bam file. It should include the 0x0002 flag (properly paired aligned to the reference).
And I would suggest you considering the mapping quality instead of the uniqueness that you mentioned here. Higher mapping quality means more uniqueness.

[all_your_base]

@qiongyi, thanks a lot for the suggestions.

I eventually solved this problem by having my Bowtie2 .SAM output sorted by read ID. This way the reads are in the order:

ID=abc123/1
ID=abc123/1
ID=abc123/2
ID=abc123/2

If I set Bowtie2 -k =3, it will report up to 3 alignments for each read. If there are exactly two alignments reported for abc123/1 and two alignments for abc123/2, I know the read has exactly two matches in the ref.

Thanks!