How to compare Read IDs from different Bam files

[jullee]

Hi

I have aligned (and filtered) my paired-end sequence data to two alternative reference genomes (e.g. different species). I would like to get a list of reads that align to both references and a list of reads that only align to one. My approach so far is as follows:

1) convert my final bam files to sam files

samtools view -h -o Ref1.sam Ref1.bam

samtools view -h -o Ref2.sam Ref2.bam

2) extract column 1 (sequence IDs) from both files (additional grep step to remove the header)

awk '{print $1}' Ref1.sam > Ref1Ids
grep -v '^@' Ref1Ids > Ref1Ids_Final

(repeat for Ref2 sam file)

3) ?

Now I'm thinking about using basic linux commands to compare the text files listing sequence IDs. But I'm having trouble figuring out how. I think the "comm" command requires that the files be "sorted" but when I use the "sort" command on my files, the resulting order of read IDs in the files doesn't make sense (e.g. read names are composed of both strings and numbers; the strings are constant but the numbers vary and are out of a logical order following the sort command).

Anyone have any suggestions? Or a better way to get a list of read IDs that differ between two bam alignment files?

Thanks!

[lindenb]

I wrote a tool to compare two BAMS: https://github.com/lindenb/jvarkit/wiki/CmpBams

[kmcarr]

Hi Jullee,

Here is how I would do it:

samtools view -F 4 Ref1.bam | cut -f1 | sort -u > Ref1_reads.txt
samtools view -F 4 Ref2.bam | cut -f1 | sort -u > Ref2_reads.txt

comm -12 Ref1_reads.txt Ref2_reads.txt > AlignToBoth.txt
comm -23 Ref1_reads.txt Ref2_reads.txt > AlignToRef1Only.txt
comm -13 Ref1_reads.txt Ref2_reads.txt > AlignToRef2Only.txt

Let me break this down:

The samtools command is converting the the BAM to SAM format but filtering out any reads which are not aligned to the reference (-F 4).

You should not output the header information (-h) since it is meaningless for your purposes (and you got rid of it later in your pipeline anyway).

I do not bother saving the intermediate .sam file since I really don't need to, just pipe the output into the next step.

Just take the read ID in column 1 with cut. (You used awk for this but whatever you like.)

Sort the ID's and only keep one, unique copy of each (-u). Save the output.

You were concerned about the order output by sort; don't be. sort in this case is outputting the read IDs in lexical (dictionary) order. This is the order which is required by comm.

The comm commands are pretty straightforward, outputting just one of the three columns in each run. You could run comm just once and then separate the columns but I find this method a little easier.

ONE BIG NOTE: This pipeline will collapse the paired read IDs into a single entry meaning that only one of the reads from a pair need align to either of the references to be counted.

[jullee]

Thank you kmcarr and lindenb for the helpful replies! I will try these out.