Find bams with identical names and merge them

Question

[hylicase]

I'm trying to merge different bam files with the same name (and, theoretically, the same sequence). They are the same set of libraries sequenced at two different times, and I want to merge the matching pairs in order to simulate greater sequencing depth.

I can do this just fine with two bams using picard MergeSameFiles or samtools merge, but the issue is that I have 96 bams in each folder. I'd like to do this programatically, not manually, and be able to reproduce the process with different datasets in the future.

My hunch says that the simplest way to do this would be to use the bam filenames: loop over my two folders, find the bams that share a filename, and merge each pair into a single output file, but my shell chops are still rough and I am hitting a wall.

I've been starting with getting a list of bams of interest (I've also attempted dumping this list to a file with -fprint0):

find_bams() {
    find "$run_folder" -type f -name "*.bam"
 }

Then, what I think I should do is loop over the list:

for i in $(find_bams)
    do
        s=$(basename "$i" .bam)
        picard MergeSamFiles I="$i" O="$s".bam
    done

This is where I get stuck. First, there needs to be two input files to merge, and second, those two inputs must have matching filenames.

This is more of a shell scripting problem than a bioinformatics software problem, but I imagine I'm not the only one who has had to do this. Any help would be greatly appreciated.

Edit: Solved with the help of h.mon and finswimmer; see comments

[swbarnes2]

People probably don't want to write this for you...why don't you share what you have already, and why it's not working?

[hylicase]

Definitely not my intention to have people write it for me. I updated the post with what I've got. Thanks for following up.

Answer 1

[h.mon]

Unless you want to a write a more general script to tackle the problem in a more robust fashion, I would try a simpler approach. If the sets of bams have exactly the same names (as you said they have), I would do something like:

cd /path/to/bams1
for i in *.bam
do
    picard MergeSamFiles I=/path/to/bams1/"$i" I=/path/to/bams2/"$i" O=/path/to/merged/"$i"
done

If you keep the same structure of folders and filenames for future datasets, this simple solution will work just fine.

[hylicase]

I adapted this to work for my situation:

folder1=$1
folder2=$2

# since i expect both folders to have the same filenames, 
# checking just first folder should be fine
find_ds_bams () {
    find "$folder1"/*/full -type f -name "*.bam" -exec basename {} \;
}

for i in $(find_ds_bams)
do
    s=$(basename "$i" .bam)
    $picardcmd MergeSamFiles \
        I="$(readlink -f $folder1)/$s/full/$i" \
        I="$(readlink -f $folder2)/$s/full/$i" \
        O="./merged/$s.merged.bam"
done

I must have been thinking too much. Thanks a lot!

Answer 2

[finswimmer]

You can try this:

$ find /path/to/run_folder/ -type f -name "*.bam" \
| awk '{n=split($0, filename, "/"); input[filename[n]] = input[filename[n]]$0" "} END { for(file in input) {gsub(/ $/, "", input[file]); print file" "input[file]}}' \
| parallel --colsep " " 'samtools merge {}'

fin swimmer

[hylicase]

Thank you for your response. I found a working solution via the shell, but will explore awk further.

Answer 3

[cpad0112]

 $ ls test1/*.bam
test1/bam1.bam  test1/bam2.bam  test1/bam3.bam  test1/bam4.bam  test1/bam8.bam

 $ ls test2/*.bam
test2/bam1.bam  test2/bam2.bam  test2/bam3.bam  test2/bam4.bam  test2/bam5.bam

 $ find test1 test2 -type f  -name '*.bam' | xargs basename -a | sort | uniq -d| parallel --dry-run  picard MergeSamFiles I=test1/{} I=test2/{} O={.}.bam 

picard MergeSamFiles I=test1/bam1.bam I=test2/bam1.bam O=bam1.bam
picard MergeSamFiles I=test1/bam2.bam I=test2/bam2.bam O=bam2.bam
picard MergeSamFiles I=test1/bam3.bam I=test2/bam3.bam O=bam3.bam
picard MergeSamFiles I=test1/bam4.bam I=test2/bam4.bam O=bam4.bam