Run Samtools on multiple files

Question

[V]

I'm trying to run samtools sort in a folder containing about 380 bam files.

I've tried the following command and nothing happens. No errors, it just returns blank. Can anyone help me with what I've done wrong? Thank you :)

samtools sort -@ 7 *.bam > *.sorted.bam

Answer 1

[Medhat]

Try this;

for f in *.bam; do filename="${f%%.*}"; samtools sort -@ 7 $f ${filename}.sorted.bam; done

Or

for f in *.bam; do  samtools sort -@ 7 $f   ${f/.bam/sorted.bam}; done

[V]

Seems to be working, hasn't thrown out an error! thanks alot!!

[yolandaly]

I modified a bit based on Medhat's script. Seems better. for f in *.bam; do samtools sort -@ 7 $f ${f/bam/sorted}; done

[fatih]

hi all, I am working Linux cluster and ı have 1000 bam files under bam/ directory and they have pretty similar name "bam_runid_bf2ff0593235864c9d423d28f1746a42d62df29e_0_0" and just first "0" changing into folder like 1,2,3,.. I am wondering how can ı sort them as a single and merge them after indexing. I tried "for f in bam/*_0.bam; do samtools sort -@ 7 $f ${f/bam/sorted}; done" what u advised but it did not worked. sorry ı am very new in this field .:( could you explain to me more simple? thanks

[Pierre Lindenbaum]

post this as a new question with a reference to this thread. Show us any error message.

Answer 2

[WouterDeCoster]

Using gnu parallel in its simplest form:

ls *.bam | parallel -j 8 'samtools sort -@ 7  {} > {.}_sorted.bam'

To show progress and expected time of completion

ls *.bam | parallel --progress --eta -j 8 'samtools sort -@ 7  {} > {.}_sorted.bam'

[ole.tange]

--eta implies --progress, so --progress is not needed. Also try the newer --bar. If you want one per cpu core, leave out -j8.

[WouterDeCoster]

Thanks! I am huge fan of parallel, but so many options :o

Answer 3

[Pierre Lindenbaum]

create the following 'Makefile'

SHELL=/bin/bash
BAMS=$(shell ls -1 *.bam)
define run

$$(addsuffix .sorted.bam,$$(basename $(1))): $(1)
    samtools sort $$< $$(basename $$@)
    samtools index $$@
endef

all: $(addsuffix .sorted.bam,$(basename ${BAMS}))

$(eval $(foreach B,${BAMS},$(call run,${B})))

and run it in parallel for you 380 bam using the option -j for the number of parallel operations

make -j 16

[lakhujanivijay]

On a separate note, this is really cool, suggestions to learn how to use "make" (tutorial links) would be helpful for all.

[lakhujanivijay]

May be I am asking silly question, does that "16" means the no. of threads?

[Pierre Lindenbaum]

does that "16" means the no. of threads?

yes

Answer 4

[Charles Plessy]

The * wildcard character can only expand to already existing files (*.bam in your example). The command-line interpreter can not guess your intent and provide contents for *.sorted.bam. Moreover, samtools sort is not designed to run on multiple files in parallel.

However, there are generic tools for doing what you want. See for instance the two following links.

• https://davetang.org/muse/2013/11/18/using-gnu-parallel/
• Gnu Parallel - Parallelize Serial Command Line Programs Without Changing Them

[V]

Hi Charles,

Sorry My bad, I seem to have explained what I want to do wrong. I didn't mean I want to run it on all of them in parallel. I meant I want samtools to sort them out one by one. Usually what I would do would be:

samtools sort -@ 7 File1.bam File1.sorted.bam
samtools sort -@ 7 File2.bam File2.sorted.bam

and then as soon as it finishes with file1 it would go automatically to file2. But having 380 or so files, you can see the issue with having to write all of them out manually :)

[Charles Plessy]

Indeed, a for loop better answers to your question. Nevertheless, if you run parallel and limit the number of concurrent jobs to 1 (--jobs 1) you will also be able to run the commands sequentially, basically like in a for loop. And you will have learned a new tool :)

Answer 5

[lakhujanivijay]

try this:

for in in *.bam; do sample_name=`echo $i | awk -F "." '{print $1}'`; samtools sort -@ 7 $i >  ${sample_name}.sorted.bam ; done

[venu]

Just a comment, when you are defining output filename, it's always better to use basename in order to remove just the extension and add output format. Something like outFile=$(basename "$BAM" .bam | awk '{print $1".sorted.bam"}').

Why? Because sometimes the filenames include several . which explain a lot about the file. E.g. ICGC_File1.cohortA.rmdup.bam. With your approach you'll get an output which looses some information ICGC_File1.sorted.bam.

[lakhujanivijay]

+1 Thanks venu for the suggestion, though I am clueless why the loop I suggested does not work for V

[V]

Hello,

Thanks. I've tried it but keep getting this error appearing multiple times:

awk: read error (Is a directory)

Usage:   samtools sort [options] <in.bam> <out.prefix>

Options: -n        sort by read name
         -f        use <out.prefix> as full file name instead of prefix
         -o        final output to stdout
         -l INT    compression level, from 0 to 9 [-1]
         -@ INT    number of sorting and compression threads [1]
         -m INT    max memory per thread; suffix K/M/G recognized [768M]

[lakhujanivijay]

try this: Removed ">" and its for i in not for in in ; i edited

for i in *.bam; do sample_name=`echo $i | awk -F "." '{print $1}'`; samtools sort -@ 7 $i ${sample_name}.sorted; done

I assume you run it for a directory having files:

sample1.bam
sample2.bam
sample3.bam

and so on

[V]

Same Error coming up :/

[V]

yes, thats correct. They aren't named "sample" they have a serial number but yes all bam