Is it possible to split BAM without sorting it?

Question

[tomas4482]

I have some large bam files for a pipeline. But the RAM ran out when I was performing the pipe. Sorting bam will cause fatal bugs to my pipeline and cannot be debuged for now, hence not the solution. The only option is to downsize the input (upgrading the RAM cannot be done for now due to the high expense).

Then I thought I could split the original bam to several smaller bam files. But I cannot skip the sorting step. E.g. samtools view requires sorting and indexing before spliting.

Is there other way to do this?

Thank you.

[GenoMax]

samtools view -h should not need sorted BAM. If you split the BAM file (creating muiltiple intermediate SAM files) make sure you add the header to all the pieces otherwise the files may be unusable.

[tomas4482]

I've tried. It requires an indexed bam. But indexing requires sorted BAM.

[GenoMax]

How so? You are simply converting a BAM file to SAM using samtools view -h.

[tomas4482]

Here is the command I used to split the bam. samtools view -h -@ 48 sampleAligned.out.bam chr1 > sample_1.bam. Because I specified the region, it requires sorted and indexed BAM.

May I ask how to create splitted BAM referring to your method? Thanks.

[GenoMax]

Because I specified the region

This is new information. Your post originally had not asked about splitting a BAM not based on a specific region. If you need to do this then indeed the file will have to be sorted.

If you don't have the resources, you could split the unsorted file first. Grab regions you need from the pieces after sorting/indexing them, merge the region specific files and then sort them again. This will be a lot more work but will eventually get you what you need.

Answer 1

[LChart]

Using samtools alone the bam file needs to be sorted. However at the cost of materializing the actual string values, you can use awk to do this:

samtools view -h $bam | awk 'substr($1,0,1) == "@" || $3 == "chr1"' | samtools view -hb > chr1.bam

etc.

[tomas4482]

Thank you. It is what I need. Besides, I found bamtools split -reference has similar effect and is more convenient for such purpose.

[tomas4482]

May I ask one more question? Is it possible or reasonable to split BAM by lines? E.g a BAM has 10000 lines. Can I split the BAM into two 5000-line BAM?

[LChart]

Not all in one go, because you need to replicate the header for each file; but awk works here too:

awk 'substr($1,0,1) == "@" || (NR >= 5001 && NR < 10000)'

You will need to increase 5001 and 10000 to also include the number of lines in your header; but you get the idea.

[tomas4482]

Thank you!