Compare two BAM file

[hthl85]

Hello everyone,

I have some questions about BAM file.
I used samtools to convert BAM file to CRAM, then I used samtools convert CRAM file back to BAM. The size of the new BAM file is larger than the old one. Why the BAM file increase the size? I also want to compare the two BAM files to see are they different or not. Thank you.

[GenoMax]

Have you tried comparing md5sums (after freshly sorting the files)?

Are both versions on the same storage partition?

[hthl85]

Have you tried comparing md5sums (after freshly sorting the files)?

Are both versions on the same storage partition?

I didn't check md5sum when I posted the question. But the size increase significantly.
For example:
Old BAM Size(byte) | New BAM Size(byte) | Space Increase(byte) | %
8398197 | 8442685 | 44488 | 0.53
1595537237 | 1605593870 | 10056633 | 0.63
3944613 | 4031703 | 87090 | 2.21
1975987646 | 1989339186 | 13351540 | 0.68

From your suggest, I check md5sums they are different. The old and new BAM files in the same directory.

[dpryan]

Try piping both as SAM to md5sum.

[gandalf886]

If I understand correctly, the md5 should be different anyway even if the two bam file are of the same size? Because they are created by different programs.

[hthl85]

If I understand correctly, the md5 should be different anyway even if the two bam file are of the same size? Because they are created by different programs.

Do you have any idea? or Why the two BAM files are different? I found a blog mention something similar about the size of BAM file. http://davetang.org/muse/2014/09/26/bam-to-cram/

[Brian Bushnell]

As Devon said, compare them as sam files.

Convert both of them to sam and compare them. You can't get trustworthy results from comparing compressed files. If they both yield identical sam files, they are identical. If not, they differ.

[jkbonfield]

There is a compare_sam (not bam, but that's a trivial conversion) script in the htslib test harness.

https://github.com/samtools/htslib/blob/develop/test/compare_sam.pl

It's pretty crufty and not really designed for use outside the test system, but if you're familiar with perl then you should be able to drive the different options easily enough.

As for why the size is larger, it may come down to NM and MD auxiliary tags. These are not stored verbatim in CRAM, but worked out on-the-fly from the sequence delta to the reference. It doesn't store a flag to indicate whether there were present in the original input data, so by default you'll get them whether you wanted them or not.

[hthl85]

Thank you jkbonfield and everyone. I used the script jkbonfield suggested to compare two SAM file. The one generated from CRAM had extra MD tags and that caused the file increased its size.

There is a compare_sam (not bam, but that's a trivial conversion) script in the htslib test harness.

https://github.com/samtools/htslib/blob/develop/test/compare_sam.pl

It's pretty crufty and not really designed for use outside the test system, but if you're familiar with perl then you should be able to drive the different options easily enough.

As for why the size is larger, it may come down to NM and MD auxiliary tags. These are not stored verbatim in CRAM, but worked out on-the-fly from the sequence delta to the reference. It doesn't store a flag to indicate whether there were present in the original input data, so by default you'll get them whether you wanted them or not.