Comparing Two Bam Files Obtained From The Same Files And The Exact Settings

Question

[roll]

Hi, I just run tophat2 with bowtie1 on the same fastq files with the exact same settings. Then i am comparing the two bam files. I used cmp command in unix and it keeps telling me that

bam1 bam2 differ: byte 27, line 1

When I look at the first line, i do not see any difference at all.

1. Is it expected to have this difference between two bam files even when obtained with the same exact settings?
2. How can i actually see what is the different bit between the two bam files?

[Devon Ryan]

Convert the files to SAM (or uncompressed BAM), there are simply too many caveats when thinking about byte by byte comparisons of compressed files.

[roll]

cool, i just did that and this time i found the bitwise flags to be different. (one is 417 and the other is 161). Why there is a difference do you think?

[Devon Ryan]

That's a multimapping read with (presumably) multiple reported alignments. If all of the alignments are equal, then one can just be randomly chosen as being primary. Presumably, then, this is just a difference in the alignment assigned as primary.

[roll]

That explains all. Thanks for pointing that

[compuTE]

Is there a way of setting a seed on this??

[Devon Ryan]

No, bowtie2 has deterministic output by default. Every aligner will be different here. If you need the exact same results for CI testing then just use a small file without multimappers.

Answer 1

[Pierre Lindenbaum]

• you cannot just use cmp: if the SAM header contains something like a date, a timestamp, your whole files will be different.

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

example:

$ java -jar dist/cmpbams.jar -F -C tmp1.bam tmp2.bam tmp3.bam

#READ-Name  tmp1.bam tmp2.bam|tmp1.bam tmp3.bam|tmp2.bam tmp3.bam   tmp1.bam    tmp2.bam    tmp3.bam
HWI-1KL149:20:C1CU7ACXX:1:1101:17626:32431/1    EQ|EQ|EQ    K01:2136=83/43M3I54M    K01:2136=83/43M3I54M    K01:2136=83/43M3I54M
HWI-1KL149:20:C1CU7ACXX:1:1101:17626:32431/2    EQ|EQ|EQ    K01:2059=163/100M   K01:2059=163/100M   K01:2059=163/100M
HWI-1KL149:20:C1CU7ACXX:1:1102:16831:71728/1    EQ|EQ|EQ    K01:2133=83/100M    K01:2133=83/100M    K01:2133=83/100M
HWI-1KL149:20:C1CU7ACXX:1:1102:16831:71728/2    EQ|EQ|EQ    K01:2059=163/100M   K01:2059=163/100M   K01:2059=163/100M
HWI-1KL149:20:C1CU7ACXX:1:1105:3309:27760/1 EQ|EQ|EQ    K01:2213=83/100M    K01:2213=83/100M    K01:2213=83/100M
HWI-1KL149:20:C1CU7ACXX:1:1105:3309:27760/2 EQ|EQ|EQ    K01:2081=163/100M   K01:2081=163/100M   K01:2081=163/100M
HWI-1KL149:20:C1CU7ACXX:1:1106:2914:12111/1 EQ|EQ|EQ    K01:2136=83/43M3I54M    K01:2136=83/43M3I54M    K01:2136=83/43M3I54M
HWI-1KL149:20:C1CU7ACXX:1:1106:2914:12111/2 EQ|EQ|EQ    K01:2059=163/100M   K01:2059=163/100M   K01:2059=163/100M
HWI-1KL149:20:C1CU7ACXX:1:1107:11589:17295/1    EQ|EQ|EQ    K01:2123=83/56M3I41M    K01:2123=83/56M3I41M    K01:2123=83/56M3I41M
HWI-1KL149:20:C1CU7ACXX:1:1107:11589:17295/2    EQ|EQ|EQ    K01:1990=163/100M   K01:1990=163/100M   K01:1990=163/100M
HWI-1KL149:20:C1CU7ACXX:1:1110:14096:95943/1    EQ|EQ|EQ    K01:2123=83/56M3I41M    K01:2123=83/56M3I41M    K01:2123=83/56M3I41M
HWI-1KL149:20:C1CU7ACXX:1:1110:14096:95943/2    EQ|EQ|EQ    K01:1990=163/100M   K01:1990=163/100M   K01:1990=163/100M
HWI-1KL149:20:C1CU7ACXX:1:1110:15369:59046/1    EQ|EQ|EQ    K01:2213=83/100M    K01:2213=83/100M    K01:2213=83/100M
HWI-1KL149:20:C1CU7ACXX:1:1110:15369:59046/2    EQ|EQ|EQ    K01:2081=163/100M   K01:2081=163/100M   K01:2081=163/100M
HWI-1KL149:20:C1CU7ACXX:1:1111:8599:97362/1 EQ|EQ|EQ    K01:2213=83/100M    K01:2213=83/100M    K01:2213=83/100M
HWI-1KL149:20:C1CU7ACXX:1:1111:8599:97362/2 EQ|EQ|EQ    K01:2081=163/100M   K01:2081=163/100M   K01:2081=163/100M
HWI-1KL149:20:C1CU7ACXX:1:1113:10490:30873/1    EQ|EQ|EQ    K01:2120=83/100M    K01:2120=83/100M    K01:2120=83/100M
HWI-1KL149:20:C1CU7ACXX:1:1113:10490:30873/2    EQ|EQ|EQ    K01:1990=163/100M   K01:1990=163/100M   K01:1990=163/100M
HWI-1KL149:20:C1CU7ACXX:1:1113:12360:36316/1    EQ|EQ|EQ    K01:2213=83/100M    K01:2213=83/100M    K01:2213=83/100M
HWI-1KL149:20:C1CU7ACXX:1:1113:12360:36316/2    EQ|EQ|EQ    K01:2081=163/100M   K01:2081=163/100M   K01:2081=163/100M

(...)

[roll]

this looks like a nice tool. I will try it.

[Yijun Tian]

Jvarkit is the mightiest. Hi Pierre. Do you think it's possible for CmpBams to give leftmost coordinates difference for certain read as well as CIGAR in two different alignment? I am trying to map clipped reads to find some circle split reads.