Tophat bug -- excessively long match

[ebartom]

I'm running Tophat v1.2.0 with Bowtie 0.12.7.0 and Samtools 0.1.12a, on a particular RNAseq sample. Everything runs correctly until Tophat tries to use Samtools, because it creates this nonsensical alignment that chokes samtools:

HWUSI-EAS552R_0001:1:4:10279:10944#0 163 chr17 7357209 255 78M97N536870908M92N22M = 7357535 0 CAACTATAGTCCCACATCACCCAGCTATTCGCCAACTTCACCCAGCTACTCACCCACTTCTCCCAGCTACTCACCTACCTCTCCAAGCTACTCACC CCCCCCCCCCCCCCCCCCCCCCCCCCCCCC;CC?CC?CBACCCCCCCC2C7*;;'8?2??=@?=CC@@=BA<AA@+:>50::>A+:>@:A###### NM:i:5 XS:A:- NH:i:1

If you look at the CIGAR string you can see the problem. There's a match region that's 536870908 bp in length, in the middle of this 96 bp read. Obviously a bug.

Has anyone else encountered this? Know if it's specific to this version of tophat? Since tophat only chokes at the final conversion of accepted_hits.sam to accepted_hits.bam, I'm planning on filtering the accepted_hits.sam to remove this, then use samtools to do the .bam conversion outside of tophat.

Is there a better place to report tophat bugs?

[ayfchang]

Similar problem noted in Tophat v1.1.4 as well

http://seqanswers.com/forums/showthread.php?t=8817&highlight=tophat

[jb2]

Yep, in order to handle this, I just wrote a quick script to remove the read with the buggy CIGAR string and then ran samtools from there to create the bam files. You should be able to find the .sam files in the tophat tmp folder and can go from there.

[Daehwan]

I'm running Tophat v1.2.0 with Bowtie 0.12.7.0 and Samtools 0.1.12a, on a particular RNAseq sample. Everything runs correctly until Tophat tries to use Samtools, because it creates this nonsensical alignment that chokes samtools:

HWUSI-EAS552R_0001:1:4:10279:10944#0 163 chr17 7357209 255 78M97N536870908M92N22M = 7357535 0 CAACTATAGTCCCACATCACCCAGCTATTCGCCAACTTCACCCAGCTACTCACCCACTTCTCCCAGCTACTCACCTACCTCTCCAAGCTACTCACC CCCCCCCCCCCCCCCCCCCCCCCCCCCCCC;CC?CC?CBACCCCCCCC2C7*;;'8?2??=@?=CC@@=BA<AA@+:>50::>A+:>@:A###### NM:i:5 XS:A:- NH:i:1

If you look at the CIGAR string you can see the problem. There's a match region that's 536870908 bp in length, in the middle of this 96 bp read. Obviously a bug.

Has anyone else encountered this? Know if it's specific to this version of tophat? Since tophat only chokes at the final conversion of accepted_hits.sam to accepted_hits.bam, I'm planning on filtering the accepted_hits.sam to remove this, then use samtools to do the .bam conversion outside of tophat.

Is there a better place to report tophat bugs?

I thought I fixed the bug, but it still exists. I'll be grateful if you provide me with your data set so I can fix it? (Daehwan Kim - [email protected])

BTW, you can report a bug to [email protected]