Incompatibility tophat2.0.11 - bowtie2.2.3

[Dario Beraldi]

Hello,

(First, many thanks for making tophat/bowtie available!)

I might have found an incompatibility between tophat2.0.11 and bowtie2.2.3.

I get the error below:

tophat2 --no-coverage-search --no-novel-juncs -o ./tmp -p 8 genomes/iGenomes/Homo_sapiens/UCSC/hg19/Sequence/Bowtie2Index/genome reads.fq.gz

[2014-06-03 08:50:41] Beginning TopHat run (v2.0.11)
-----------------------------------------------
[2014-06-03 08:50:41] Checking for Bowtie
          Bowtie version:     2.2.3.0
[2014-06-03 08:50:41] Checking for Samtools
        Samtools version:     0.1.18.0
[2014-06-03 08:50:41] Checking for Bowtie index files (genome)..
[2014-06-03 08:50:41] Checking for reference FASTA file
[2014-06-03 08:50:41] Generating SAM header for /data01/sblab/users/berald01/reference_data/genomes/iGenomes/Homo_sapiens/UCSC/hg19/Sequence/Bowtie2Index/genome
[2014-06-03 08:50:41] Preparing reads
     left reads: min. length=50, max. length=50, 1081452 kept reads (14 discarded)
[2014-06-03 08:50:59] Mapping left_kept_reads to genome genome with Bowtie2
    [FAILED]
Error running bowtie:
Error while flushing and closing output
terminate called after throwing an instance of 'int'
(ERR): bowtie2-align died with signal 6 (ABRT)


While with tophat2.0.11 and bowtie2.2.1 it doesn't happen (see below). I'm using the indexes obtained from iGenomes. I also tried to recreate the indexes with bowtie2.2.3 but it made no difference (in fact, the new indexes where identical to the original ones.)
For both tophat and bowtie I'm running the pre-compiled binaries you made available. The OS I'm on is "Red Hat Enterprise Linux Server release 5.4 (Tikanga)".

tophat2 --no-coverage-search --no-novel-juncs -o ./tmp -p 8 genomes/iGenomes/Homo_sapiens/UCSC/hg19/Sequence/Bowtie2Index/genome reads.fq.gz

[2014-06-03 08:56:36] Beginning TopHat run (v2.0.11)
-----------------------------------------------
[2014-06-03 08:56:36] Checking for Bowtie
          Bowtie version:     2.2.1.0
[2014-06-03 08:56:36] Checking for Samtools
        Samtools version:     0.1.18.0
[2014-06-03 08:56:36] Checking for Bowtie index files (genome)..
[2014-06-03 08:56:36] Checking for reference FASTA file
[2014-06-03 08:56:36] Generating SAM header for /data01/sblab/users/berald01/reference_data/genomes/iGenomes/Homo_sapiens/UCSC/hg19/Sequence/Bowtie2Index/genome
[2014-06-03 08:56:38] Preparing reads
     left reads: min. length=50, max. length=50, 1081452 kept reads (14 discarded)
[2014-06-03 08:56:56] Mapping left_kept_reads to genome genome with Bowtie2
[2014-06-03 08:57:49] Mapping left_kept_reads_seg1 to genome genome with Bowtie2 (1/2)
[2014-06-03 08:57:58] Mapping left_kept_reads_seg2 to genome genome with Bowtie2 (2/2)
Warning: junction database is empty!
[2014-06-03 08:58:07] Reporting output tracks
-----------------------------------------------
[2014-06-03 09:00:49] A summary of the alignment counts can be found in ./tmp/align_summary.txt
[2014-06-03 09:00:49] Run complete: 00:04:13 elapsed

Can you reproduce this issue? Apologies if I missed something...

All the best
Dario

[Jenny Barna]

On Tuesday, June 3, 2014 9:12:20 AM UTC+1, Dario Beraldi wrote:

[...]
 

I might have found an incompatibility between tophat2.0.11 and bowtie2.2.3.

Error running bowtie:
Error while flushing and closing output
terminate called after throwing an instance of 'int'
(ERR): bowtie2-align died with signal 6 (ABRT)

[...]

We got the same error and I can confirm it appeared with 2.2.3 and is not there with 2.2.2. It is not bowtie2 per se. The new bowtie2.2.3 works fine with indexes build in March. We get the new error with data that works fine with bowtie2.2.2, keeping everything else constant (server, OS, index, tophat2).

Thanks.

[Christine Brennan]

I'm seeing the same error with running tophat-2.0.11 & bowtie-2.2.3 - I had thought it might be because I'm running the --fusion-search option, but from your post it sounds like that might not be the problem:

 tophat --fusion-search --num-threads 8 --library-type fr-firststrand --transcriptome-index /TranscriptomeIndex --GTF /Homo_sapiens/Ensembl/GRCh37/Annotation/Archives/archive-2013-03-06-14-23-04/Genes/genes.gtf  /Homo_sapiens/Ensembl/GRCh37/Sequence/Bowtie2Index/genome read1.fq.gz read2.fq.gz
...
...
...

Error while flushing and closing output
terminate called after throwing an instance of 'int'

(ERR): bowtie2-align died with signal 6 (ABRT) (core dumped)

[Tonis Org]

I can second to that - getting exactly the same error with running tophat-2.0.11 & bowtie-2.2.3. It works with bowtie 2.2.2. 

[Christopher Sifuentes]

Same here. Thanks for posting how you figured this out!

[mattb...@gmail.com]

Hi All,

There's a bug in Tophat 2.0.11 bowtie header generation. Bowtie2 2.2.3 is more stringent in requiring the "-x" parameter when specifying the /path/to/ref/index and fails, causing an incomplete *.samheader.sam file to be created which breaks downstream steps.  Attached is a patch that adds the "-x" option prior to listing the path to the reference index. It should fix the problem. I've confirmed that it works fine with Bowtie 2.2.2 as well.

--- tophat.orig 2014-06-13 16:37:07.987360700 -0500
+++ tophat      2014-06-13 16:55:55.543802747 -0500
@@ -1415,7 +1415,7 @@
         if read_params.color:
             bowtie_header_cmd.append('-C')
 
-        bowtie_header_cmd.extend([idx_prefix, '/dev/null'])
+        bowtie_header_cmd.extend(['-x', idx_prefix, '/dev/null'])
         if noSkip:
            subprocess.call(bowtie_header_cmd,
                    stdout=temp_sam_header_file,

If there's a better place to submit the patch please let me know.

Matt

[Rohan Chaubal]

Thanks a bunch Matt! How does one "apply" this patch? Should I edit a particular file and write this in that file, or run this as a shell script or edit the source code?

[Rohan Chaubal]

Geez. stupid question. sorry for the blonde moment :/ fixed it. works beautifully! thanks for the patch!

[mattb...@gmail.com]

Glad you got it working :)

I did further testing and noticed that this patch breaks bowtie 1 integration. It's such a small patch it shouldn't be too hard to fix, but I'll need to take a closer look. In the meantime, it should be fine for bowtie 2.

Matt

[Boyko]

Hi Matt,

This older thread should help:

https://groups.google.com/forum/#!topic/tuxedo-tools-users/vG5Rn0IgxoA

Bowtie 1 does not accept the -x parameter, so there needs to be a quick check for the version in use beforehand.

Best,
Boyko

[Rachel Raynes]

Hello!  I am getting a similar problem, but I actually do need to be told where to apply the patch. (I am COMPLETELY new to this.)  Can you point me in the right direction?