Tophat2.03: error

[mrfox]

Hi all,

I just tested Tophat2.03 released on May 26th using a small data set (paired-end data, 5000 reads, read len = 50). Unfortunately, I got the error message below. I am not sure if I made something wrong or it is a bug...



[ao@xxxx tmp]$ tophat2 -r 150 -p 20 -g 1 -o ./ ../../bowtie2_index_hg19/hg19 tmp1.fq tmp2.fq

[2012-05-28 23:28:16] Beginning TopHat run (v2.0.3)
-----------------------------------------------
[2012-05-28 23:28:53] Checking for Bowtie
Bowtie version: 2.0.0.6
[2012-05-28 23:28:54] Checking for Samtools
Samtools version: 0.1.18.0
[2012-05-28 23:28:54] Checking for Bowtie index files
[2012-05-28 23:28:54] Checking for reference FASTA file
[2012-05-28 23:28:54] Generating SAM header for ../../bowtie2_index_hg19/hg19
format: fastq
quality scale: phred33 (default)
[2012-05-28 23:30:36] Preparing reads
left reads: min. length=50, max. length=50, 2498 kept reads (2 discarded)
right reads: min. length=50, max. length=50, 2494 kept reads (6 discarded)
[2012-05-28 23:35:13] Mapping left_kept_reads to genome hg19 with Bowtie2
[FAILED]
Error running:
/../tools/tophat-2.0.3/bam2fastx --all --fastq .//tmp/left_kept_reads.bam|/RIS/home/hzhao1/tools/bowtie2-2.0.0-beta6/bowtie2-align -q -k 1 -D 15 -R 2 -N 0 -L 20 -i S,1,1.25 --gbar 4 --mp 6,2 --np 1 --rdg 5,3 --rfg 5,3 --score-min L,-0.6,-0.6 -p 20 --sam-no-hd -x ../../bowtie2_index_hg19/hg19 -|/../tools/tophat-2.0.3/fix_map_ordering --bowtie2-min-score 10 --sam-header .//tmp/hg19_genome.bwt.samheader.sam - .//tmp/left_kept_reads.mapped.bam .//tmp/left_kept_reads_unmapped.bam --index-outfile .//tmp/left_kept_reads.mapped.bam.index

[TrueSEQr]

I have seen this error as well. However, I may have something to add that a potential solver will find relevant.

When I see the error above in the stdout log, I see the following in the "tophat.log" under the log directory.

[2013-07-30 00:03:55] Mapping left_kept_reads_unmapped to genome segment_juncs with Bowtie2 (1/1)
open: No such file or directory
Error: bam2fastx failed to open BAM file ./tophat_out/tmp/left_kept_reads_unmapped.bam
[FAILED]

Further, I see this in the bowtie.left_kept_reads_unmapped.log

0 reads
0.00% overall alignment rate

This seems to be telling me that there is nothing to be captured in the "left_kept_reads_unmapped" file. However, I would expect Tophat to be able to anticipate something like that--maybe make a dummy file. Or, I would expect it somehow to be impossible.

I use version 2.0.8, BTW.

[jp.]

Hi
I have have problem with Tophat2.. I have successfully analysed using tophat and cufflinks with insert size 250.
do you know how to give commands for read length 101 x 2 (insert size 150) ?

I have seen this error as well. However, I may have something to add that a potential solver will find relevant.

When I see the error above in the stdout log, I see the following in the "tophat.log" under the log directory.

[2013-07-30 00:03:55] Mapping left_kept_reads_unmapped to genome segment_juncs with Bowtie2 (1/1)
open: No such file or directory
Error: bam2fastx failed to open BAM file ./tophat_out/tmp/left_kept_reads_unmapped.bam
[FAILED]

Further, I see this in the bowtie.left_kept_reads_unmapped.log

0 reads
0.00% overall alignment rate

This seems to be telling me that there is nothing to be captured in the "left_kept_reads_unmapped" file. However, I would expect Tophat to be able to anticipate something like that--maybe make a dummy file. Or, I would expect it somehow to be impossible.

I use version 2.0.8, BTW.

[TrueSEQr]

Here's the closest thing to a solution to the "unmapped" problem I've found so far: use Bowtie1. There is a --bowtie1 option on tophat v2. Use that, and make sure that bowtie is in your path. Oddly, when I first tried this, I found that tophat complained (and failed to run because of) that the bowtie version was not correct. It stated it had to be at least 0.12 something. I had to compensate by borrowing and copying the tophat script which is writtin in python, and commenting out that check. It seemed to run clear to the end for the first time.

So that's two separate problems: 1--why is it failing to find this file if I run bowtie2 underneath. 2--why doesn't it like bowtie 1.0.0?

[TrueSEQr]

BTW, jp, Sorry to say I am no expert at tophat. I am just trying to learn it now. Still, I would encourage you to start your own thread on this topic, because that question is quite different from the one I started here. It might get more attention that way, too.

[TrueSEQr]

Bowtie1 is not universally solving this. I may have a bad install, here. I have downloaded data that was used for testing in writing the original TopHat paper, but that failed. Plus the strange workarounds I have had to do lead me to believe something in the software stack is not right. The "tophat test data" I downloaded, seemed to work just fine.

So, bottom line: this was not strictly a bowtie2 / tophat 2 issue. Could also be the index I used, which I had built myself (I did make sure to build one such index per bowtie version, and point the tophat at the relevant version). But, then again, I have not yet found the "explicit instructions" for building such indices, and could have left a parameter out.

[TrueSEQr]

I finally resorted to going through the tophat2.py code. I am not a Python programmer, but was able to get a hint. When I used the command-line option "--keep-tmp", this problem with the disappearing left_kept_reads_unmapped.bam did not occur.

Instead, I see this:

"Error while running bowtie:
Error while flushing and closing output"

and that last line is repeated seven times (BTW, I ran tophat with eight threads).

So this raises the question: why does the left_kept_reads_unmapped.bam file get deleted if you are not clearing away temporary files, even though there are still downstream dependencies on that file? When I ran this against the test data, I did not have to use --keep-tmp to get it to run.

Still a deeper problem.