Write Failure In Tophat 1.3.1 X86_64

Question

[Gabrielw]

using tophat 1.3.1 x86_64...

this is my entire command:

tophat -o /storage/gwajnberg/outputtophat/ \
          /storage/gwajnberg/bowtie_splitted/ \
          /storage/gwajnberg/SRP000614/SRX002932/SRR013983.fastq

where .../bowtie_splitted/ is the directory with the reference .fa file that originated the bowtie indexes files ( they are in the same directory). My reads have between 250 and 300 bp.

thats what happened:

[Fri Jul  1 12:49:59 2011] Beginning TopHat run (v1.3.1)
-----------------------------------------------
[Fri Jul  1 12:49:59 2011] Preparing output location ./tophat_out/
[Fri Jul  1 12:49:59 2011] Checking for Bowtie index files
[Fri Jul  1 12:49:59 2011] Checking for reference FASTA file
    Warning: Could not find FASTA file /storage/gwajnberg/bowtie_splitted/.fa
[Fri Jul  1 12:49:59 2011] Reconstituting reference FASTA file from Bowtie index
Executing: /storage/app/bowtie-0.12.5/bowtie-0.12.5/bowtie-inspect /storage/gwajnberg/bowtie_splitted/ > ./tophat_out/tmp/.fa
[Fri Jul  1 13:07:57 2011] Checking for Bowtie
    Bowtie version:             0.12.5.0
[Fri Jul  1 13:07:57 2011] Checking for Samtools
    Samtools Version: 0.1.16
[Fri Jul  1 13:07:58 2011] Generating SAM header for /storage/gwajnberg/bowtie_splitted/
[Fri Jul  1 13:08:00 2011] Preparing reads
    format:         fastq
    quality scale:     phred33 (default)
    Left  reads: min. length=25, count=121868
[Fri Jul  1 13:08:08 2011] Mapping left_kept_reads against  with Bowtie 
[Fri Jul  1 13:08:40 2011] Processing bowtie hits
[Fri Jul  1 13:10:32 2011] Mapping left_kept_reads_seg1 against  with Bowtie (1/33)
.......
[Fri Jul  1 13:34:37 2011] Mapping left_kept_reads_seg33 against segment_juncs with Bowtie (33/33)
[Fri Jul  1 13:34:38 2011] Joining segment hits
sort: write failed: standard output: Broken pipe
sort: write error
[Fri Jul  1 13:37:11 2011] Reporting output tracks
    [FAILED]
Error: [Errno 2] No such file or directory
Found 0 junctions from happy spliced reads

Does anyone know how to help me? thanks!

[Michael Schubert]

A quick guess: you forgot to provide the path to your reference fasta file.

[David Quigley]

Haven't run tophat (yet), but it looks like the "joining segment hits" step failed. Broken Pipe means one command was expecting to eat the output of another command via a Unix pipe, and one of the two commands failed. Then tophat tried to write some output at "Reporting output tracks", but that probably failed because of the earlier problem.

[Gabrielw]

Well, there isn´t any option to specify where is the fasta file....the command is :

tophat [options] [?] < read file> there is only an option to choose the output directory, but its not necessary, because there is a default directory!

somewhere I read that the reference file ( fa or fasta) must be in the same directory of bowtie indexes... and that´s what I did!

I built the indexes from this specified index! Somwhere I also read that if there isn´t a ref file, the programa builds a reference file from the recognized indexes files. I´ll try some othere things later, s

Answer 1

[seidel]

From the tophat documentation: When you run tophat, the first argument should be the full path to the directory containing the index plus the prefix of the index files.

Your command has a path, but no index listed. For instance, if you're studying Nematostella, and your index files all start with nematostella, then presumably you would also have a fasta file with that same prefix. So your index argument would look like this: /storage/gwajnberg/bowtie_splitted/nematostella

[Gabrielw]

I´m moved my ref to /bowtie_splitted/hg19 no my reference name is hg19.fa and I´m building a new index from it...with prefix hg19 .... so I´ll try to run tophat again tomorrow ...so pray :)

[Gabrielw]

I voted for you because your answer helped me to understand a little bit more, but I realized that my installation isn 't working with the test file from tophat

Answer 2

[Docroberson]

I think Michael is probably right.

Warning: Could not find FASTA file /storage/gwajnberg/bowtie_splitted/.fa

The fasta file was not specified correctly. The error is actually rather cryptic. If specifying the FASTA file correct doesn't work try posting the command you used to dissect any problems there.

[Gabrielw]

Well, there isn´t any option to specify where is the fasta file....the command is :

tophat [options] [?] < read file> there is only an option to choose the output directory, but its not necessary, because there is a default directory!

somewhere I read that the reference file ( fa or fasta) must be in the same directory of bowtie indexes... and that´s what I did!

I built the indexes from this specified index! Somwhere I also read that if there isn´t a ref file, the programa builds a reference file from the recognized indexes files. I´ll try some othere things later, s

Answer 3

[Darked89]

You have to post your entire command used to run tophat. The message:

Could not find FASTA file /storage/gwajnberg/bowtie_splitted/.fa

indicates that you may have not specified your bowtie index correctly. If your index has the name prefix: your_genome.bowtie then your reference fasta should be your_genome.bowtie.fa. BTW, if the your_genome.bowtie.fa is absent it should be not a big deal except for some wasting some CPU time for recreating it from indexes.

The more troubling thing is the length of your sequences. If you got 25bp reads these can not be split with tophat, for sure not running with default options, and possibly not even with "--segment-length 12" if you do not have genome annotation and large genome. Either use untruncated, full length reads (preferably longer than 50bp) or split reads into 50bp+ vs shorter ones and map these in different modes (spliced with tophat, unspliced with bowtie).

Last but not least: update to bowtie 0.12.7

[Gabrielw]

size of reads : between 250 and 300 bp .fa file in the same directory of bowtie indexes .... using this command : >tophat [options] [?] < read file>

Answer 4

[Gabrielw]

this is my entire command: tophat -o /storage/gwajnberg/outputtophat/ /storage/gwajnberg/bowtie_splitted/ /storage/gwajnberg/SRP000614/SRX002932/SRR013983.fastq

where .../bowtie_splitted/ is the directory with the reference .fa file that originated the bowtie indexes files ( they are in the same directory). My reads have between 250 and 300 bp...

I'm trying using 0.12.7 version of bowtie, I was using 0.12.5 ....

[Gabrielw]

the same thing happened :(

[Darked89]

/storage/gwajnberg/bowtie_splitted/ does not correspond to the bowtie index prefix. You are missing the indexed genome for mapping.

Answer 5

[Gabrielw]

I Tried what everyone said and continued happening this.... my splitted index bowtie is with prefix hg19 and my fasta file is hg19.fa in the same directory.... :(

[seidel]

Can you verify if your installation of tophat works? At http://tophat.cbcb.umd.edu/tutorial.html there is a section called: "Testing the Installation". It's a simple test. If that works, you can add complication until it breaks. If nothing else, at least you made it one step further.

[Gabrielw]

you were right !!! I used "testing the installation" and happened the same thing!!! thanksss now I'll try to install again :(

Answer 6

[Gabrielw]

Well thanks all for helping me....that's what happened... tophat 1.3.1 x86_64 and tophat 1.3.0 x86_64 are bugged and when I used them with the test data from tophat homepage , it happened the same error that happened with my files... so I downloaded the version 1.2.0 x86_64 and worked very well... thanks all for the support :)

[Rodrigo]

Sorry, should have comment here: Hey gabrielw. Where did you found that tophat 1.3.1 x86_64 and tophat 1.3.0 x86_64 are bugged?

Thanks!

Answer 7

[Rodrigo]

Hey gabrielw. Where did you found that tophat 1.3.1 x86_64 and tophat 1.3.0 x86_64 are bugged?

Thanks!