Possible tophat 2.0.6 bug?

[panagiotis]

Hi all,

I am trying to use tophat to map SOLiD colorspace pair-end (50x35) reads with tophat. However, when I tried with the standard procedure which looks something like:

tophat --color --quals --num-threads 8 --coverage-search --library-type fr-secondstrand --output-dir ./tophat_out --prefilter-multihits --keep-tmp --GTF /opt/iGenomes/Homo_sapiens/UCSC/hg19/Annotation/Genes/genes.gtf --bowtie1 /opt/NGSTools/Bowtie/indexes/hg19_c ./CON_BR1_F3.csfasta ./CON_BR1_F5.csfasta ./CON_BR1_F3_QV.qual ./CON_BR1_F5_QV.qual

I got the following error:

[2012-12-20 16:07:24] Beginning TopHat run (v2.0.6)
-----------------------------------------------
[2012-12-20 16:07:24] Checking for Bowtie
		  Bowtie version:	 0.12.8.0
[2012-12-20 16:07:24] Checking for Samtools
		Samtools version:	 0.1.18.0
[2012-12-20 16:07:24] Checking for Bowtie index files
[2012-12-20 16:07:24] Checking for reference FASTA file
	Warning: Could not find FASTA file /opt/NGSTools/Bowtie/indexes/hg19_c.fa
[2012-12-20 16:07:24] Reconstituting reference FASTA file from Bowtie index
  Executing: /opt/NGSTools/Bowtie/bowtie-inspect /opt/NGSTools/Bowtie/indexes/hg19_c > ./tophat_out/tmp/hg19_c.fa
[2012-12-20 16:11:08] Generating SAM header for /opt/NGSTools/Bowtie/indexes/hg19_c
	format:		 fasta
[2012-12-20 16:11:30] Reading known junctions from GTF file
[2012-12-20 16:11:36] Pre-filtering multi-mapped left reads
[2012-12-20 16:11:36] Mapping CON_BR1_F3 to genome hg19_c with Bowtie 
	[FAILED]
Error running bowtie:
Error: reads file does not look like a FASTA file
terminate called after throwing an instance of 'int'

I checked and re-checked many things regarding my file formats and checked the logs. Eventually, I found that the tophat wrapper is launching prep_reads which does not produce fasta output (rather fastq output), while the subsequent bowtie command assumes fasta format from stdin. So what I did was to edit tophat.py, and commented out the lines 2164-2165

        #elif reads_format == "fasta":
        #    bowtie_cmd += ["-f"]

and saved to my_tophat.py in order to use specifically with my data.

This was not of course the right way to do, so my question is, am I doing something wrong, or somebody else here has encountered this before. Is it a tophat bug?

Thanks!

[panagiotis]

My workaround fixes the problem but I also just noticed that the problem does not appear without the --prefilter-multihits option and the untouched tophat python script. So it is quite possible that it is a tophat 2.0.6 bug with csfasta input and the --prefilter-multihits option...