Tophat 2.0 error

[EGrassi]

Hello, after 1 week of alignments (100bp paired reads, 176mil of them, 6 cores was all I had...) tophat2.0 failed after the last step of alignments (I believe):

[2012-05-22 11:16:12] Joining segment hits
[2012-05-22 14:23:48] Reporting output tracks
Traceback (most recent call last):
File "/home/data/rnaseq/bin/tophat", line 3778, in <module>
sys.exit(main())
File "/home/data/rnaseq/bin/tophat", line 3750, in main
params.gff_annotation)
File "/home/data/rnaseq/bin/tophat", line 2526, in compile_reports
os.rename(sorted_bam_parts[0], output_dir + "accepted_hits.bam")
IndexError: list index out of range
make: *** [S1.th] Error 1


I'm going to investigate if in the /tmp dir I will find all the needed files to go on with my analysis but I wanted to know if this is a known issue and if it is possible to fix it in some way...

The command line was:
tophat -p 5 -G /rogue/bioinfotree/prj/ewing-rnaseq/local/share/data/Homo_sapiens/UCSC/hg19/Annotation/Genes/genes.gtf --transcriptome-index=transcriptome -o S1.th ../alignment/expanded_genome2 <(zcat /rogue/bioinfotree/R1_001.fastq.gz),list of 11 fastq <(zcat /rogue/bioinfotree/R2_001.fastq.gz),list of 11 fastq

[EGrassi]

Big datasets (~200M paired reads, 100bp) with tophat 2.0

Ok, just to point other people in the right direction: thanks to the quick reply of the tophat developers I discovered that this was an out of memory issue...in this case tophat_reports failed with an alloc error but the python wrapper went on and crashed because there were no accepted_hits files in the tmp dir. I'm trying to avoid losing all that cpu time reproducing by hand on a machine with more memory the last steps of the analysis, if anyone is interested ask me.

The developers also suggested me for datasets of this size to split up fastq in fragments of 40~60M to avoid this kind of problems.
The funny thing is that for the first couple of splitted fastq (which were present in my first big run, this is somewhat puzzling) tophat now fails earlier in the pipeline, at the segment_juncs portion:

[2012-05-23 12:20:30] Searching for junctions via segment mapping
        [FAILED]
Error: segment-based junction search failed with err =1
Error: could not get read# 1 from stream!

I've found this error cited in several threads but with old tophat versions or suggesting causes that here don't apply (it is not a memory issue, I've checked, not a permissions issue...). Any clue?