Cufflinks 2.0.2 segmentation fault

[bmicro_mit1]

I am using 100 bp paired end Illumina Hi-seq data with about 50M reads and trying to use tophat / cufflinks for RNA-seq analysis for human data, using Ensemble v68 gtf along with Gencode v13 lncRNA gtf annotations. These files were concatenated together to run both tophat 2.0.6 with bowtie2 2.0.6:

tophat -p 4 --solexa1.3-quals --read-realign-edit-dist 0 --no-novel-juncs --library-type fr-unstranded -G $GTF -o $OUT $GENOME $FASTQ_1 $FASTQ_2

and Cufflinks 2.0.2
cufflinks -o $OUT -p 4 -G $GTF -b $FASTA --multi-read-correct $OUT/accepted_hits.bam

A segmentation fault has continued to occur with multiple samples at similar locations as Cufflinks is re-estimating abundacnes with bias and multi-read correction. Below is the output:

[09:41:21] Learning bias parameters.
> Processed 635 loci. [*************************] 100%
[09:45:58] Re-estimating abundances with bias and multi-read correction.
> Processing Locus chr16:5289802-6826015 [******* ] 31%Segmentation fault (core dumped)

Any input would be greatly appreciated.

[GenoMax]

How much RAM do you have on this machine?

[bmicro_mit1]

This was run on a couple of different machines in a SGE cluster. Some of the nodes had up to 48Gb of RAM, but in the SGE email reporting the program had died, it never reported more than 5Gb of memory usage.

[GenoMax]

Are there any other error messages (stderr output)?

[bmicro_mit1]

No, just the segmentation fault. I am running with verbose mode right now to see if there is more output.

[GenoMax]

Are you using the -o and -e directives with your qsub (or SGE job submission script) to capture the output/stderr output? Contents of those would be useful as well.

[bmicro_mit1]

Yes, nothing in the 'o' file, and only what I had put in the initial post from the 'e' file.

[chrisjohn86]

I have had a lot of segmentation faults with tuxedo over the last few weeks. I finally figured out it was due to bad RAM, I removed 8GB of 16GB total and it started working fine. Memoryxp is a good RAM diagnostic tool. This is one possible reason for seg faults. There are others:

http://www.cyberciti.biz/tips/segmentation-fault-on-linux-unix.html

[Anelda]

I have found that the order of the reference chromosomes in the genome.fasta file and the chromosomes in the GFF/GTF file, must be exactly the same otherwise a segmentation fault occurs. This is specifically valid in the case of Cufflinks. To demonstrate..

grep ">" genome.fasta > fasta.order
cut -f 1 genome.gff | uniq > gff.order

diff fasta.order gff.order

If the order of the chromosomes are not the same, you'll have to reshuffle. Easiest might be to reshuffle the GFF/GTF - I'm not sure if there are any scripts that can sort fasta/gff files. I just grep each chromosome from the GFF file and send it to a separate file, then cat the individual chromosome.gff files in the correct order and create new genome.gff.

Hope this helps someone!

[sterding]

I have found that the order of the reference chromosomes in the genome.fasta file and the chromosomes in the GFF/GTF file, must be exactly the same otherwise a segmentation fault occurs.

Thanks. I am testing this. I am also curious how you found the trick

[sterding]

hi Anelda,

I made the genome.fa and gtf file in the same order, but still I got the " Segmentation fault" error in the step of " Learning bias parameters" if I use -b option. Without the "-b" option, I don't get the error. So, I think the bug is in the "-b" option. Hopefully cufflinks team can get attention to the problem.

[biocomputer]

hi Anelda,

I made the genome.fa and gtf file in the same order, but still I got the " Segmentation fault" error in the step of " Learning bias parameters" if I use -b option. Without the "-b" option, I don't get the error. So, I think the bug is in the "-b" option. Hopefully cufflinks team can get attention to the problem.

I'm using Cufflinks 2.2.1 and having the same problem with cufflinks and cuffdiff. -b causes a segfault, it works fine without it. I ensured the genome.fa and gtf file have their chromosomes in the same order and contain the same chromosomes and there is lots of free memory available.

[offspring]

Please file an issue report at https://github.com/cole-trapnell-lab/cufflinks containing a description of the problem and how to reproduce it, otherwise the cufflinks team won't even be aware of the problem.

[biocomputer]

I was able to solve my problem. Despite the "-b genome.fa" seemingly being the cause of the problem it's actually the .gtf file. See here how to modify the .gtf file:

https://groups.google.com/d/msg/tuxedo-tools-users/FTKA4qozJIc/p47AwnCXxvwJ

https://groups.google.com/d/msg/tuxedo-tools-users/V9eI65dVzyU/LJhbCHBsITAJ