malloc error running cufflinks

32 views

Skip to first unread message

Madeleine Aase-Remedios

unread,

Oct 12, 2023, 5:20:52 AM10/12/23

to Tuxedo Tools Users

I am running cufflinks on my HPC with a slurm script for 3 datasets, using the same script for each. One dataset ran successfully for a few hours while the other two encounter this error:

[13:37:24] Loading reference annotation.

[13:37:28] Inspecting reads and determining fragment length distribution.

> Processed 12109 loci. [*************************] 100%

Warning: Using default Gaussian distribution due to insufficient paired-end reads in open ranges. It is recommended that correct parameters (--frag-len-mean and --frag-len-std-dev) be provided.

> Map Properties:

> Normalized Map Mass: 58409.00

> Raw Map Mass: 58409.00

> Fragment Length Distribution: Truncated Gaussian (default)

> Default Mean: 200

> Default Std Dev: 80

[13:37:32] Assembling transcripts and estimating abundances.

> Processing Locus NC_052520.2:14287409-143368 [*** ] 13%malloc(): invalid size (unsorted)

/var/spool/slurmd/job6265637/slurm_script: line 14: 3414828 Aborted cufflinks --num-threads=2 -o ${outdir}${OUT} -g $gtf $BAM

I also tried to run it on a single file, in case something in my script were causing an error, and got the same error:

> Processing Locus NC_052520.2:14287409-143368 [** ] 11%malloc(): invalid size (unsorted)

srun: error: cn025: task 0: Aborted

It always crashes at the same point, at that locus.

The two datasets that failed are both from Drosophila simulans, so are using the same reference genome and annotation, while the one that succeeded are D. mauritiana. I looked into that locus and there are two things going on that could be the reason for this error. One, it's annotated as a tRNA so maybe it could have a very high number of reads mapped to it that is overloading the memory? Or, two, the locus is complex, with overlapping transcripts, which is causing the problem.