Hello All,
I have 20 paired read files (RNA-Seq by Illumina) that range from 10 Gb to 30 Gb.
Chimerascan works on the majority of them, but for 3 it runs for days then the system kills it. The log files suggest that the initial runs complete through. The program completes for files as large as 30Gb as well as smaller files near 10Gb but fails on 3 intermediate sized files. I though it is a memory problem and I split the files into two smaller files but the problem always occurs.
Does someone know why?
Best,
Luciano
If the size is not the reason, i might suggest to search for strange or unwanted chars inside the read files that are killed.
Also: do you have any log or response by the system or the code when it is killed?
Hello Paolo, Thank you for replying. Here you go the last line of log file:
OSError: [Errno 12] Cannot allocate memory
I am gonna check strange chars into the files.
Hello Rohan,
Here you go the files created by the algorithm:
-rw-rw-r-- 1 cascione cascione 2.6G Jan 21 15:14 aligned_reads.bam
-rw-rw-r-- 1 cascione cascione 6.1K Jan 21 16:07 isize_dist.txt
drwxrwxr-x 2 cascione cascione 137 Jan 21 22:59 log
-rw-rw-r-- 1 cascione cascione 1.4K Jan 21 14:35 runconfig.xml
-rw-rw-r-- 1 cascione cascione 3.4G Jan 21 16:05 sorted_aligned_reads.bam
-rw-rw-r-- 1 cascione cascione 8.3M Jan 21 16:06 sorted_aligned_reads.bam.bai
drwxrwxr-x 2 cascione cascione 4.0K Jan 21 22:59 tmp
I used the default settings and I am go to re-run following your suggestions. Thank you very much.
Bye,
Luciano
Let's see if Rohan suggestion works, otherwise: Chimerascan should be python, and your problem could depend on your Os version and Python version.
So we can check:
Is it Fedora or another OS? Which version?
What python version are you using?
Ciao Luciano,
even if you tried with smaller input files the problem could still be related to memory if Chimerascan has memory leaks problems. It can happen also with Python programs if they rely on external programs or library for example.
Googling a bit I found this tool that might be handy to trace down Chimerascan and see if memory leaks is really the problem.
http://www.lshift.net/blog/2008/11/14/tracing-python-memory-leaks
Hello all,
I tried the Rohan suggestion but the computation has been killed. I could conclude there is a memory leak somewhere. It seems very strange to me, the algorithm works properly for 20 samples except for 3 samples. I am going to identify the leak.
For Paolo, my OS is a RedHat 5 and Python is 2.7.
Thank you all for your suggestions,
bye,
Luciano
Yes, Rohan suggestion about the TMP dir could be also the deal braker :) i had issues with fedora, but redhat should be fine. Let us know!
Thanks again for your help guys,
The Computation crashes after "Extracting single-mapped reads that may span breakpoints". Here you go the last two lines of log file:
2014-01-29 11:11:02,535 - root - INFO - Extracting single-mapped reads that may span breakpoints
2014-01-29 11:11:02,535 - root - DEBUG - Matching single-mapped frags to 5' chimeras
I am sorry but i did not find any options in chimerascan to set the tmp directory.
Hi Rohan,
The file you mentioned "bowtie_single_spanning.log" is not present into log folder of the samples which are not getting completed but it is present for the others samples (computation completed).
Dear All,
Finally I fixed the problem. I switched my bowtie version, bowtie 2.0.4 to bowtie 0.12.9, and the problem disappeared. Thank you for your suggestions.
Bye,
Luciano
- Badges
- Science method