Tophat Slow : Searching For Junctions Via Segment Mapping

Question

[Nicolas Rosewick]

Hi,

Is it normal that my tophat jobs are running the step "Searching for junctions via segment mapping" since 4 days ? I'm running tophat2 with 76 bp paired-end reads with -p 4 and 15Gb of memory. here's my log file :

[2012-05-03 11:19:56] Beginning TopHat run (v2.0.0)
-----------------------------------------------
[2012-05-03 11:19:56] Checking for Bowtie
          Bowtie version:     2.0.0.5
[2012-05-03 11:19:56] Checking for Samtools
        Samtools version:     0.1.18.0
[2012-05-03 11:19:56] Checking for Bowtie index files
[2012-05-03 11:19:56] Checking for reference FASTA file
[2012-05-03 11:19:56] Generating SAM header for DB
    format:         fastq
    quality scale:     phred33 (default)
[2012-05-03 11:20:00] Preparing reads
     left reads: min. length=76, count=14027314
    right reads: min. length=76, count=14048352
[2012-05-03 11:34:13] Mapping left_kept_reads against DB with Bowtie2 
[2012-05-03 13:07:51] Mapping left_kept_reads_seg1 against DB with Bowtie2 (1/3)
[2012-05-03 13:59:08] Mapping left_kept_reads_seg2 against DB with Bowtie2 (2/3)
[2012-05-03 14:58:46] Mapping left_kept_reads_seg3 against DB with Bowtie2 (3/3)
[2012-05-03 16:02:58] Mapping right_kept_reads against DB with Bowtie2 
[2012-05-03 17:37:00] Mapping right_kept_reads_seg1 against DB with Bowtie2 (1/3)
[2012-05-03 18:24:02] Mapping right_kept_reads_seg2 against DB with Bowtie2 (2/3)
[2012-05-03 19:16:00] Mapping right_kept_reads_seg3 against DB with Bowtie2 (3/3)
[2012-05-03 20:25:54] Searching for junctions via segment mapping

Anyone has the same problem ?

Thanks

[Sean Davis]

Does the output of "top" show that the process is actually using compute resources? What is the memory usage showing?

[Nicolas Rosewick]

here's when I do a qstat on my sge job : cpu=17:10:27:53, mem=2525712.89942 GBs, io=544.86174, vmem=2.282G, maxvmem=4.171G. I don't know how to know the memory usage in real time (like in Top) on a sge grid

[Sean Davis]

I should have been less specific. The question is whether your job is swapping. You may need to ask a system administrator for help in answering this question.

Answer 1

[Nicolas Rosewick]

I think I found the problem. On of my job successfully ended this morning. But when I check the sge log file, I observed that the job used : Max vmem = 20.210G but I only allowed 15Go for the job. So I think it will use a lot of swap in/ swap out time to finish the process. I will try with much more memory .