rsem-calculate-expression with STAR takes forever but without error

Question

[Fleur]

Hi all,

I am running rsem-calculate-expression after succesfully generating the reference files with rsem-prepare-reference. However, the command never proceeds past: "started mapping" like in the example below

Sep 16 08:44:43 ..... started STAR run

Sep 16 08:44:43 ..... loading genome

Sep 16 08:44:46 ..... started mapping

The code that I'm using is this:

rsem-calculate-expression -p 14 --paired-end --star HNLMFDSXY_104305-001-002_CGATGCTT-CGTAGATG_L003_R1.fastq HNLMFDSXY_104305-001-002_CGATGCTT-CGTAGATG_L003_R3.fastq ref/human_ensembl results/CD4_CLL2

I've let it run for days on end but it doesn't finish, any clue on what it going wrong or how to troubleshoot? Since I don't get an error either, it's a bit a mystery what is going on.

I'm using a SSH with 16 cores and 64 GB RAM and the sample is paired-end with ~20M reads.

Thanks a lot for any help or suggestions.

Best, Fleur Peters

[godth13teen]

I think you can try STAR mapping first with option to output transcriptome bam, and then run RSEM on it

[swbarnes2]

Indeed, the options for STAR are hard coded into RSEM; if you want to choose your running options, you need to run it separately.

[Fleur]

Thanks both! I will try :) :)

Answer 1

[Michael]

Few things here:

1. When you run a job via a remote login, always run the job via nohup, that's what I normally do:

   nohup mycommand & tail -f nohup.out
   

or better

    nohup nice -5 mycommand &

if you are sharing the machine with others.

Otherwise, if you log out or your connection breaks, your job terminates. This command brings the output of your job to the console. But you can logout and comeback and it will still run.

1. use top -u loginame to see if your process is still running, and check nohup.out for errors

2. STAR makes log files, so look for them: ls *Log.*.out, then look at the file named *Log.progress.out. It should give you some progress information and mapping speed.

[Fleur]

Ah yes, sounds similar to tmux which is what I'm using. That also makes sure that the job keeps running when you close the terminal. Thanks!

And good point, I will have look at the STAR log files to see is I can figure out what's going wrong