STAR "freezes" while processing multiple files

[Alexandra Kitz]

Hi! 

This problem was discussed before but none of the solutions worked for me...

I have multiple .fq-files that I'd like to align, I did plate-based scRNAseq on a Nextseq500 so I have 8 files per cell (4 lanes, R1 and R2 each). I wrote a loop over the files in a shell script for our cluster:

#!/bin/bash

#SBATCH --partition=general

#SBATCH --job-name=align11f

#SBATCH --ntasks=1

#SBATCH --nodes=1

#SBATCH --cpus-per-task=8

#SBATCH --mem=100000

#SBATCH --time=96:00:00

#SBATCH --mail-type=ALL

#SBATCH --mail-user=email

for i in $(ls 11Fat/*_R1_001_trimmed.fq.gz | sed s/_R1_001_trimmed.fq.gz// | sort -u); do STAR --runMode alignReads \

--genomeLoad  NoSharedMemory \

--genomeDir indices/STAR \

--readFilesIn ${i}_R1_001_trimmed.fq.gz,${i}_R2_001_trimmed.fq.gz \

--runThreadN 8 --outFileNamePrefix results/STAR/$i.fq.gz \

--outSAMtype BAM SortedByCoordinate \

--readFilesCommand zcat ; done

It works fine for about the first 12-14 files - also fairly quickly, which is awesome! - but then it seems to "freeze" but doesn't produce any (for me visible) error. Attached is the log.out file for the last generated file in my loop. If I just let it go on, the cluster will time out eventually.

Where's my mistake?

Thanks,

Alex

[Alexander Dobin]

Hi Alexandra,

it looks like there was some kind of trouble with genome generation step, the --genomeChrBinNbits parameters appear to be set to 0.

Since you are mapping to the transcriptome, the recommended value for this parameter would be log2(AverageTranscriptLengh)~12.

Please check your command line, and if it looks OK, send me the Log.out file from the genome generation step.

Also, if you are mapping reads as paired-end, you would need to separate R1 and R2 files with space, not the comma.

Cheers

Alex

[Alexandra Kitz]

Update: The problem disappeared after doing the following: 

1. Re-generating the genome with --genomeChrBinNbits 12

2. Using dSQ to run the script on the cluster.

Thanks!

On Wednesday, March 28, 2018 at 6:15:21 PM UTC-4, Alexandra Kitz wrote: