#339 Bowtie2 segfault

[Geoffrey Wong]

I ran this command:
bowtie2 --very-sensitive -k 2 -x hg19 -U .trimmed.fastq -S .trimmed.fastq.sam

Segmentation fault Jul 27 2015 1:41 PM
(ERR): bowtie2-align exited with value 139

I've tried this on both Ubuntu 12.04 and Mac OS X Yosemite(where I get (ERR): bowtie2-align died with signal 11 (SEGV) ).

Any help with this? Thanks!

[Val]

Hi Geoffrey,

Can you please try release 2.2.6 and let me know how it goes? https://github.com/BenLangmead/bowtie2/tree/master/latest_release

thanks,
Val

[Geoffrey Wong]

I am still getting the same error.

Segmentation faultJul 28 2015 11:05 AM
(ERR): bowtie2-align exited with value 139

[Val]

Is it possible to get your datafile (.trimmed.fastq) ?

thanks,
Val

[Geoffrey Wong]

Here is the file:

https://dl.dnanex.us/F/D/Yf7Yy72FPyZZ1p87Byf1KQQ1vx53G6V028j3gPQY/vegf3_sense.assembled.trimmed.fastq

[Val]

Hi Geoffrey,

Thank you for the file. I tried running it on my laptop. I tried 2 versions of the command you were using:
./bowtie2 --very-sensitive -k 2 -x ~/data/hg19 -U vegf3_sense.assembled.trimmed.fastq

and

./bowtie2 --very-sensitive -k 2 -x ~/data/hg19 -U vegf3_sense.assembled.trimmed.fastq -p 3

They both finished well:
1145771 reads; of these:
1145771 (100.00%) were unpaired; of these:
97161 (8.48%) aligned 0 times
396673 (34.62%) aligned exactly 1 time
651937 (56.90%) aligned >1 times
91.52% overall alignment rate

I will try to run it tomorrow on a server as well. Did you try to run on another machine? Just to eliminate the possibility that there is some hardware malfunctioning somewhere.

thanks,
Val

[Geoffrey Wong]

GWong-MBA:hg19.bt2-index geoffrey.wong$ bowtie2 --very-sensitive -k 2 -x hg19 -U ~/Downloads/vegf3_sense.assembled.trimmed.fastq -S file.trimmed.fastq.sam
(ERR): bowtie2-align died with signal 11 (SEGV)

I ran it locally on my mac and the seg fault still popped up.

[Val]

Hi Geoffrey,

I managed today to run it on some servers with Scientific Linux 5.8 (Boron). This is the error I got:
Error: Read M02528:17:000000000-AFTVK:1:2106:23570:21123 1:N:0:0 has more read characters than quality values.
It seems either the trimming was not done properly or something made that sequence to lose most of its quality values.

Val

[Geoffrey Wong]

Hi Val,

I looked for the line, which is here:

@M02528:17:000000000-AFTVK:1:2106:23570:21123 1:N:0:0
ATATTTTTTTAATTAAAAAATCTACGGCTCACTAGTGGTTCAAGAAAATAATTCCATAGGACACAGCTACTATTTTAACCTATTAACTAGAATAAAAACTATGGTGCCAGGCGCGGTGGCTCACGCCTGTAATCCCAGCACTTTGGGAGGCCGAGGCGGAC
+
CCCC8EFEGG<9FGFDA<CFFGFGGG7FBFC,IIIICIIIIIIIIIIIIIIIIIIIIIIGIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIHIIIIIIIIIIGIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII@,F@FB7EGF;

I don't recognize any issue with this line, and it seems that there are the same number of read chars and quality vals are there? I'm unsure of how to proceed.

[harijay]

Hi Val,
I work with Geoff on similar datasets and have seen the segfault 11 on Ubuntu Linux 14.04 and 12.04 with 16 GB and 98 GB memory with reads that came off the Miseq demultiplex software ( hence hopefully free of the trimming induced errors) . I can reproducibly get the error with 2.2.5 ( have not tried 2.2.6) with Ubuntu. If I run the same dataset on SUSE linux , it always works . Ubuntu mostly gives the SIG 11 error or in some cases the "(ERR): bowtie2-align exited with value 139" error.
I can send you these data as well if it helps troubleshooting.

Also posted on seqanswers with a little more details: http://seqanswers.com/forums/showthread.php?t=61545

[Val]

Hi Geoff and Harijay,

You are right, the sequence is ok. For some reason my file did not copy correctly over to the servers I am using. I moved it again and, of course, bowtie finished with no problems this time too. I need to run it on an OSX and/or Ubuntu. I might have to install an ubuntu on a virtual machine.
Thank you both for helping me figure this problem out.

I will be back after I try to run it on these 2 OS as well.

thanks again,
Val

[Val]

Hi guys,

I installed a Ubuntu 14 under VirtualBox and I also got a time share on a OSX Yosemite. I ran the test as described previously and they both succeded like before. I plan to try next an Ubuntu 12 as well. In the meantime, Harijay, would it be possible to get some part of your dataset that fails like this as well?

thanks,
Val

[harijay]

Hi Val,
I am getting the Ubuntu segfaults during a run using a command that aligns it against an amplicon index.
The index is built using command:

bowtie2-build amplicons_dk.fasta amplicons &> index.err

The alignment of the Illumina Miseq reads to this amplicon index is done using command

bowtie2 --very-sensitive -k 2 -x amplicons -1 [fastafile] -2 [fastafile] | samtools view -bS - > initial_aligns.bam' % (bowtie_options, ','.join(read1_fastqs), ','.join(read2_fastqs))

The amplicons fasta file is at :
https://s3-us-west-2.amazonaws.com/miseqdata/141218dm/amplicons_dk.fa

Read 1 fastq file is at :
https://s3-us-west-2.amazonaws.com/miseqdata/141218dm/VEGF1_S1_L001_R1_001.fastq.gz

Read 2 fastq file is at :
https://s3-us-west-2.amazonaws.com/miseqdata/141218dm/VEGF1_S1_L001_R2_001.fastq.gz

I can reproducibly get a crash while processing whole 10 sample dataset with sig11 in Ubuntu with 2.2.5 , but never on SUSE. I have attached the data for Sample 1 of 10 above.

Thanks
Hari

[harijay]

Hi Val...
Thank you for your help with troubleshooting this. The segfaults went away when we switched the way I was calling bowtie2. Something about the way python subprocess with shell=True interacts with bowtie2 was creating the segfaults only on Ubuntu but not on SUSE. I have switched the way I call bowtie2 to subprocess.call ["bowtie2",...] and that has resolved the OS depended crashes.