This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Bwa Sampe Segmentation Fault

Hi everyone,

I'm running bwa in the sampe mode and, after successfully processing >10M reads it fails with a segmentation fault (as follows) on what appears like a set of poorly-alignable reads. Any suggestions on what can be done to overcome this problem would be much appreciated. Many thanks!

#chunk processed ok
[bwa_read_seq] 2.8% bases are trimmed.
[bwa_sai2sam_pe_core] convert to sequence coordinate... 
[infer_isize] (25, 50, 75) percentile: (492, 529, 561)
[infer_isize] low and high boundaries: 354 and 699 for estimating avg and std
[infer_isize] inferred external isize from 55618 pairs: 523.901 +/- 55.549
[infer_isize] skewness: -0.798; kurtosis: 0.914; ap_prior: 2.38e-04
[infer_isize] inferred maximum insert size: 897 (6.71 sigma)
[bwa_sai2sam_pe_core] time elapses: 9.46 sec
[bwa_sai2sam_pe_core] changing coordinates of 9766 alignments.
[bwa_sai2sam_pe_core] align unmapped mate...
[bwa_paired_sw] 76039 out of 91346 Q17 singletons are mated.
[bwa_paired_sw] 5151 out of 16392 Q17 discordant pairs are fixed.
[bwa_sai2sam_pe_core] time elapses: 52.09 sec
[bwa_sai2sam_pe_core] refine gapped alignments... 4.60 sec
[bwa_sai2sam_pe_core] print alignments... 1.93 sec
[bwa_sai2sam_pe_core] 11010048 sequences have been processed.

# failed chunk
[bwa_read_seq] 3.1% bases are trimmed.
[bwa_sai2sam_pe_core] convert to sequence coordinate... 
[infer_isize] fail to infer insert size: too few good pairs
[bwa_sai2sam_pe_core] time elapses: 11.07 sec
[bwa_sai2sam_pe_core] changing coordinates of 0 alignments.
[bwa_sai2sam_pe_core] align unmapped mate...
[bwa_paired_sw] 0 out of 0 Q17 singletons are mated.
[bwa_paired_sw] 68090 out of 140740 Q17 discordant pairs are fixed.
[bwa_sai2sam_pe_core] time elapses: 63.65 sec
[bwa_sai2sam_pe_core] refine gapped alignments... 
/spool/1339511835.1833481: line 8:  9921 Segmentation fault      (core dumped) 
../../bwa-0.6.1/bwa sampe -a 2000 ../myGenome.fa myReads.PE500.4.1.sai myReads.PE500.4.2.sai myReads.PE500.4.1.fq myReads.PE500.4.2.fq > myReads.PE500.4.bwape.sam
bwa paired-end
• 12,446 views • link
updated by deepue deepue • written by a11msp a11msp

I had some seg fault problems with sampe awhile ago and after trying 101 different things, I eventually gave up and used bowtie2. So I'm curious to see what happens here.

That said, I would suggest mentioning what version of bwa you're using (or have tried) and hopefully post part of your failing data file somewhere if you want others to take a look.

• 0 views • link
written by Madelaine Gogol Madelaine Gogol

Haha. We both posted the same time with the same answer.

• 0 views • link
written by Zev.Kronenberg Zev.Kronenberg

I posted this question to the bwa mailing list and received the following patch from Mark Kelly that may fix this bug:

diff -crN bwa-0.6.1-orig/bwase.c bwa-0.6.1-fix/bwase.c
*** bwa-0.6.1-orig/bwase.c      Mon Nov 28 08:55:02 2011
--- bwa-0.6.1-fix/bwase.c       Tue Jun 12 13:05:17 2012
***************
*** 245 - 251 ****
                       }
               }
       } else { // no gaps
!               for (z = u = 0; z < (bwtint_t)len; ++z) {
                       c = pacseq[(x+z)>>2] >> ((~(x+z)&3)<<1) & 3;
                       if (c > 3 || seq[y+z] > 3 || c != seq[y+z]) {
                               ksprintf(str, "%d", u);
*** 245 - 251 ***
                       }
               }
       } else { // no gaps
!               for (z = u = 0; z < (bwtint_t)len && x+z < l_pac; ++z) {
                       c = pacseq[(x+z)>>2] >> ((~(x+z)&3)<<1) & 3;
                       if (c > 3 || seq[y+z] > 3 || c != seq[y+z]) {
                               ksprintf(str, "%d", u);

Am re-running with the patched version now and if it works, will post it as an update to the question and close it.

• 0 views • link
written by a11msp a11msp

Update: With this patch, the run lasted a little longer but eventually died in a very similar setting. Perhaps the patch is in the right place, but is not fully adequate...

• 0 views • link
written by a11msp a11msp

I had this exact problem and tried all the usual suspects. Sadly I couldn't figure it out and ended up using bowtie 2.

• 0 views • link
written by Zev.Kronenberg Zev.Kronenberg

Same here:

bwa sampe -s -A -r "@RG\tID:id\tLB:foo\tSM:id\tPL:ILLUMINA" hg19 1.sai 2.sai 1.fq 2.fq ... [bwasai2sampe_core] convert to sequence coordinate... Segmentation fault (core dumped)

This happens within 2 seconds of starting the program. For me, this result is reproducible:

  • Across bwa-0.5.9, bwa-0.6.1 and bwa-0.6.2
  • Across 50 different pairs of .sai files
  • With a variety of optional parameters: with or without -s and -A,
    different @RG string lengths, outputting to .sam file or piping to samtools.
• 0 views • link
written by urbio • urbio

I had this exact same seg fault 2 seconds after running. The sai file was produced on a cluster using bwa 0.7, and I was running sampe on another server with bwa 0.6. When I installed and ran bwa sampe using bwa 0.7, it worked perfectly. Might be worth trying the new version. -- Hope that helps people with this same issue.

• 0 views • link
updated by Ram Ram • written by rebekah.rogers • rebekah.rogers

5 answers

This exact error can also occur when wrong reference files are used (so when the chromosome id's in the reference index do not match those in the .sai)

• 2 views • link
written by evelyne.c88 • evelyne.c88

I was having the same problem and after I rebuilt the BWA indexes everything worked fine

• 0 views • link
written by perry.ridge • perry.ridge

Multiple users reported problems with bwa in similar circumstances. See comments to the question.

The main purpose of this answer is to label this question answered so that more people look at it and the excellent comments to the question

• 1 views • link
written by Istvan Albert Istvan Albert

We were getting a segmentation fault with sampe in version 0.5.9 because n_occ and rest were overflowing in bwase.c and becoming negative, so I changed from line 46:

if (n_multi) {
    int k, rest, n_occ, z = 0;
    for (k = n_occ = 0; k < n_aln; ++k) {
        const bwt_aln1_t *q = aln + k;
        n_occ += q->l - q->k + 1;
    }

to:

if (n_multi) {
    unsigned int k, rest, n_occ, z = 0;
    for (k = n_occ = 0; k < n_aln; ++k) {
        const bwt_aln1_t *q = aln + k;
        n_occ += q->l - q->k + 1;
    }

as q->l and q-> are defined as uint_32 if you look in the appropriate header. It worked after the above change. I don't know if this is related to your problem.

• 0 views • link
written by nick.shrine • nick.shrine

I had similar problem with bwa version 0.7.4. but then I switched to bwa ver 0.6.2 and it worked fine.

• 0 views • link
written by mghandi • mghandi

I had a sample run paired-end sequenced in two lanes, sample data of first lane has run successfully but the second lane data is having "Segmentation fault" error. If this is due to version issue or memory, the first one should have failed. What might be the issue ? Could someone please help me.

Thanks

• 0 views • link
written by deepue deepue

Log in to answer this question.

More posts like this