bwa mem segfault; bwa bwasw breaks MarkDuplicates

[ElenaN]

Hi,
I have run into various problems with various versions of bwa
1. bwa mem gives a segmentation fault right away. Not sure how to debug that, no matter what memory (we have 48G) and thread settings. Has anyone else run into this problem?
The call to bwa is
bwa-0.7.5a/bwa mem -M -R "@RG\tID:$ID\tLIB:SeqCapEZ.v3\tID:$ID\tPL:ILLUMINA" -t 24 ucsc.hg19.fasta $READS1 $READS2 > $ALN_OUT.sam

2. bwasw produces a file which does no work with MarkDuplicates.
The call to bwa is
/old_home/enabieva/bwa-0.7.5a/bwa mem -M -R "@RG\tID:$ID\tLIB:SeqCapEZ.v3\tID:$ID\tPL:ILLUMINA" -t 24 ucsc.hg19.fasta $READS1 $READS2 > $ALN_OUT.sam

The MarkDuplicates error is:
Exception in thread "main" net.sf.picard.PicardException: Value was put into PairInfoMap more than once. 1: null:HWI-ST992:135:C0YGDACXX:2:1108:12651:98485

Grepping for that read brings up the following:
HWI-ST992:135:C0YGDACXX:2:1108:12651:98485 113 chr1 564775
0 43S47M chr2 55113029 0TATCGTC
ACAGCCCATGCATTTGTAATAATCTTCTTCATAGTAATACCCATTACAATCTCCAGCATTCCCCCTCAAACCTAAGAAATATG
@DDDDEDBDDDDCCDFFFFFEGHHHGHIIGII
HJIIGGFGAGEGGBJIIGHFBJJGEHHJGGGCIIIJJJIGGIIHHHFFHHFDFDD@@@
AS:i:47 XS:i:47 XF:i:1 XE:i:1 NM:i:0
HWI-ST992:135:C0YGDACXX:2:1108:12651:98485 113 chr1 566621
0 51M39S chr2 55113029 0TATCGTC
ACAGCCCATGCATTTGTAATAATCTTCTTCATAGTAATACCCATTACAATCTCCAGCATTCCCCCTCAAACCTAAGAAATATG
@DDDDEDBDDDDCCDFFFFFEGHHHGHIIGII
HJIIGGFGAGEGGBJIIGHFBJJGEHHJGGGCIIIJJJIGGIIHHHFFHHFDFDD@@@
AS:i:51 XS:i:51 XF:i:1 XE:i:1 NM:i:0
HWI-ST992:135:C0YGDACXX:2:1108:12651:98485 145 chr2
55113029 168 90M * 0 0TGAGCCA
CCACGTCCAGCCCACTTTTTGATTTTCAAAGAGTCAGAGCCCCAAAATTCCATCTCTTCCTTGCCTCTCATCGTAGGGTTCAC
A:<?@<,8<=DBA@@?==3;;BBFEE>HEEEH
CAHG@HFFEEEE<GIGIIGHGJJIHF>BIIEIIHGJIIIGGCFD?HDGGGFDDFF@@@
AS:i:90 XS:i:0 XF:i:3 XE:i:2 NM:i:0

So it seems that the problem is that the first read is mapped twice, but this is not marked in any way.

I found a discussion of this problem, but the suggested solution seems roundabout.
Does anyone else encounter this?

Thanks,
Elena

[Zengquan]

Yes, I met exactly the same error with you. Although I used the -M option, it seems also not compatible well with picard. I'm looking forward to an effective solutions.

[sdriscoll]

As far as I know the picard people are working with the bwa people to get the two programs to be more compatible. BWA MEM 0.7.4 should still work. The -M option is going to be gone and instead they are going to use a new flag which will have to be picked up by samtools and picard - but like I said they are all working together on this. I can't explain the seg-fault but I do know that the latest version of MEM has done away with what the -M option used to do. If you back up to a previous version of MEM it should still work. I'm using '0.7.4-r389-beta' and it mostly works. The bwa aln/sampe pipeline is broken in that it fails to properly annotate the 'XA' field of the alignments which triggers a bug in samtools if you attempt to run those alignments through it. However, that version's 'MEM' works perfectly for me (on Mac and Linux).