bwa sampe segmentation fault

[papori]

Hi all,
i am using bwa-0.5.9 for illumina output.

i have reads from 2 samples of same organism.
So i have 2 databases.

when i run BWA on the first, everything went well. the SAM file is 3.3GB.

On the second, i got segmentation fault in the last part, when converting to *.sam.
it succeed to convert 2.2GB , and then seg` fault.

My steps are:
bwa index -a is database.fasta

bwa aln database.fasta read_1.fastq > database_aln_sa_1.sai
bwa aln database.fasta read_2.fastq > database_aln_sa_2.sai

bwa sampe database.fasta database_aln_sa_1.sai database_aln_sa_2.sai read_1.fastq read_2.fastq > database_aln.sam

When i run bwa with samse on each end separately , i got 2 sam file , each in size 1.6GB , as expected.


My questions are:
How can i solve the Segmentation fault?
Or, Can i merge the 2 sam files that have been resulted separately to one sam?without loosing any data?

Thanks in advance..

[giorgifm]

Dear papori,

my best solution at BWA segmentation fault has always been to change cpu and increase RAM. Otherwise, you may try splitting your original paired fastq files in half and try again.

And then merge the resulting (paired) BAM files with Picard Tools MergeSamFiles

Good luck!

[CompBio]

I'm running into a similar problem. I've got about 1 million paired-end reads (~500,000 pairs) that I'm aligning to about 6,000 transcript sequences. bwa aln works fine for both of them, but bwa sampe exits immediately with a segmentation fault, both with and without the -P option.

I'm using the following format:

bwa sampe -f part_1_part_2.sam transcripts.fa part_1.aln part_2.aln part_1.fq part_2.fq

The only feedback I get is:

Segmentation fault (core dumped)

I'm working on a machine with 128GB memory, so I really can't imagine it's a memory issue. I've used bwa sampe successfully with similar data (500,000 pairs and 28,000 transcripts), but not with this particular data set.

It's frustrating, to say the least.

[stylz2k]

I'm having the same trouble with bwa 0.6.2. Its really frustrating.

I have a machine with Intel Core i5 and 8 GB RAM.

[fongchun]

Did you guys ever find solutions to your problems of segmentation fault?

[CompBio]

In our case the problem went away when we reduced the size of the headers in the FASTA file. We were storing meta-data in the headers that sometimes grew pretty long (100's of characters). By storing the meta-data in a table and using only the indexes as FASTA headers, the problem went away.