Remapping 1000 Genomes Data with BWA

[nexgengirl]

Hi,

I'm working now to remap some 1000 genomes exome data with BWA to compare with other exomes I've aligned with BWA. These were originally mapped with Mosaik. I've hit a bit of a snag and I'm checking to see if anyone has also experienced this or has some suggestions on how to fix. Basically, I took the bam file and converted it to the two fastq files (since it's paired end). I then aligned these using BWA aln to the reference (it worked great and fast). However, when I run the sampe step it hits the message below and after that the analysis slows drastically (sam file is only on chromosome 1 after running for two days on a 12 core machine with over 40 Gb of memory). Anyone else seen this?

[bwa_read_seq] 10.0% bases are trimmed.
[bwa_read_seq] 7.5% bases are trimmed.
[bwa_sai2sam_pe_core] convert to sequence coordinate...
[infer_isize] (25, 50, 75) percentile: (4084, 11785, 30818)
[infer_isize] low and high boundaries: 76 and 84286 for estimating avg and std
[infer_isize] inferred external isize from 207594 pairs: 19383.946 +/- 21062.583
[infer_isize] skewness: 1.370; kurtosis: 0.892; ap_prior: 1.00e-05
[infer_isize] inferred maximum insert size: 148287 (6.12 sigma)
[bwa_sai2sam_pe_core] time elapses: 3.36 sec
[bwa_sai2sam_pe_core] changing coordinates of 385 alignments.
[bwa_sai2sam_pe_core] align unmapped mate...
[aln_local_core] Potential bug: (155,166) > 65

[xied75]

Hi, nexgengirl,

1, You don't need to convert bam into fastq first, BWA can take bam directly, with -b -1, -b -2 you can let it aln directly from the bam.
2, The error is mismatched pair, you can see that those infered insert size all went mad.

Best,

dong

[swbarnes2]

You might not be able to use a single .bam file as a paired end file.

Try using Picard to sort your .bam file by name, then use samtools to pull out read1 .bam and read2 bams, and then you can align them separately and use sampe to put them together, as usual.

[xied75]

If the bam is made from paired end sam and sorted with name other than coordinate, then yes you can run BWA ALN on it directly, without all these extra steps.

bwa aln -b -1

will take the first read

bwa aln -b -2

will take the second read

Only if the bam is a mixture of single reads and paired reads then this might not work.

[nexgengirl]

Hi all,

Thanks for your suggestions. I reran the bwa sampe with the -A option to get it to run. It ran successfully in 16 minutes!

Thanks again.