Problems with bwa on Q2 trimmed paired end reads

[curious_mapper]

Hi

I used bwa before on both 75 mers and 100mers PE reads successfully. However, I have been noticing some issues when I try to run reads that are trimmed off of low quality (residues whose quality correspond to B). The input reads are still paired ends, but the only difference is that they are not of the constant length anymore. Each mate in a pair could be of different sizes (upto 75 mer) based on its quality. BWA ran successfully several times on the same data set before without the quality trimming.

This brings me to the question if bwa can handle inconsistent read lengths during alignment. I also tried to run a small subset of sample reads to check if it finishes successfully, but it fails every time.

The only parameter I am using is n=7. For the larger set it exits saying "User defined signal 2" after trying to process a few million reads while creating the .sai files. I am also pasting the console log for a small subset of trimmed paired end reads I have been trying to run. In either case, it has not even started to write the .sam files.

Any help would be highly appreciated. Thanks!


[bwa_aln_core] calculate SA coordinate... 2016.01 sec
[bwa_aln_core] write to the disk... 0.03 sec
[bwa_aln_core] 100000 sequences have been processed.
[bwa_aln_core] calculate SA coordinate... 3217.48 sec
[bwa_aln_core] write to the disk... 0.02 sec
[bwa_aln_core] 100000 sequences have been processed.
[bwa_sai2sam_pe_core] convert to sequence coordinate...
[infer_isize] (25, 50, 75) percentile: (58334509, 565854931, 1207105846)
[infer_isize] low and high boundaries: 75 and -2147483648 for estimating avg and std
[infer_isize] inferred external isize from 60 pairs: 632688892.850 +/- 606565808.842
[infer_isize] skewness: 0.710; kurtosis: -0.690
[infer_isize] inferred maximum insert size: -1108636348 (4.21 sigma)
[bwa_sai2sam_pe_core] time elapses: 31.80 sec
[bwa_sai2sam_pe_core] changing coordinates of 22 alignments.
[bwa_sai2sam_pe_core] align unmapped mate...

[maubp]

For those interested in the Q2 trimming for recent Illumina data, see this thread:

http://seqanswers.com/forums/showthread.php?t=4721

[jmartin]

I'm wondering about this too. If we rely on external means to trim pair end illumina data, and that results in the pair end mates being trimmed to different lengths in some cases, would that cause bwa to crash?

I know about the -q option built into bwa. But in some cases we will have pair end files that were externally trimmed to accomodate other analysis that was done. Can bwa sampe handle that?