cpu usage in pindel for samples aligned with bwa mem

Question

[Xavier Pastor]

I'm using pindel on a set of sample pairs aligned using bwa mem but it doesn't make a good usage of the specified threads. In addition it seems to take too long to process a single chromosome. When I run pindel on the same samples but aligned using bwa aln the profile of multithreaded usage is completely different and much more efficient.

Is it possible that pindel doesn't make a good usage of the new flags in the bam reported by bwa mem, since now it's possible to find multiple primary alignments for a split read and the new supplemental flag 0x800?

A typical call to pindel is:

pindel-0.2.5a3 -f /path/to/hg19.fa -i /path/to/config.txt -o my_sample -c Y -T 8

Also mention that the argument -M was not set in the alignment of the samples with bwa mem.

Best,
Xavier

[Xavier Pastor]

I'd just like to mention that after removing the supplemental reads in the bwa mem alignment (samtools view -bh -F 0x800) pindel seems to behave similar as to bwa aln alignments, at least computationally. I still need to confirm how about the final results.

If this is true I guess my initial suspicion was true and pindel cannot deal well with bam files reporting several primary alignments for the same read.

Best,
Xavier