Parallelization for Bismark using Bowtie 1

[KlChen]

Thank you very much!

Nothing is wrong with Bismark. When you run bowtie1 using multiple threads, you can't guarantee what order the reads will be returned in. Since Bismark runs multiple instances of bowtie at once and then processes their outputs, it needs the reads to be in the same order (otherwise, you end up comparing different reads and everything breaks down). In bowtie2, this can be enforced with --reorder, but no such option exists in bowtie1. Just use bowtie2.

[dpryan]

Nothing is wrong with Bismark. When you run bowtie1 using multiple threads, you can't guarantee what order the reads will be returned in. Since Bismark runs multiple instances of bowtie at once and then processes their outputs, it needs the reads to be in the same order (otherwise, you end up comparing different reads and everything breaks down). In bowtie2, this can be enforced with --reorder, but no such option exists in bowtie1. Just use bowtie2.

[KlChen]

For Bowtie 1, I can use the -p option to do parallel search, but when I run Bismark with Bowtie 1 and the option -p 7, output is

"The parallelization switch '-p' only works for Bowtie 2. Please respecify! at bismark/bismark_v0.9.0/bismark line 6497

What's wrong with -p and Bismark? How can I speed up my alignment with Bowtie 1? Shall I use Bowtie 2 instead?