Can I use a gpu to demultiplex with bcl2fastq?

Question

[diltsjeri]

Hi,

Just brainstorming ways to speed up our pipeline. Demultiplexing is definitely a time hog. I'm not too familiar with exactly how demultiplexing works computationally. Is bcl2fastq written in such a way that it could utilize our gpus? I'm not finding much documentation on the subject.

[GenoMax]

Demultiplexing is definitely a time hog.

You will have to quantify along with information about what type of run data (sequencer, cycles etc) you are processing and on what compute infrastructure.

Otherwise I don't think it is particularly so. Latest version of bcl2fastq v.2.19 is much much faster compared to previous versions and if you have access to a cluster/high performance storage system you can process runs at a good clip using multiple threads.

[diltsjeri]

Yeah the documentation says cpus only with threading options.

[eperry2688]

Can you contact me via email?

Answer 1

[Devon Ryan]

No, it can't use your GPUs. Just use multithreading, it runs reasonably quickly then given how long it takes to sequence things.