search

pjedge / longshot

diploid SNV caller for error-prone reads
MIT License
182 stars 26 forks source link

Memory usage on PromethION genomes #92

Open wdecoster opened 1 year ago

wdecoster commented 1 year ago

Hi,

I notice that longshot is using quite a bit of memory on our PromethION (human) genomes, at the moment 140Gbyte RAM. I don't know in which step this happens, but this of course limits the number of samples I can progress in parallel on our server... I assume the suggestion is to use --region and process chromosome by chromosome, but is there a reason this cannot be implemented internally in longshot to process chromosomes separately, freeing up memory after a chromosome is done?

Let me know if this is something you could use some help with!

Wouter

vibansal commented 1 year ago

It can be implemented but would require someone with expertise in Rust programming. If you are interested, please email me.

Also, a python wrapper that can process chromosomes in parallel exists: https://github.com/vibansal/longshot2