search

andersen-lab / Freyja

Depth-weighted De-Mixing
BSD 2-Clause "Simplified" License
102 stars 29 forks source link

Reduce runtime #184

Closed erinyoung closed 6 months ago

erinyoung commented 11 months ago

I like using Freyja in my SARS-CoV-2 nextflow workflows. I am encountering issues, however, with how long Freyja takes to run.

Do you have tips or suggestions on how to reduce the time that Freyja variants demix or boot takes to run?

joshuailevy commented 11 months ago

Hi @erinyoung!

Glad you've been enjoying using Freyja! The demix step has definitely been taking a bit longer as of late now that the list of candidate lineages has ballooned. I'd worked on a few tricks to speed that up a while back (e.g., on dropping lineages with few signature mutations present before solving the demixing problem), but they didn't quite make it into the codebase. Of course using a barcode with only recent lineages will speed things up, but in general I wouldn't recommend that as you risk missing things like cryptic circulation.

The variants step generally takes more time overall- and I haven't really tried speeding that one up- I'll take a hack at that next week. Just to check- is all of your sequencing just SARS-CoV-2 amplicon seq? Or do you have multiple schemes/references?

Best, Josh

erinyoung commented 11 months ago

Just one scheme.