Biggest project I’ve ever undertaken, a modern fast I/O bioinformatics library, written in rust. Identical commands to samtools and BCFtools, but way faster and unified.
https://t.co/i1GVLlnPUP
Affirmative, you could either spend a few months handwriting the most optimized SIMD for single threaded performance, or make it work efficiently in parallel. You should consider porting Zflate to goose, that one was all handwritten cpu instruction code, and is the codec that pasta relies on.
@ducktapeprogra1@jpelbers Suprised by the goose results, very interesting. I will say that pasta was designed for parallel work, HTSlib parses, decodes, formats and encodes on the same thread, pasta splits the work and often delivers more then 10x when parallel.
@ducktapeprogra1@jpelbers It started as a fork of noodles, when I realized how subpar noodles was, making no real contribution apart from async rust rewrite, so I somewhat developed it from the ground.
@ducktapeprogra1@jpelbers Hi, seqair and noodles use libdeflate, pasta has a rust only codec that I developed alongside pasta, it’s the fastest I/O library (and codec) that I know of today, give it a shot
@beboelhosary@nicbstme What you can do if you’re concerned about heritage, is run a pangenome alignment. A pangenome reference takes references from all over the world (instead of European ones like hg38). That should give you a better and less noisy result.
@beboelhosary@nicbstme More or less, they align your genome to a reference, find the differences, and show you them in a VCF (variant call format) which you can then annotate to find which variants are pathogenic. If you want more info I’ll be happy to show you