Does it help to have them "check each other's work"? Eg, run it with two different tools, if you get the same result there's a higher chance it's right? Or there's something about how the PDF is typeset that makes some sequences particularly hard to extract correctly?

If you're willing to try nextflow instead of snakemake I put together a generic module for split-apply-combine that can be used to easily parallelize single-threaded FASTQ processing programs. You just need to write a little wrapper process, and specify fan-out number github.com/shenkers/nf-...

Bouillon Bilk, it was very cool!

Impressive speedup! It's the same fundamental data-structures, it's just faster because it's written in Rust? Haven't used rust for anything yet, but I would have expected performance to be pretty similar to bedtool's cpp.

Why do you think that is? Are we measuring the wrong thing? We are measuring the right thing, but we don't know how to interpret it? Something else?

Is that an individual knockout performed in adult mice? Is it a failure to develop testes vs failure of germ stem cells to differentiate? If you subsequently turn it back on is fertility restored?

@chrisamiller.bsky.social @lwpembleton.bsky.social I'm trying to bootstrap a platform exactly along these lines, hoping to help small strappy companies or research labs. Would love show and hear more about your pain points if you have time.