top | item 41624499

(no title)

dunk010 | 1 year ago

It's useful to have as a resource that you can dip into; for instance to check a specific thing like "how fast do I metabolise caffeine (bad example as youc can get that from a SNV checking service like 23&Me. However, never ever EVER dump your whole genome into a program that will output every match for some potential condition. Now that's not for the reason you might immediately think ("Oh no look at all the bad stuff"), but rather because the databases that tools like this use are pretty chock full of junk where some researcher threw in a reference - along with a few thousand others - to say that their particular piece of research implicates this gene at this poistion with this variant. I've seen it with my own eyes, and a lot of it is very low quality. You therefore need highly curated databases where the genotype-phenotype associations are well-defined and thoroughly researched. You'll spend the rest of your life either trying to prove/disprove every potential false positive, or just worrying yourself needlessly. So, yeah, it's not a bad resource to have if you can be responsible with it.

discuss

order

bick_nyers|1 year ago

Hmm, now I'm interested in seeing if an LLM could be finetuned to give an effective quality score on such a study.