Looking good

Yea but you should be able to do equal or better than traditional methods, if you feed llms more data, they do get better

This might get a lot better with >100x data but imagine UKB is too small.

One (unfortunately shitty) model to rule them all 😬

Support Mamdani dipshit

Support Mandani

What I meant is that eQTls may happen randomly (RNA noise). The nearest gene is a good benchmark for causal GWAS genes, so a method that points to causal genes would ideally be better than that, which you can measure by for example by testing pathway enrichment, rare gene burden enrichment etc.

So you’re not just finding more, but the overlap Is significantly more than baseline ? And are the genes also more biologically enriched, (accounting for nearest gene)?

Incredible. One speaker is here WHILE BATTLING CANCER because she’s one of the thousands of people enrolled in a clinical trial that was senselessly cut by Trump/Musk