Four generations of protein design models chart the road to fully programmable protein design
AllThingsApx · x · 2026-11-05
miangoar posted an 18-tweet thread tracing the evolution of four generative biomolecular design models — Proteína, La-Proteína, Proteína-Complexa, and RFD4-Proteina — arguing the field is heading toward fully programmable protein design.
Key arcs per the accompanying diagram: Proteína uses a non-equivariant transformer for large-scale Cα backbone generation; La-Proteína jointly generates sequence and full atoms via a VAE decoder with joint flow and separate clocks; Proteína-Complexa adds target-conditioned denoising with test-time search (pAE/H-bond rewards, prune-and-branch) for binder design; RFD4-Proteina uses an energy-supervised decoder with triangle updates to reason across amino acids, modified AAs, and small-molecule RNA via temporal search.
More from Research
- RoboColosseum launches: 5,699 pairwise real-world robot evals across 124 sites — DJiafei · 2026-11-05
- New Book on Generative AI and Stochastic Thermodynamics by Welling Team Out This Month — mjdramstead · 2026-11-05
- New paper uses 288M job postings to measure where AI raises the value of work it can't do — robseamans · 2026-11-05
- Matthew Honnibal: LLMs can 'see' weight matrices — but only one layer at a time — honnibal · 2026-11-05
- RicursiveAI claims 16x compute efficiency gain on code pretraining vs Olmo 3 baseline — samiramanabi · 2026-11-05
- Claude drifts from 🌀 to 💖: refined AI symbols become less legible to humans (67%→44%) — mimi10v3 · 2026-11-05