sokrypton predicts protein structures from LLM contact maps, admits the structure part is a gimmick
sokrypton · x · 2026-11-18
Protein-AI researcher sokrypton (author of ColabFold) notes this experiment unexpectedly blew up. He admits the structure part is a bit of a gimmick: once you have a good contact map, converting it into a structure is trivial. Ubiquitin happened to have one of the more accurate contact predictions, so he gave it a try. He also shares the full prompt, the artifact, and additional examples.
More from Research
- sokrypton shows models can reason about which beta sheets interact — sokrypton · 2026-11-18
- OpenAI study: persona-style CoT training drops monitorability from 0.94 to 0.22 — dair_ai · 2026-11-18
- LLM protein-folding test on fresh PDB entry lands near-random structures — MoAlQuraishi · 2026-11-18
- Lean-PEP turns worst-case optimization bounds into formally verified Lean proofs — prof_grimmer · 2026-11-18
- Two arms, two token streams, one shared vocabulary: new robot manipulation work open-sourced — DJiafei · 2026-11-18
- NUS's BiCAT action tokenizer for bimanual robots: 113/150 real-robot wins, 2.7x lower latency than OAT — DJiafei · 2026-11-18