Reproduce “ProteoCraft: Protein Design as Proposal-Verification Search” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Fang Wu, Hongbin Lin, Bingchen Duan, Yuchen Wang, Li Erran Li, Yanjie Huang, Mingxuan Cao, Yijun Ma and 19 more
ProteoCraft uses approximate gradients to propose sequence edits, then verifies candidates with complete structure predictions and refines promising sequences through coarse-to-fine search. In 210 de novo candidates, MCTS raised mean Pair-chain iPTM from 0.660 to 0.721; 114/210 met a strict predicted-confidence screen, versus 40/210 for Best-of-N. Experiments found binding among selected PD-L1 and IFNAR2 designs, not establishing broader specificity or generalization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.