Reproduce “Calpha: Toward Native Biomolecular Intelligence in a General-Purpose Language Model” 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

Yong Liu, Zhanpeng Shi, Peicong Lin, Zhenyi Zhong, Yizhou Dang, Wendi Wu, Haoling Li, Shuangjia Zheng

Calpha connects a chat model’s residue states to a structure generator, letting it choose text or structure output. ESM2-3B features and a conditioning front-end pass Qwen’s states to a SimpleFold-3B head. On CASP16 domains, mean Cα TM-score is 0.635. On 300–600-residue chains, Calpha scores lower than a direct ESM2 baseline; differing training recipes prevent attributing the gap to Qwen alone.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.