A trimodal protein language model from Fajie Yuan's lab that aligns protein sequence, structure and function via contrastive learning (bioRxiv 2024; accepted Nature Biotechnology 2025). Its search server holds precomputed embeddings for 5B+ proteins, powering large-scale protein/text retrieval, and was expanded in 2025 with OMG/MGnify (3B sequences) and NCBI (700M) coverage — a companion retriever to the SaProt PLM.

scientific-fmproteinopen-weight