Su, B.; Huang, K.; Peng, Z.; Amunts, A.; Yang, J.

Constructing atomic models from cryogenic electron microscopy (cryo-EM) density maps is essential for interpreting molecular mechanisms. In this study, we present CryFold, an approach for de novo model building in cryo-EM, leveraging recent advancements in AlphaFold2 to improve the state-of-the-art method ModelAngelo. To accommodate the cryo-EM map information, CryFold replaces the global attention mechanism in AlphaFold2 to local attention, which is further enhanced by a novel 3D rotary position embedding. It reduces the resolution requirement, accurately distinguishes between paralog sequences in noisy maps, detects previously uncharacterized proteins with unknown functions, precisely compartmentalizes the map for isolation of non-protein components, and better resolves conformational changes in difficult map regions. We showcase these advantages and overall higher model completeness and accuracy of CryFold with example maps in 3-4.5 Angstrom resolution of large multi-protein membrane macromolecular complexes. A particular case includes a 104-protein complex that has been modeled within only 5.5 hours, and a minor conformational change of a single protein domain has been detected at the periphery when models from two different maps were compared. CryFold stands as an accurate method currently available for model building of proteins in cryo-EM structure determination. It is available at https://github.com/SBQ-1999/CryFold.