Automated assembly of protein complexes from cryo-EM maps with structure-informed Monte Carlo Tree Search
Dilip, R.; Qu, S. J.; Chen, Z.; Van Valen, D. A.
Show abstract
Structural cell biology aims to visualize and identify functional molecules in their native environment. Macromolecular complexes have thus far been resolved predominantly at intermediate resolutions. This poses a major challenge for modeling due to the vast combinatorial space of possible components within a proteome. Here, we developed Cryosearch, a system for automated modeling of macromolecular complexes from proteome-scale monomer libraries. We implemented Monte Carlo tree search with correlation-based rewards to identify combinations of protein domains that collectively best explain a density map. This approach enabled autonomous de novo assembly of molecular complexes from intermediate-resolution maps, a task that has been difficult to perform manually.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.