Automated assembly of protein complexes from cryo-EM maps with structure-informed Monte Carlo Tree Search

Area of Science:

• Structural cell biology
• Biophysics
• Computational biology

Background:

• Visualizing functional molecules in their native cellular context is crucial for structural cell biology.
• Intermediate-resolution maps of macromolecular complexes in situ complicate protein identification and structural modeling due to proteome complexity.

Purpose of the Study:

• To develop an automated framework for assembling macromolecular complexes from proteome-scale monomer libraries.
• To enable de novo assembly of molecular complexes from intermediate-resolution maps.

Main Methods:

• Developed Cryosearch, a GPU-accelerated framework.
• Implemented Monte Carlo tree search with correlation-based rewards.
• Identified combinations of protein domains that best explain density maps.

Main Results:

• Cryosearch automates the assembly of macromolecular complexes.
• The framework successfully identifies protein domain combinations from density maps.
• Enables autonomous de novo assembly of molecular complexes.

Conclusions:

• Cryosearch advances structural cell biology by facilitating molecular complex assembly.
• The developed framework addresses challenges in protein identification and structural modeling.
• This approach allows for autonomous de novo structural determination of cellular machinery.