DeepTracer

DeepTracer reconstructs de novo multi-chain protein complex structures from high-resolution cryo-electron microscopy (cryo-EM) density maps using deep learning.

Key Features:

• Deep learning-based methodology: Employs deep learning algorithms to interpret cryo-EM density maps for rapid and precise de novo protein structure determination.
• Fully automatic processing: Operates fully automatically for de novo multi-chain protein complex structure determination from cryo-EM maps.
• Performance improvements: Demonstrated over 30% increase in residue coverage and an RMSD improvement from 1.29 Å to 1.18 Å compared with state-of-the-art methods.
• Application on coronavirus-related complexes: Validated on 62 coronavirus-related density maps, including 10 with no previously deposited structures, achieving an average residue match of 84% and RMSD of 0.93 Å against known structures.
• Computational efficiency: Capable of processing approximately 60,000 residues across 350 chains in roughly two hours.

Scientific Applications:

• Viral complex structural analysis: Determines macromolecular architecture of SARS-CoV-2 and other coronavirus protein complexes from cryo-EM density maps.
• Therapeutic and vaccine support: Provides structural models to support identification of potential therapeutic targets and to accelerate vaccine and drug development.

Methodology:

Uses a fully automatic deep learning-based approach to interpret cryo-EM density maps for de novo multi-chain protein complex structure determination.