CGBack: Diffusion Model for Backmapping Large-Scale and Complex Coarse-Grained Molecular Systems.

Area of Science:

• Computational Biology
• Molecular Modeling
• Biophysics

Background:

• Coarse-grained (CG) models accelerate molecular dynamics (MD) simulations by reducing computational cost.
• Reconstructing atomistic detail from CG models (backmapping) is crucial for structural analysis but remains challenging.
• Existing backmapping methods struggle with stereochemistry, high-energy states, and complex biomolecular systems.

Purpose of the Study:

• To develop an advanced backmapping framework for reconstructing all-atom molecular structures from CG representations.
• To address limitations of conventional backmapping pipelines in preserving molecular fidelity and efficiency.
• To enable accurate atomic detail recovery for diverse and large-scale biomolecular systems.

Main Methods:

• Developed CGBack, a backmapping framework utilizing a denoising diffusion probabilistic model.
• Implemented backmapping and refinement procedures within the CGBack framework.
• Validated CGBack across various protein systems, including single-chain, multichain, and intrinsically disordered proteins.

Main Results:

• CGBack accurately recovers atomic coordinates from CG representations across diverse protein scales.
• The framework successfully backmaps complex systems, including densely packed intrinsically disordered proteins.
• Demonstrated preservation of stereochemistry and avoidance of high-energy configurations.

Conclusions:

• CGBack offers a robust and accurate solution for backmapping CG molecular models to all-atom representations.
• The framework enhances multiscale molecular simulation pipelines, improving protein modeling efficiency.
• CGBack is a promising tool for advancing simulations of proteins and other biomolecules across various CG models.