Session: COMP Poster Session:

Molecular dynamics (MD), a powerful computational approach for studying the physical movements of and interactions between macromolecules especially when laboratory studies of such molecules are difficult, is gaining momentum with recent studies into the SARS-CoV-2 virus. All-atomic, fine-grained models are known to provide accurate results; however, simulating these models is computationally expensive, as all pairwise electrostatic and Van der Waals interactions must be considered. Coarse-grained (CG) models are more efficient alternatives to all-atom models, which accelerate simulations with an affordable sacrifice of accuracy by using a small number of particles to represent groups of particles and reducing unnecessary computations.

This research applies two methods of developing CG models for novel macromolecules with generic molecule topologies, demonstrated by using the SARS-CoV-2 spike glycoprotein (S-protein, PDB code: 6VXX). A simulation mapping method that combines different timestep integrations to accelerate CG model simulation while combating molecule instability resulting from poor numerical integration is developed. After minimizing the CG system potentials of the S-protein, GPU-accelerated simulations are conducted with our simulation mapping method. For validation, the preliminary root mean square deviation (RMSD) values are compared to the all-atomic simulations as the control under NPT conditions. The converged RMSD trends confirm that our CG models have stable structures.

For ongoing efforts, minimizing the differences of RMSD values by AI enables us to obtain more accurate parameters for longer simulations. Our CG model uses learned parameters from big data to accelerate the conventional all-atomic MD without significant loss in accuracy. As the first AI-guided multiscale CG model, it may help enable a thorough investigation of the binding dynamics of the S-protein at varied ambient conditions including temperature and pH values. Additionally, our AI-guided modeling methodology and novel simulation mapping method can be generalized to conduct stable CG simulations for other molecules and materials too.

Figure: The all-atomic model of the spike glycoprotein, our SBCG (Shape-Based Coarse Grained) model, and our RBCG (Residue-Based Coarse Grained) model, respectively.

Figure: The all-atomic model of the spike glycoprotein, our SBCG (Shape-Based Coarse Grained) model, and our RBCG (Residue-Based Coarse Grained) model, respectively.

Yuefan Deng

Appearances