Alpha carbon-only representations are sufficient for ML force fields

Summary

For ML-based MD potentials, alpha carbon-only representations are sufficient to recapitulate the dynamics of small proteins (1,2). Models that use both alpha and beta carbons were found to be less stable than those using only alpha carbons (2). These potentials were calculated using graph neural networks with TorchMD.

Figures

Ref (2)

See also

• Information about backbone atoms is lost when rebuilding full-atom protein models from alpha carbon trace alone

1.

Majewski M, Pérez A, Thölke P, Doerr S, Charron NE, Giorgino T, et al. Machine learning coarse-grained potentials of protein thermodynamics. Nature Communications. 2023;14(1). Available from: https://doi.org/10.1038/s41467-023-41343-1

2.

Doerr S, Majewski M, Pérez A, Krämer A, Clementi C, Noe F, et al. TorchMD: A Deep Learning Framework for Molecular Simulations. Journal of Chemical Theory and Computation. 2021;17(4):2355–63. Available from: https://doi.org/10.1021/acs.jctc.0c01343