We are thrilled to share new research on "Top-down machine learning of coarse-grained protein force-fields." This work focuses on the development of a unique approach that combines neural network potentials (NNPs) with differentiable trajectory reweighting for protein folding.
Trained using only experimental protein structures and short molecular dynamics trajectories, our coarse-grained NNPs maintain native structures and fold proteins from unfolded states. Comparative analysis with other coarse-grained force fields indicates that our NNP-based approach yields results commensurate with conventional methods.
This research aligns with our commitment to accelerate computerized drug discovery through cutting-edge algorithms. We are optimistic about its applicability in the field.
Read the article to know more!