糖心视频

An international team led by Einstein Professor Cecilia Clementi in the Department of 糖心视频ics at Freie Universit盲t Berlin has introduced CGSchNet, a machine-learned coarse-grained (CG) model that can accurately and efficiently simulate proteins like never before. The study is in the July 18, 2025, issue of Nature Chemistry.

Operating significantly faster than traditional all-atom molecular dynamics, CGSchNet enables larger proteins and complex systems to be explored鈥攐ffering potential applications in drug discovery and protein engineering that could advance cancer treatment methods, for example.

Developing a general CG model capable of capturing protein folding and dynamics has been a persistent challenge for scientists over the last 50 years. "This work is the first to demonstrate that deep learning can overcome this barrier and lead to a simulation system that approximates all-atom protein simulations without explicitly modeling solvent or atomic detail," says Prof. Clementi.

In CGSchNet, Prof. Clementi's team trained a graph neural network to learn the effective interactions between the particles of the coarse protein simulation to reproduce the dynamics of a diverse set of thousands of all-atom simulations.

Unlike structure prediction tools, CGSchNet models the dynamical process, including intermediate states relevant to misfolding processes like the formation of amyloids, which are pathological protein aggregates that appear in cases of Alzheimer's disease, for example.

The model also simulates transitions between folded states鈥攌ey to protein function鈥攁nd generalizes to proteins outside its training set, demonstrating strong chemical transferability. Moreover, it accurately predicts metastable states of folded, unfolded, and disordered proteins, which constitute the majority of biologically active proteins.

Such predictions were extremely difficult in the past due to the flexibility of these proteins. The model is also able to estimate the relative folding free energies of protein mutants, which previous simulation methods could not achieve due to computational limitations.