<ul><li>ConfRover is an autoregressive model introduced to simultaneously model protein conformation and dynamics from molecular dynamics (MD) trajectories.</li><li>The model supports both time-dependent and time-independent sampling and includes an encoding layer, a temporal module, and an SE(3) diffusion model as the structure decoder.</li><li>Experiments conducted on the ATLAS protein MD dataset show the effectiveness of ConfRover in learning conformational dynamics and its applicability to various downstream tasks.</li><li>ConfRover is the first model capable of sampling both protein conformations and trajectories within a single framework, providing a new and adaptable approach to learning from protein MD data.</li></ul>

Simultaneous Modeling of Protein Conformation and Dynamics via Autoregression

Discover more