RNADynNet: Unified RNA trajectory generation and dynamics representation learning.Abstract
RNADyn brings together standardized RNA molecular dynamics data and a unified model for generating and understanding RNA dynamics. RNADynBench comprises 2,585 quality-controlled, 100-ns all-atom trajectories with leakage-controlled splits. RNADynNet uses a shared backbone for trajectory generation and dynamics fingerprint extraction from a single conformer. Coordinate denoising, single-frame-to-trajectory alignment, and physical grounding connect trajectory generation with representation learning. The benchmark evaluates molecular geometry, fluctuations, and conformational distributions, while single-conformer fingerprints capture information about MD-derived motion and support transfer to RNA-ligand binding prediction.
Yiming Huang
PhD Candidate
My research focuses on topological deep learning for complex and higher-order graph structures, as well as generative models, graph representation learning, and foundation models.
