RNADyn: A Benchmark for Generating and Understanding RNA Dynamics

TL;DR

The paper introduces RNADynBench, a benchmark for RNA molecular dynamics, and RNADynNet, a unified model for RNA dynamics learning. RNADynNet generates trajectories and extracts dynamics fingerprints from a single conformer. It achieves high RMSF correlations and comparable agreement with MD-derived dynamics.

Why It Matters

Understanding RNA dynamics is crucial for understanding its functions and interactions. The proposed benchmark and model can facilitate the development of more accurate RNA dynamics simulations, leading to advancements in fields like drug discovery and RNA-based therapies. This can ultimately improve our understanding of RNA's role in various biological processes.

Key Findings

• RNADynBench provides 2585 quality-controlled RNA trajectories
• RNADynNet achieves RMSF correlations of 0.875 and 0.766
• Physical grounding improves generated dynamics and recoverable physical information