Predicting RNA Structure and Dynamics with Deep Learning and Solution Scattering.

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• RNA molecules exhibit flexibility, making it challenging to model their solution conformations under varying conditions.
• Predicting accurate RNA structures requires accounting for ions and conformational plasticity, which are often missing in current models.
• Small-angle X-ray scattering (SAXS) is a key experimental technique for validating predicted RNA structures in solution.

Purpose of the Study:

• To develop a computational pipeline, SCOPER, for predicting and validating RNA structures in solution.
• To address the challenges of missing ions and conformational plasticity in RNA structure prediction.
• To improve the accuracy of SAXS profile fitting for RNA structures.

Main Methods:

• Integration of kinematics-based conformational sampling with the deep-learning model IonNet for predicting Mg²⁺ ion binding sites.
• Benchmarking SCOPER against fourteen experimental SAXS datasets.
• Analysis of the impact of ion content on RNA plasticity.

Main Results:

• SCOPER significantly improved the quality of SAXS profile fits by incorporating Mg²⁺ ions and conformational sampling.
• Increased ion content was observed to decrease RNA plasticity.
• The study highlights the importance of adjusting plasticity and ion density to prevent overfitting experimental SAXS data.

Conclusions:

• SCOPER provides an efficient tool for validating the solution state of RNA structures.
• The pipeline generates corrected atomistic models of RNAs, including essential ions.
• Accurate validation of RNA structures requires consideration of both ion interactions and conformational dynamics.