Artificial Intelligence Resolves Kinetic Pathways of Magnesium Binding to RNA

Area of Science:

• Biochemistry
• Computational Chemistry
• Molecular Biophysics

Background:

• Magnesium ions (Mg2+) are essential cofactors in numerous biological processes, including RNA function.
• Understanding magnesium's binding pathways to biomolecules like RNA is critical for elucidating its functional roles.
• Current simulation techniques struggle to capture the millisecond-timescale dynamics of magnesium's transition from outer-sphere to inner-sphere coordination.

Purpose of the Study:

• To resolve the molecular mechanisms and pathways of magnesium binding to RNA.
• To investigate the role of solvent molecules in the magnesium-RNA binding process.
• To develop accurate predictive models for solute-solvent coupling using machine learning.

Main Methods:

• Transition path sampling (TPS) was employed to study the binding pathways.
• Committor analysis was performed to identify key states in the binding process.
• A machine learning algorithm, specifically a deep neural network, was developed and optimized using hyperparameter tuning.

Main Results:

• The study revealed that water molecules immediately fill voids created by departing phosphate oxygens during magnesium binding.
• Water molecules from the first and second hydration shells actively participate in the coordinated exchange process.
• The optimized deep learning model accurately predicted commitment probabilities by recognizing crucial solvent structures.

Conclusions:

• Detailed insights into the ubiquitous solute-solvent coupling governing magnesium-RNA interactions were obtained.
• The findings highlight the critical role of solvent dynamics in biochemical reactions involving kosmotropic ions.
• This work provides a foundation for understanding similar interactions in various aqueous biochemical systems.