Predicting ²H NMR acyl chain order parameters with graph neural networks.

Area of Science:

• Biophysics
• Computational Chemistry
• Pharmacology

Background:

• 2H NMR order parameters are crucial for understanding how molecules affect lipid membrane properties like order, mobility, and permeability.
• Current methods for evaluating these effects are time-consuming and molecule-specific, hindering rapid screening.
• A need exists for a faster, more generalizable approach to assess molecular impacts on membranes.

Purpose of the Study:

• To develop a predictive model for 2H NMR order parameters of lipid membranes.
• To enable rapid screening of various molecules, particularly drugs, and their effects on membrane characteristics.
• To provide a computational tool for drug development and understanding potential side effects.

Main Methods:

• Development of a predictive model utilizing graph neural networks (GNNs).
• The GNN model learns molecular features to predict 2H NMR order parameters.
• The model is trained on data relating molecular structures to membrane parameter changes.

Main Results:

• The graph neural network-based model demonstrates sufficient accuracy in predicting 2H NMR order parameters.
• The model facilitates rapid assessment of how different molecules influence lipid membrane properties.
• The developed model offers a practical approach for screening potential drug candidates.

Conclusions:

• The GNN model provides a viable and accurate method for predicting 2H NMR order parameters.
• This approach significantly accelerates the assessment of molecular effects on lipid membranes.
• The study lays the groundwork for future research in computational membrane biophysics and drug discovery, with the model accessible via a web application.