Mechanistic Dissection of Conformational Transition of Bicyclic Peptide via Molecular Modeling and Deep Learning.

Area of Science:

• Computational Chemistry
• Molecular Dynamics
• Deep Learning

Background:

• Molecular conformations critically influence properties like membrane permeability and binding affinity.
• Understanding conformational ensembles and transitions is key for effective molecule design.
• Current methods often provide static views, limiting insight into dynamic conformational changes.

Purpose of the Study:

• To introduce Internal Coordinate Net (ICoN) version 1 (v1), a deep learning model for analyzing cyclic peptide conformational dynamics.
• To enable the identification of transient conformations and torsional rotations.
• To generate atomistic transition pathways and provide mechanistic insights.

Main Methods:

• Trained a deep learning model (ICoN v1) on molecular dynamics (MD) simulation data.
• Utilized the model to learn the physics governing cyclic peptide conformational dynamics.
• Generated minimum-energy pathways (MEPs) in the model's latent space to create transition pathways.

Main Results:

• ICoN v1 successfully identified transient conformations and torsion rotations between energy minima.
• The model generated smooth, fully atomistic transition pathways, capturing detailed interactions.
• ICoN v1 demonstrated strong generalization beyond sampled MD data, producing novel pathways.
• Analysis revealed key residues involved in conformational transitions.

Conclusions:

• ICoN v1 provides a powerful tool for understanding and visualizing molecular conformational dynamics.
• The model's ability to generate transition pathways offers mechanistic insights for cyclic peptide design.
• This approach can significantly inform drug discovery by elucidating how molecular conformations transition.