Structure-Kinetic Relationship for Drug Design Revealed by a PLS Model with Retrosynthesis-Based Pre-Trained Molecular Representation and Molecular Dynamics Simulation

Deciphering the Selectivity of CBL-B Inhibitors Using All-Atom Molecular Dynamics and Machine Learning

We employ a combination of accelerated molecular dynamics and machine learning to unravel how the dynamic characteristics of CBL-B and C-CBL confer their binding affinity and selectivity for ligands from subtle structural disparities within their binding pockets and dissociation pathways. Our predictive model of dissociation rate constants (k off) demonstrates a moderate correlation between predicted k off and experimental IC50 values, which is consistent with experimental k off and τ-random accelerated molecular dynamics (τRAMD) results. By employing a linear regression of dissociation trajectories, we identified key amino acids in binding pockets and along the dissociation paths responsible for activity and selectivity. These amino acids are statistically significant in achieving activity and selectivity and contribute to the primary structural discrepancies between CBL-B and C-CBL. Moreover, the binding free energies calculated from molecular mechanics with generalized Born and surface area solvation (MM/GBSA) highlight the ΔG difference between CBL-B and C-CBL. The k off prediction, together with the key amino acids, provides important guides for designing drugs with high selectivity.

Fingerprinting Interactions between Proteins and Ligands for Facilitating Machine Learning in Drug Discovery

Molecular recognition is fundamental in biology, underpinning intricate processes through specific protein-ligand interactions. This understanding is pivotal in drug discovery, yet traditional experimental methods face limitations in exploring the vast chemical space. Computational approaches, notably quantitative structure-activity/property relationship analysis, have gained prominence. Molecular fingerprints encode molecular structures and serve as property profiles, which are essential in drug discovery. While two-dimensional (2D) fingerprints are commonly used, three-dimensional (3D) structural interaction fingerprints offer enhanced structural features specific to target proteins. Machine learning models trained on interaction fingerprints enable precise binding prediction. Recent focus has shifted to structure-based predictive modeling, with machine-learning scoring functions excelling due to feature engineering guided by key interactions. Notably, 3D interaction fingerprints are gaining ground due to their robustness. Various structural interaction fingerprints have been developed and used in drug discovery, each with unique capabilities. This review recapitulates the developed structural interaction fingerprints and provides two case studies to illustrate the power of interaction fingerprint-driven machine learning. The first elucidates structure-activity relationships in β2 adrenoceptor ligands, demonstrating the ability to differentiate agonists and antagonists. The second employs a retrosynthesis-based pre-trained molecular representation to predict protein-ligand dissociation rates, offering insights into binding kinetics. Despite remarkable progress, challenges persist in interpreting complex machine learning models built on 3D fingerprints, emphasizing the need for strategies to make predictions interpretable. Binding site plasticity and induced fit effects pose additional complexities. Interaction fingerprints are promising but require continued research to harness their full potential.