Jul 2026· Computational biology and chemistry· Vol 125, pp.
109263
· 0 citations· 34 references
Medicine
TL;DR
CMD-PLA is proposed as a dynamics-aware framework for protein-ligand affinity prediction, and the conformational evolution of the ligand is explicitly modeled as a pocket-dependent dynamical process rather than an isolated static update.
Abstract
Protein-ligand affinity prediction is fundamental to structure-based virtual screening and lead discovery. However, most existing methods rely on a single static conformation of the complex and exhibit high sensitivity to pose uncertainty and conformational noise. To address this limitation, CMD-PLA is proposed as a dynamics-aware framework for protein-ligand affinity prediction. Within this framework, pocket-conditioned molecular dynamics refinement is performed, and the conformational evolution of the ligand is explicitly modeled as a pocket-dependent dynamical process rather than an isolated static update. Furthermore, a dual-view atomic representation is adopted to separately capture the intra-molecular covalent structure and the inter-molecular interaction geometry. Global representations of the ligand and the pocket are also incorporated to complement the local geometric modeling. Experimental results demonstrate that CMD-PLA achieves robust performance across various settings. The provided case study further illustrates the interpretability of the model.
Scaffold-Guided Structure Refinement is presented, leveraging information on known binders within ligand series targeting a specific protein, based on the observation that shared molecular scaffolds among binders exhibit conserved binding modes.
J. Pletzer-Zelgert, Matthias Rarey, Bernd Kuhn· Journal of Medicinal Chemist...· 0 citations
A category-stratified, statistically powered benchmark comparing pose prediction from receptor conformational ensembles against AlphaFold2, used as a matched static-structure baseline, across 29 protein–ligand systems spanning cryptic-pocket, induced-fit, water-mediated, and autoimmune-indication target classes is presented.
Ryan Varghese, Pooja Tiwary, Krishil Oswal· bioRxiv· 0 citations
The quality of protein-ligand binding affinity prediction is often limited by the accuracy of positioning the ligand correctly inside the binding site. However, pose accuracy is a secondary concern in high-throughput virtual screening, which is the application scenario in mind when most docking protocols are developed. On the other hand, similar protocols are also applied to position ligands in the binding pocket prior to free-energy calculations, and the accuracy of docking protocols is not well known in this case, especially when structures of analogues are available as templates to guide the docking as is typical in lead optimization. Docking benchmarks typically focus on the ability of docking methods to generate poses with a low RMSD to crystal structures, but for physics-based affinity prediction methods, we also need the ability to identify potential alternate binding modes of a new ligand that might be viable. This is especially relevant for ligand modifications which break local symmetry, leading to multiple potential substituent orientations, such as substitutions of phenyl rings. Here, our focus is on assessment of pose prediction methods when the bound structure of a reference ligand is known (typical in structure-based drug design) and the likely binding mode(s) of a related compound are needed, and we focus on cases where the new compound has multiple potential binding modes. To assess templated docking protocols on their ability to identify binding modes when starting from a solid reference structure, we collected a set of 60 complex structures from the PDB which have more than one ligand binding mode – shown in the PDB records as alternative locations in the ligand. Our results suggest success rates of only 30-50% for finding alternative binding modes, which are modest compared to benchmarks of the same docking programs on the Astex diverse set (70-90% success rates). Overall, we conclude that docking methods would benefit from further tuning or improvement to become more effective in lead optimization.
Ažbeta Kubincová, S. S. Çınaroğlu, Jianna Ongsioco et al.· Journal of Chemical Informat...· 0 citations
RNA molecules explore heterogeneous conformational ensembles that are essential for their biological function and molecular recognition, yet this intrinsic flexibility poses a major challenge for structure-based drug discovery. In particular, the absence of well-defined binding pockets in static structures limits the identification of ligandable sites. Here, we present an integrative ensemble-based approach that combines enhanced-sampling molecular dynamics simulations with Nuclear Magnetic Resonance data to characterize the conformational landscape of the HIV-1 TAR RNA at atomic resolution. Starting from extensive sampling, we refined the resulting conformational distribution through maximum-entropy reweighting to achieve quantitative agreement with experimental data. Analysis of the reweighted ensemble reveals a diverse set of conformational substates, including compact arrangements that exhibit pocket features compatible with ligand recognition and overlap with known ligand-bound structures. At the same time, highly ligandable conformations, which are only marginally populated, might nonetheless be critical for RNA recognition. Our results demonstrate that integrative ensemble modeling can reveal pharmacologically relevant RNA conformations that are not apparent from experimental static structures, providing a framework for ensemble-based strategies in RNA-targeted drug discovery.
Stefano Bosio, Vincent Schnapka, Mattia Bernetti et al.· bioRxiv· 0 citations
The kinetics of protein-ligand binding systems are increasingly recognized as a key determinant of drug efficacy, yet remain far harder to compute than binding affinities. Existing kinetics methods either bias the dynamics along a collective variable (CV), demanding careful system-specific CV design, or use path sampling, which keeps the dynamics unbiased but can struggle to converge rates out of deep free-energy wells and often relies on hand-engineered descriptors. By combining the `best of both worlds', we propose a method to compute accurate kinetics for general ligand-unbinding problems at modest computational expense and minimal fine tuning, building on the AI for Molecular Mechanism Discovery (AIMMD) path sampling framework. To avoid the need for feature engineering, we opt for modelling the committor with a single descriptor-free, equivariant graph neural network shared across all systems. We also partially flatten deep bound-state wells with a static, basin-restricted bias potential. This improves convergence by lifting the path sampling state boundary out of regions, where the committor is hard to learn, while leaving the reactive region strictly unbiased. Across host-guest and protein-ligand systems spanning roughly 17 orders of magnitude in residence time, the method robustly recovers rates in line with reference and experimental values. Simultaneously, and without further sampling, it also reconstructs the underlying unbinding mechanisms. We additionally find that accurate rates do not require globally accurate committor models, allowing for efficient kinetics estimation even in a low-data training regime. Requiring little system-specific setup, our approach offers an efficient and broadly generalizable route to binding kinetics, and its shared committor architecture lays crucial groundwork for probing structure-kinetics relationships across ligand series in drug discovery.
This work introduces a method to smoothly transition from physics-based to knowledge-based predictions based on the uncertainty of each model and shows that combining structure-based and ML models significantly improves the prediction accuracy if training data is limited, whereas the weighting smoothly shifts from docking to ML as more data is acquired.
Ažbeta Kubincová, David L. Mobley· Journal of Chemical Informat...· 1 citation