Biophysical Society Thematic Meeting | Bucharest 2026

Biophysics of Membrane Reactions in Brian

Wednesday Speaker Abstracts

FROM CORRELATION TO CAUSATION IN MOLECULAR DYNAMICS: PROBING INFORMATION TRANSFER IN GPCR LIGAND UNBINDING Edina Rosta 1,2 ;

1 UCL, Physics, London, United Kingdom 2 ELTE, Chemistry, Budapest, Hungary

Understanding how information is transferred through complex molecular systems is essential for developing a mechanistic picture of biomolecular function, particularly in systems such as G protein–coupled receptors (GPCRs), where long-range communication underpins activation, signaling, and ligand kinetics. Despite major advances in molecular simulation and machine learning, identifying the key interactions that govern dynamical processes such as ligand unbinding, and establishing their causal roles remains a central challenge. Here, we build upon recent advances in enhanced sampling and machine learning approaches for ligand unbinding kinetics and their application to GPCR systems, extending these frameworks toward a causal interpretation of molecular dynamics. Our approach first integrates high-dimensional simulation data with machine learning-based feature extraction to identify candidate variables associated with transition-state dynamics and kinetic bottlenecks. We then introduce a novel framework for assessing causal relationships among these variables, enabling the systematic identification of key drivers of information flow and mechanistic change during ligand dissociation. Applying this methodology to the muscarinic M3 receptor, we analyze the dominant interactions governing unbinding pathways and focus on a subset of critical variables that emerge from the underlying dynamics. We infer causal structure using statistical approaches we develop, providing a principled way to distinguish correlated features from mechanistically relevant determinants of kinetic behavior. This work establishes a bridge between molecular simulation, machine learning, and causal inference, offering a new ways of interpreting complex biomolecular processes. By moving beyond correlation-based analyses toward causal understanding, our framework might enable new avenues for rational drug design and for elucidating the fundamental principles governing molecular machines.

33