HITS-SIMPLAIX Joint Colloquium Edina Rosta – From Correlation to Control: Causality-Guided Enhanced Sampling of Biomolecular Dynamics

by Edina Rosta, University College London

Molecular simulations can reveal how proteins move, bind ligands and switch between functional states, but many biologically important transitions occur far beyond the timescales accessible to straightforward molecular dynamics. Enhanced-sampling methods address this limitation by biasing selected collective variables, yet a fundamental question remains: does a coordinate that correlates with a transition actually control it, and can we accelerate a rare event without changing its physical mechanism?

I will discuss our recent efforts to approach enhanced sampling from a kinetics- and causality-driven perspective. In Kinetics-Optimized Markovian Bias Integration (KOMBI), the bias is optimized directly to shorten the mean first-passage time to a target state, providing an efficient kinetic route to rare configurations and a natural starting point for further exploration of the underlying free-energy landscape. Extensions using reproducing-kernel Hilbert-space representations provide a more flexible description of the underlying kinetics, while transition-path diagnostics allow us to test whether acceleration preserves the original reactive pathways rather than creating artificial shortcuts.

A complementary direction is to distinguish correlated molecular motions from causal drivers through controlled molecular perturbations, and ultimately to use intervention-validated coordinates for sampling and control. I will illustrate these ideas with ligand-coupled conformational transitions in GPCRs and with nucleotide-dependent switching and phosphate release in Ras, where phosphate chemistry provides a particularly challenging example of coupled molecular motion, solvation and electrostatic control.

Together, these approaches suggest a route from observing rare molecular events to identifying their drivers, accelerating them, and ultimately learning how to control biomolecular function.

Dr. Edina Rosta is a Professor of Molecular Modelling at UCL. After completing her PhD at USC in the group of Arieh Warshel (2013 Chemistry Nobel Prize Laureate), she joined the Hummer lab as a Postdoctoral Research Fellow at the NIDDK, NIH. She took up a lecturer position at KCL Chemistry in 2012. In 2020 she joined UCL Physics. Current research in her group focuses on atomistic modelling, including hybrid quantum mechanics/molecular mechanics (QM/MM) simulations. To quantitatively and accurately assess how enzymes achieve their extraordinary efficiency and specificity in performing chemical reactions, she develops modern enhanced sampling methods including novel algorithms to calculate molecular kinetics from biased molecular simulations using the theoretical framework of kinetic networks and machine learning tools. Biological applications aim at understanding and molecular design concerning the most prominent chemical reactions of living organisms: phosphate transfer and cleavage.

REGISTRATION: