Research Areas
Our lab develops and applies advanced computational methods to understand and predict the molecular interactions that determine drug action. We work at the interface of computer-aided drug design (CADD), molecular simulation, and machine learning enhanced modelling, focusing on physically rigorous predictions and automated workflows that can directly inform experimental discovery.
Computer Aided Drug Discovery
Our research in computer-aided drug discovery focuses on developing physics-based and data-driven computational methods that allow accurate prediction of molecular recognition, a cornerstone of rational drug design.
A major part of this work involves free energy perturbation (FEP) methodologies. We have developed and applied automated workflows such as QligFEP and QresFEP, which enable quantitative estimation of binding affinities and mutational effects. These methods provide rigorous predictions of ΔΔG values for ligand binding and stability changes and are validated against experimental measurements.
Our FEP workflows are used to:
- Rank and optimise lead compounds.
- Validate binding modes and selectivity mechanisms.
- Understand resistance mechanisms driven by receptor or enzyme mutations.
Mutated G proteins as drug targets
We apply our computational approaches to study how mutations in G protein coupled receptors (GPCRs) affect ligand binding, receptor activation, and signalling bias. These mutations often underlie altered drug efficacy, selectivity or resistance, making them key targets for rational design.
Using QresFEP and molecular dynamics simulations, we investigate how single point mutations influence receptor ligand binding free energies, identify structural determinants of altered activity, and design ligands that compensate for mutation induced effects.
Machine Learning and FEP Integration
Our lab is actively developing machine learning (ML) methods tailored to FEP workflows (Jespers et al., J. Chem. Inf. Model. 2025). We use ML to increase the efficiency and scalability of molecular simulations and improve predictive accuracy.
Specifically, we integrate ML to:
- Prioritise compounds for free energy calculations, reducing computational cost.
- Develop ML derived potentials and sampling strategies to accelerate convergence of FEP simulations.
- Automate parameter optimisation and uncertainty estimation in free energy predictions.
These developments allow us to bridge physics based and data driven modelling, creating a hybrid framework that combines the accuracy of molecular simulation with the efficiency of modern machine learning.
Together, our research areas aim to connect molecular level understanding with practical drug discovery, enabling quantitative, data driven and reproducible design of new therapeutics.