Multiscale Molecular Modelling Center

Research Overview

Overview of the center's topics and methods.

Method Development & AI Data-Driven Modelling

The central pillar of our research is the development and refinement of physics-based and data-driven methods. In force-field optimization, we focus on formulating tailored loss functions, tuning weighting parameters, and leveraging multi-objective optimization algorithms that preserve physical validity. Concurrently, we harness modern machine learning (ML) to explore complex parameter landscapes (utilizing advanced techniques like quality-diversity optimization), train surrogate models for molecular dynamics, and predict high-dimensional potential energy surfaces. We also apply ML to evaluate molecular similarity, generating novel hypotheses for experimental validation (e.g., drug repurposing). To maximize computational efficiency, we integrate data-driven models directly with quantum mechanical calculations to build hybrid ML → QM workflows.

Molecular Structure & Conformational Prediction

Our group also predicts the molecular structures and conformations of small molecules. We explore conformational space by systematically rotating torsion angles, performing QM optimizations, and by integrating modern generative ML methods. For example, we embed QM calculations directly into the conformational search — enabling end-to-end ML → QM workflows that are applied to complex biopolymers (e.g., lignin). Our research investigates fundamental conformational phenomena, including ring-puckering dynamics in small cycles (e.g., cyclopentane), conformer ensembles for flexible hydrocarbons (e.g., octane), and σ-hole interactions between halogenated compounds and various organic molecules.

Membrane Proteins & Molecular Mechanisms

Molecule Dynamics Modelling

A major focus of our research is investigating membrane proteins and understanding their underlying molecular mechanisms. Using all-atom and coarse-grained molecular dynamics (AA-MD and CG-MD), we analyze the structure, dynamics, and function of membrane transport and receptor systems — including the nicotinic acetylcholine receptor (nAChR) across distinct functional states, the sodium-proton exchanger NHE1, and organic cation (OCT) transporters. Our work also encompasses characterizing potential substrates for monocarboxylate transporter 1 (MCT1) to elucidate their role in congenital metabolic disorders.

Variant Effect Prediction & Cross-Species Modeling

Beyond direct molecular simulations, we develop ML models for variant effect prediction (VEP). We examine missense variant functional consequences across species boundaries (e.g., in ENaC and nAChR) and extend curated human VEP datasets by integrating data from mouse and rat models. This cross-species perspective helps bridge sparse-data gaps and improves predictive power.

Drug Design & Drug Delivery

Our group researches computer-aided strategies towards drug design and drug delivery. One focus is the analysis of structure-activity relationships and virtual screening — leveraging our custom MolFuSE framework. In drug delivery, we investigate carrier-mediated transport and controlled release mechanics, with a special emphasis on modeling cyclodextrin complexes. We study these systems using hybrid ML-driven molecular dynamics simulations (ML-MD) to bridge theoretical predictions with empirical data.

Polymers & Materials

A further research area is dedicated to the modelling of polymers and materials. The focus is on the investigation of polymer melts, in particular of high-density polyethylene (HDPE) for extrusion blow moulding, which is simulated using coarse-grained molecular dynamics (CG-MD). These efforts are extended to additives and to polypropylene (PP). In addition, modified membranes are analytically characterized and structurally interpreted.