Rosta Research

Research

Multiscale mechanisms and molecular design.

We integrate QM/MM reaction-path sampling, atomistic molecular dynamics, enhanced sampling and kinetic models to connect microscopic mechanisms with thermodynamic and experimental observables.

01

Catalysis & reaction mechanisms

Resolve chemical reactivity in heterogeneous biomolecular environments.

Enzymes accelerate reactions through a precisely organised network of electrostatics, proton transfers, metal coordination and conformational motion. We use quantum mechanics/molecular mechanics calculations to resolve these contributions and test competing mechanistic hypotheses.

Phosphate chemistry

Hydrolysis and transfer reactions in nucleotide-processing and signalling enzymes.

Metalloenzymes

Metal identity, coordination geometry and their influence on selectivity and reaction barriers.

Protein redesign

Mechanism-based sequence changes that recover or reshape catalytic activity.

QM/MM workflows

Reaction paths, free-energy profiles and high-level electronic-structure refinement.

Phosphatase and pyrophosphatase reaction schemes with enzyme-family statistics
01Phosphatase and pyrophosphatase mechanisms converge on selective Mg2+ coordination.
Mg-pinch coordination in a phosphatase active site
02In phosphatases, Mg2+ clamps the phosphate groups flanking the scissile P–O bond.
Mg-pinch coordination in a pyrophosphatase active site
03Pyrophosphatases reproduce the Mg-pinch geometry to polarise the target bond for cleavage.
Mg-pinch coordination in NTP phosphatase and pyrophosphatase chemistry. Dudás et al., JACS 2026, CC BY 4.0.
02

Free energies, kinetics & pathways

Recover thermodynamic landscapes and rare-event kinetics.

Biomolecular function depends on rare events and distributed ensembles. We build analysis and enhanced-sampling approaches that recover equilibrium populations, kinetic pathways and barrier-crossing rates from simulation data.

Enhanced sampling

Biasing strategies that access states and transitions beyond conventional simulation timescales.

Markov models

Network descriptions of molecular state-to-state dynamics.

DHAM methods

Dynamic histogram analysis for free energies and kinetics from biased trajectories.

Electron transfer

Rate calculations that connect electronic coupling, solvent response and molecular motion.

KOMBI position-dependent optimal bias The black unbiased free-energy landscape from the KOMBI paper remains fixed. Its exact optimised profiles for starting states 5, 35 and 65 fill the valleys to create a downhill biased landscape toward state 88. state FES (kcal/mol) Starting on the unbiased landscape start 05 → target 88 KOMBI · Wei, Dudás & Rosta
01The bias fills the valleys. The optimised landscape then slopes downhill towards the target.
Diffuse exploration Targeted KOMBI pathway KOMBI · two-dimensional benchmark
02A kinetic route emerges. Optimising first passage replaces diffuse exploration with directed motion.
Pathway conservation

fij = πiKij(1 − qi)qj

J+ij = [fij − fji]+,   Ĵ = J+/ΣJ+

ΩJ = Σij min(Ĵ(0)ij, Ĵ(b)ij),   ΩJ = 1 ⇔ conserved pathway

Reactive-current comparison Wang, Fadhluddin & Rosta
03The objective changes the flux. Current overlap quantifies how strongly the pathway is conserved or redirected.
Distance-resolved Ras phosphate-release pathway from magnesium coordination to solvent 3.3Å7.0Å Ras phosphate release · Wang et al.
04Phosphate exits the active site. The Mg–P coordinate resolves departure from the coordination sphere.
03

Molecular design & discovery

Translate molecular free energies into prospective design decisions.

We combine binding free energies, molecular dynamics and scalable screening to evaluate ligands, protein mutations and cooperative complexes. The goal is not only to predict a score, but to understand which interactions make a design work.

Molecular glues

Cooperativity and ternary-complex stabilisation in targeted protein degradation.

Ligand discovery

Structure-based screening and community benchmarks for challenging targets.

Mutational scans

Mechanistically guided evaluation of protein sequence space.

Computational design

Workflows that combine simulation accuracy with decision-making at scale.

Molecular glues in practice

Published structures and free-energy calculations connect ternary-complex geometry to measurable cooperativity.

CRBN and IKZF2 ternary complex with the white-stick molecular glue bound in the CRBN pocket 01Ternary-complex geometry. The glue stabilises a new protein–protein interface between CRBN and IKZF2. Six molecular-glue ligand structures and their binding poses in the CRBN and IKZF2 interface 02A ligand series probes the interface. Small chemical changes reorganise the ternary binding mode. Calculated molecular-glue cooperativity plotted against experimental ternary-complex measurements 03Cooperativity becomes a design signal. Free energies distinguish stabilising and destabilising glue candidates.

Dudas, Athanasiou, Mobarec & Rosta · J. Chem. Theory Comput. 2025

Scientific workflow

Prediction grounded in mechanism.

Each project uses the level of theory and sampling strategy required by the question—not a one-size-fits-all pipeline.

01
An observable selected along a molecular coordinate

Frame the observable

Define the experimental quantity, uncertainty and molecular hypothesis.

02
Multiple molecular trajectories representing an ensemble

Construct the ensemble

Prepare chemically realistic states and sample the relevant conformational space.

03
A transition across a molecular free-energy barrier

Resolve the mechanism

Calculate pathways, thermodynamics, kinetics or binding cooperativity.

04
Calculated values compared with a validated trend

Test the prediction

Compare against experiment and use disagreement to refine the model.

Methods

Tools across electronic, atomistic and statistical scales.

QM/MMMolecular dynamicsFree-energy perturbationUmbrella samplingMarkov state modelsDHAMElectronic structureReaction-path optimisationMachine learning
Multiscale molecular modelling illustration connecting chemical states, an energy landscape and a ligand-bound protein structure
Electronic rearrangement, rare-event pathways and biomolecular structure within a common multiscale framework.

Collaborate

Mechanistic hypotheses that require quantitative molecular resolution.

We collaborate on problems where multiscale simulation can discriminate reaction pathways, quantify free-energy differences or interpret kinetic observables.

Discuss a project →