Publications

In this drug design project, we combine funnel metadynamics, NMR and biochemical assays to improve an inhbitor of Human peroxiredoxin 5, a target enzyme for innovative treatments against strokes.

We use a custom bonded model and replica-exchange simulations to look at how Ag+ binding promotes the folding of an 11-residue peptide into an $\alpha$-helix. We train an autoencoder neural network to map the folding landscape. There is even some NMR and Mass Spec in there.

With eABF and the string method, we find that the recovery stroke of myosin VI is initiated by lever-arm re-priming with only weak coupling to ATPase activation.

With extensive eABF calculations, we reveal the mechanism of the 36 deg rotation step of the ATP synthase rotor, clarifying the water-mediated proton transfer pathway and the role of experimentally-detected intermediates.

We combine cryo-EM, enhanced sampling MD simulations, machine learning, and in vivo and in vitro reconstitution to reveal the mechanisms and pathways for Pi release from F-actin, both from filament interior and barbed end.

Using data from the early days of the pandemic, we show that surface accessibility to antibodies -estimated using a simple computational protocol- is a predictor of mutation sites found in later variants of concern.

We build one of the first complete models of the SARS-CoV-2 spike protein, simulate a viral membrane patch with 4 spike copies for 2.5 microseconds, and devise a machine learning approach to predict immunogenic sites from the simulation.

We combine ABF and the string method to look at how a synthetic molecular muscle contracts, and show that an asymmetric pathway in which subunits bend and rearrange sequentially dominates despite the symmetry of the supra-molecule.

In situ cryo-ET and molecular modelling reveal the hinges of the SARS-CoV-2 a few months into the pandemic.

With crystallography, conventional MD and ABF, we characterise a structural state of the myosin motor domain that suggests a “ratchet-like” mechanism in which lever-arm repriming precedes active site closure and is not strongly coupled to it.

Crystallography, single-molecule studies and implicit solvent simulations reveal how myosin X is optimised to walk fast on F-actin bundles.

Crystallography, binding affinity assays and MD simulations reveal “molecular tinkering” at play in the evolution of microtubule-binding domains in myosin tails.