Understanding enzymes well enough to redesign them
I work on halohydrin dehalogenases, light-controlled enzymes and enzymes that build molecules on DNA. I simulate them with QM, QM/MM and MM, develop the methods those simulations need, and work with experimental collaborators who test the designs.
Research
Five scales, one enzyme
Quantum mechanics for the electrons, molecular mechanics for the protein, QM/MM to join them, and the methods I develop at each level. The ruler on the left tracks length and time; the panels on the right run live in your browser. Panels marked real data play my own calculations; the others are schematics of the method.
Where are the electrons, and where do they go when a bond breaks?
Semi-empirical and machine-learned potentials make it cheap enough to scan, optimise and search for transition states on hundreds of atoms, then check the key points with DFT.
What does the reaction look like at the atomic level, and what does it cost?
A 303-atom cluster model of a halohydrin dehalogenase active site, computed with AIQM2 through ORCA. Cyanide is pulled towards the epoxide carbon one step at a time. At the last step the epoxide C–O bond breaks and the catalytic tyrosine hands its proton to the leaving oxygen, while the arginine re-protonates the tyrosine.
How does the enzyme polarise the substrate before anything reacts?
Charges fitted in the gas phase miss the field of the protein. Fitting them inside the enzyme, with the protein as point charges, gives ligand parameters that match where the ligand actually sits.
Which conformations does the enzyme visit, and which ones are active?
Long simulations, projected onto a few collective variables, show which states an enzyme samples and how mutations shift the balance, such as the flexible N-terminal loop that controls HheG's activity and enantioselectivity.
Simulations point to the residues that matter; design tools and the lab test what to put there. In HheC, the single change F12Y in sequence motif 1 raised specific activity up to 5-fold and the apparent melting temperature by 10 K.