Designing molecules that take a protein apart, and working out whether the ones that look good on screen survive being simulated.
A molecular glue degrader for ROS1
ROS1 gene fusions drive a share of non-small-cell lung cancer, and the inhibitors that work on them tend to fail in the same two ways: on-target solvent-front resistance, G2032R in particular, and progression in the brain.
ROS1 has been degraded before, but only with heterobifunctional PROTACs — big molecules that struggle to cross into the central nervous system. A molecular glue degrader would be considerably smaller, which is exactly what you want against a cancer that so often metastasises to the brain. Nobody has reported one for ROS1 yet, and ROS1 offers little to build on: no native degron, and no validated glue-competent E3 surface.
So with Dr. KM Tahsin Hassan Rahit at the Bose Laboratory, Arnie Charbonneau Cancer Institute, University of Calgary, I have been working through the kinase molecular-glue literature systematically — a pre-registered review of degraders reported against ROS1-analogous kinases — asking which discovery routes actually transfer to a new target, which E3 ligases keep recurring, and what that implies for designing a ROS1 glue from scratch. Manuscript in preparation.
Nine candidate inhibitors, in simulation
Running alongside: nine novel inhibitor candidates against DPP4 and SARS-CoV-2 targets. Docking has come back promising across the set — binding energies from −8 to −11 kcal/mol — and they are now going through molecular dynamics, so the poses get tested rather than trusted, which is the whole point of the exercise.
The comparison that matters is against the active pharmaceutical ingredients already on the market. A good docking score measured against nothing in particular is not worth much.
How the work is done
Structure preparation and ligand docking to find candidate binding poses, with the pocket and its contact residues examined directly, and surface properties — including hydrophobicity — used to characterise the site. Poses worth keeping then go into molecular dynamics in explicit solvent, to see whether a binding mode that looks favourable also holds once the system is allowed to move.
