A Kinetic Framework to Unravel the Principles of Covalent Inhibition

Covalent inhibitors, i.e., small-molecule drugs that form covalent bonds with their targets, have made for some of the most successful stories in pharmacology. In principle, their tunable reactivity provides another dial for drug design on top of the noncovalent scaffold, offering selectivity and potency that scaffold alone cannot reach. Only after decades of eschewal during the late 1900s have their rational design and systematic application become possible. While the last few decades have seen a rapid acceleration to this end, the initiative is far from over, and many theoretical challenges stand in the way of a satisfactory understanding.

Understanding covalent inhibitors requires fundamentally different concepts from those that describe noncovalent inhibitors. An experimental evaluation of inhibition strength is forced to measure a timescale leading to equilibrium rather than a population distribution at equilibrium. However, covalent inhibition is controlled by a hierarchy of steps including diffusional encounter, noncovalent association, conformational rearrangements of the enzyme–inhibitor complex, covalent bond formation, and possibly bond cleavage. These processes can occur on vastly disparate timescales, spanning nanoseconds for fast conformational fluctuations to days for slow covalent bond turnover. The challenge is that an experimental assay yields a nonlinear combination of the underlying rates, from which the individual contributions are inseparable without additional insight.

To unravel the principles between the macroscopic mechanisms we observe and the microscopic mechanisms we predict, we introduce a kinetic framework comprising numerical simulations of the time evolution of arbitrary kinetic schemes, lossless experimental fitting, and Bayesian parameter exploration. The goal is to relate the data collected by atomistic simulations to the data collected by biochemical assays. In doing so, two questions become answerable. Which microscopic step must change to produce a desired effect, and which structural modification changes it?

The first part of this dissertation demonstrates the importance of this approach with a formal description of the surprisingly large effect that rapid conformational intermediates can have on experimental measurements of irreversible inhibition. The second part applies the same kinetic modeling machinery to the binding mechanism between Bruton's tyrosine kinase and the irreversible covalent inhibitor ibrutinib, where free energy calculations of the chemical steps and of the conformational preferences of the unlinked ligand together recover the measured covalent efficiency. The third part combines a general workflow of kinetic modeling with bespoke analysis to interpret unconventional experimental data on the binding of novel covalent kinase probes to ABL1 kinase, resolving kinetic parameters with less data and more accuracy. The fourth part applies the framework developed over the previous three parts to understand how small structural differences among a triad of ibrutinib-derived reversible inhibitors can cause drastic changes in residence time. Taken together, these four parts show that productive drug design requires resolving the microscopic steps underlying an observable and identifying their molecular determinants.