Design of Allosteric Inhibitors for Mutant EGFR by Combined use of Machine Learning and Molecular Dynamics Simulations

ABSTRACT The non-small cell lung cancer (NSCLC)-associated Epidermal Growth Factor Receptor (EGFR) mutant L858R/T790M confers resistance to first- and second-generation tyrosine kinase inhibitors (TKIs). Allosteric inhibitors, binding outside the ATP-binding site, have emerged as alternative therapeutic agents. Unlike orthosteric inhibitors, they preferentially stabilize EGFR in an inactive conformation. Hence, understanding the mechanistic basis of this inhibition is essential for designing potent allosteric inhibitors. In this study, we performed microsecond-scale molecular dynamics (MD) simulations on the inactive conformations of apo-EGFR and apo-EGFR L858R/T790M to explore how cancer-associated mutations induce a conformational switch toward the active kinase state. Simulations of allosteric inhibitor (EAI001) bound EGFR L858R/T790M revealed that inhibitor binding enhances the inactive state population by disrupting the K745-E762 salt bridge and modulating key structural elements. These findings revealed the structural basis of allosteric inhibition in EGFR L858R/T790M, . It also emphasized the importance of MD simulations in allosteric drug design for assessing the ability of the inhibitor to enhance the population of the inactive state of the mutant EGFR. We have also standardized a virtual screening protocol involving screening of an allosteric TKI library by docking, re-ranking them with a machine learning-based scoring function (SG-ML-PLAP), and evaluating top-scoring molecules by MD and MM/GBSA to identify molecules stabilizing the inactive state of the EGFR L858R/T790M . This approach identified 10 novel allosteric kinase inhibitors predicted to be more potent than EAI001. Overall, our results not only elucidate the mechanism of allosteric inhibition in EGFR L858R/T790M but also offer promising leads for the development of next-generation therapies to overcome TKI resistance.