Transfer seeding reveals divergent activation ensembles across tyrosine kinases

Abstract

Protein kinase activation involves coordinated rearrangements of conserved regulatory elements, including the Asp--Phe--Gly (DFG) motif and activation loop (A-loop), but whether tyrosine kinases follow a universal activation mechanism remains unclear. This motivates the need for a common protocol to generate comparable conformational ensembles despite uneven experimental structural coverage across kinases. Here we combine over 8.3 milliseconds of distributed unbiased molecular dynamics simulations with Markov state models (MSMs) to compare the conformational landscapes of the ABL, EGFR and MET kinase domains. To expand conformational coverage while retaining unbiased dynamics, we use a transfer seeding strategy that steers AlphaFold2 models derived from homologous templates to sample MET conformational states absent from available experimental databases. Our MSMs resolve metastable conformations not represented in the experimental structural set, including low-population DFG-inter intermediates along the MET activation pathway. We find that activation pathways are kinase-specific rather than universal. A-loop rearrangements contribute to the rate-limiting steps in ABL and MET, whereas EGFR is rate limited by a local rearrangement of the DFG motif. Building on established DFG geometry nomenclature, ensemble level characterization of the A-loop further distinguishes metastable conformations that share the same DFG geometry and reveal kinase-specific differences in phosphosite accessibility along their activation landscapes. Our results provide a systematic framework for comparative kinase ensemble characterization, integrating DFG geometry with A-loop and other regulatory elements to resolve kinase-specific states and the atomic-resolution structural rearrangements underlying their interconversion.