Project Summary/Abstract Our goal is to integrate structure and sequence-based approaches grounded in statistical mechanics to understand key features of molecular recognition by proteins, as well as protein fitness and function more generally. There are three aims. The first is focused on mapping complex conformational and fitness landscapes of proteins. We will integrate machine learning (ML) sequence co-variation and molecular dynamics structure- based approaches to analyze the sequence dependent conformational landscapes of proteins with a particular focus on the landscapes that govern the transitions of kinase family proteins from the active to functionally important inactive states. With our collaborators we will investigate the sequence dependent origin of the alternative binding modes for peptide substrates that TKs have compared with STKs. This work has important implications for the design of therapeutics targeting Src and other cytoplasmic TKs which appear to bind peptide substrates in an unusual way. Also, as part of our first aim, we will map the free energy landscape of the set of kinase P-loop active conformational states; this information is needed for efforts to develop anti-cancer therapeutics with higher specificity. The thrust of the second aim is to realize the power of the sequence- covariation ML methods we are developing to detect and decompose multi-residue allosteric interaction motifs within kinases and kinase protein complexes by evaluating connected mutational correlations that carry signatures of indivisible units of biological information flow. These methods are designed to identify allosteric pathways and will be generalizable to other protein targets we are working on including GPCRs and the HIV Intasome. Our third aim is to build on the structure-based molecular dynamics approaches we recently developed to determine the excess chemical potential of water molecules at the surface of proteins. The excess chemical potential provides quantitative information about position specific thermodynamic features of interfacial water molecules and their networks. Together with our experimental Cryo-EM collaborators we will use this information to refine solvent at the protein-solvent interface in Cryo-EM density distributions in an iterative and self-consistent way; this will substantially improve upon current methods for locating and refining solvent in Cryo- EM maps of proteins and their assemblies. This new refinement tool will be made available to the structural biology community. We will build on our recent development of classical density functional methods to evaluate how the displacement of specific solvent molecules located in protein binding sites affects the affinities and specificities of the small molecule ligands targeting these sites.