Humans have 493 typical protein kinase domains. 437 of these are catalytically active kinases and 56 are pseudokinases. From structural bioinformatics of substrate-bound kinases, we have devised criteria for identifying the active form of a protein kinase. These criteria include the conformation of the DFG motif as well as the positions of the N and C-terminal segments of the kinase activation loop that accommodate the binding of substrate. With this information, we have modeled all 437 catalytic kinases in their active form with template-based modeling in AlphaFold2. By clustering the activation loop conformations of inactive forms of catalytic kinases, we have identified biologically relevant inactive forms of dozens of kinases. This has enabled modeling of many human kinases in their dominant inactive forms. We have undertaken similar steps to understand and model the 56 human kinases. With a novel score that utilizes AlphaFold's PAE matrix to score domain-domain and protein-protein interactions (github.com/DunbrackLab/IPSAE), we are modeling intramolecular/interdomain interactions within kinases and PPIs of human kinases, including activating, inhibiting, and substrate interactions. Where possible, these models are validated with experimental data in collaboration with colleagues at Fox Chase Cancer Center and elsewhere.