Computational Methods to Identify and Target Druggable Binding Sites at Protein-Protein Interactions in the Human Proteome

Abstract

Protein-protein interactions are fundamental in cell signaling and cancer progression. An increasing prevalent idea in cancer therapy is the development of small molecules to disrupt protein-protein interactions. Small molecules impart their action by binding to pockets on the protein surface of their physiological target. At protein-protein interactions, these pockets are often too large and tight to be disrupted by conventional design techniques. Residues that contribute a disproportionate amount of energy at these interfaces are known as hot spots. The successful disruption of protein-protein interactions with small molecules is attributed to the ability of small molecules to mimic and engage these hot spots. Here, the role of hot spots is explored in existing inhibitors and compared with the native protein ligand to explore how hot spot residues can be leveraged in protein-protein interactions. Few studies have explored the use of interface residues for the identification of hit compounds from structure-based virtual screening. The tight uPAR•uPA interaction offers a platform to test methods that leverage hot spots on both the protein receptor and ligand. A method is described that enriches for small molecules that both engage hot spots on the protein receptor uPAR and mimic hot spots on its protein ligand uPA. In addition, differences in chemical diversity in mimicking ligand hot spots is explored. In addition to uPAR•uPA, there are additional opportunities at unperturbed protein-protein interactions implicated in cancer. Projects such as TCGA, which systematically catalog the hallmarks of cancer across multiple platforms, provide opportunities to identify novel protein-protein interactions that are paramount to cancer progression. To that end, a census of cancer-specific binding sites in the human proteome are identified to provide opportunities for drug discovery at the system level. Finally, tumor genomic, protein-protein interaction, and protein structural data is integrated to create chemogenomic libraries for phenotypic screening to uncover novel GBM targets and generate starting points for the development of GBM therapeutic agents.