Project Summary Protein-protein interactions (PPIs) are essential in cellular processes and human diseases. It is estimated that 80% of proteins rely on PPIs to perform their primary functions. Thus, modulating PPIs should be a powerful way to interfere with pathological pathways and treat human diseases. However, PPIs are challenging targets for small-molecule drugs due to the lack of druggable pockets to achieve sufficient drug affinity. Recent advances in covalent inhibition of kinases have shown that targeting hyperreactive cysteines with small electrophilic molecules can achieve increased potency, prolonged target engagement, and improved selectivity. We hypothesize that electrophilic molecules forming covalent bonds with hyperreactive cysteines on PPI interfaces may represent a new avenue for developing PPI-modulating drugs. Building on the recent development in Artificial Intelligence (AI) methods for protein structure modeling and analysis, the ongoing efforts to predict and model 3D structures of human PPIs in my sponsor’s lab, and my co-sponsor’s expertise in developing covalent inhibitors, this project aims to predict and validate druggable hyperreactive cysteines located within PPI interfaces. Using a Convolutional Neural Network trained on a large dataset of reactive cysteines to integrate the physiochemical environment around a cysteine in the 3D space, I will develop a predictor for cysteine reactivity. I will identify reactive cysteines on human PPI interfaces by integrating experimental results and predictions on PPI structures and cysteine reactivity. Based on my preliminary data, I expect to find thousands of PPI interfaces with hyperreactive cysteines. Next, I will analyze the protein surface pockets around hyperreactive cysteines on PPI interfaces using both established tools to evaluate pocket druggability and new AI methods to characterize the geometric and chemical fingerprints of a pocket. Comparing the fingerprint of a potential drug pocket against the surface pockets of the entire human proteome will allow me to identify pockets with unique features for specific drug targeting. Results from these analyses will be incorporated into a comprehensive online database of reactive cysteines on PPI interfaces and their druggability. Combining the above analyses with other structural (such as interface size and other components in a complex) and functional considerations, I will choose several dozen target PPIs to perform virtual screens using multiple established methods to identify covalent drug candidates. Several promising drug candidates supported by multiple methods will be tested experimentally through pull-down assays. Experimentally tested candidates will be further studied by affinity chromatography and mass spectrometry to evaluate the off-target binding partners. Overall, this project will provide valuable insights into the identification and targeting of hyperreactive cysteines within PPI interfaces, offering new...