Cancer Interactome: An <i>in-silico</i> Novel Approach for Elucidating Cancer Protein-Protein Interactions (CPPIs) Using Structural Graph Learning Augmented with Attention

Protein–protein interactions (PPIs) are intricate and vital components of cellular processes, coordinating the different biological processes. The accurate prediction of PPIs is significant for unraveling the functionality of known cancer proteins, comprehending the underlying oncogenesis and determining the possible therapeutics. Although cancer- specific PPI datasets are crucial, but are currently unavailable. The research addresses the gap by curating a novel structural Cancer PPI (CPPI) dataset encompassing significant cancer proteins to represent an extensive picture of Cancer Protein interactions. The paper introduces an innovative approach integrating graph-based methods, viz: GraphSAGE and GIN with attention, to identify and interpret intricate interactions for accurate prediction of CPPIs. Moreover, incorporating attention mechanisms allows the model to prioritize pertinent information while transmitting messages, improving interpretability and predictive capabilities. The performance of the proposed approach is systematically evaluated, compared, and cross-validated with the existing baseline models. The results demonstrate the enhanced remarkable comparative outcomes and highlight the potential of attention-augmented graph-based learning for providing vital insights into the complex world of CPPIs by capturing the structural features and the sites that regulate CPPIs.