Our lab works in the general area of computational structural biology although it also includes an experimental component. We carry out theoretical and computational research and develop software tools, with our efforts being guided by a variety of applications of biomedical importance. In the past we have elucidated the structural and energetic origins of protein-protein and protein-nucleic acid interactions, developed methods for protein structure prediction, and detected novel structural and functional relationships between proteins based on their geometric similarity. We currently focus on two distinct areas: the exploitation of structural information to predict protein function on a genome-wide scale and the molecular basis of cell-cell recognition. The former topic is the subject of the current proposal which focuses on the prediction of protein-protein interactions (PPIs) and protein interaction networks. Our overarching goal is to provide a structure-informed perspective in multiple areas of systems biology, thus filling a major gap in this rapidly growing area of biomedical research. Our research plans are derived from our development of the PrePPI algorithm and corresponding database of human PPIs. PrePPI provides proteome-wide structure-based predictions of PPIs, and discovers relationships not obtainable from other methods. The P-HIPSTer algorithm, which is derived from PrePPI, offers analogous information for virus-human PPIs for 1000 human-infecting viruses. The reliability of both resources has been validated experimentally, and both have revealed novel biological insights. PrePPI, in common with other PPI databases, is cell-context independent and, for example, does not distinguish among tissue and tumor types. To address this challenge, we developed the OncoSig algorithm that uses machine learning methods to combine PrePPI with regulatory interactions from patient genomic data. The generation of tumor-specific lists of PPIs, called SigSets, can then be mapped onto a context-dependent PPI network, or SigMap. We have also developed novel methodologies that link protein structure space with chemical compound space. The current proposal builds on these accomplishments with new methodological developments and new applications to network biology. We plan to integrate PrePPI with PPI information derived from genetic interactions derived from the correlation of gene profiles across many conditions (e.g. tumor types, cell lines or drug treatments). This will provide an unprecedented structure- and context-dependent view of protein interaction networks. Other plans include the extension of PrePPI to non-human genomes and the extension of P-HIPSTer to bacterial pathogens. Our overall vision includes the development of an integrated set of software tools and databases that will advance cutting edge biomedical applications. These tools will range in scope from protein- protein interaction networks, structure-derived protein function annotation and to the linking of network biology to chemical compound space which will suggest druggable targets within networks and provide leads for small molecules that can target individual proteins.
This project focuses on the modeling and incorporation of protein three-dimensional structures across the human proteome to further our ability to describe disease-related protein-protein interaction networks in tissue- and tumor-specific contexts. These structure-informed networks will reveal mechanistic insights regarding cellular processes underlying human diseases ranging from cancer to those resulting from viral and bacterial infection. Our results will also provide hypotheses regarding drugs that can target proteins in these networks.