2026-09-18T19:25:26Zhttps://digital.lib.washington.edu/server/oai/request

oai:digital.lib.washington.edu:1773/525802026-02-15T21:17:23Zcom_1773_4888col_1773_4943

Baker, David Humphreys, Ian 2024-10-16T03:16:56Z 2024-10-16T03:16:56Z 2024-10-16 2024 Humphreys_washington_0250E_27510.pdf https://hdl.handle.net/1773/52580 Thesis (Ph.D.)--University of Washington, 2024 The total set of potential protein-protein interactions (PPI) within an organism's proteome guides a plethora of potential biological processes at an organism’s disposal. Understanding these PPIs is critical to our understanding of biological systems, however identifying interactions with high accuracy is challenging. Medium to high-throughput experimental techniques for identifying protein interactions result in high rates of false-negatives and false-positives. However, protein interactions are typically evolutionarily conserved resulting in co-varying mutations at the interface between complexes. Deep learning based protein structure prediction models capture coevolutionary information at significantly higher resolution than statistical methods and we exploit this coevolutionary signal to computationally predict protein-protein interactions with high accuracy based on gold-standard benchmarks. We create and apply bioinformatic and deep learning pipelines to rapidly predict proteome-wide protein-protein interactions in Bacteria and Eukaryotes to identify novel interactions and provide high resolution structural models to better understand their biological ramifications. application/pdf en_US CC BY Biochemistry Bioinformatics Microbiology Molecular and cellular biology Deep learning and coevolution reveal proteome-wide protein-protein interactions Thesis Open Access open.access