Most biological processes are intrinsically coordinated through complex networks of protein-protein interactions. The diversity and scale of these networks offer a highly selective and tunable way of modulating protein function. While genetic disease-associated mutations are known to be enriched at protein interacting interfaces, the effects of these mutations on interaction dynamics and signalling pathways remain largely unexplored.
To overcome this, I applied sophisticated data analysis and machine learning techniques to develop novel methods to rapidly and accurately explore the effects of novel mutations on protein interactions. These methods were validated using independent experimental data, outperforming previous approaches.
Harnessing the insights from these novel computational tools, I explored the role of mutations in Mycobacterium Tuberculosis, SARs-CoV-2 and rare genetic diseases. This showed that up to 60% of mutations involved in drug resistance and disease phenotypes would lead to significant disruption of key protein-protein interactions.
While this supported the important role protein-protein interactions were playing, developing therapies to target these interactions is more challenging than traditional active site pockets. To improve the efficacy of efforts to identify protein-protein interaction modifiers, I developed two innovative virtual screening approaches- one using information on the small molecule alone, and the other considering the target structure as well.
This work provides a foundation for systematically characterising and analysing protein-protein interaction networks. We are currently using it to identify the important and junk interactions for any organism, as part of clinical genomic and protein engineering pipelines and as a part of larger drug discovery efforts.