KNOWLEDGE-DRIVEN ANTIVIRAL DISCOVERY

Our unpreparedness for the recent SARS-CoV-2 pandemic highlighted the need for continuous research and financial investments into antiviral research prior to likely unavoidable novel viral emergence. Broad-spectrum antiviral (BSA) drugs are a promising strategy to protect against emergent viruses; however, the development of such drugs has been challenging indicating a prime opportunity for using computational technologies to accelerate discovery. In this dissertation, we emphasize the “three R’s” of modern discovery, whereby data science enables the revision, reduction and (partial) replacement of wet-lab experimental antiviral research. We take an approach that spans from the population down to the molecular level, underscoring both the interconnectedness and importance of each level in antiviral discovery. We identify key viruses with high pandemic potential, regions of the world where the next outbreaks are more likely to occur, follow-up on viral cases that have occurred since our initial study, and summarize developments that can help us prepare for such outbreaks. We hypothesize that conserved binding sites in key coronavirus proteins can be explored for the development of BSA compounds, identified such conserved binding site residues across coronaviruses and validated our hypotheses with existing experimental data. Over the course of this thesis project, we have built a curated, annotated, and publicly available database of compounds tested in both phenotypic and target-based assays against high-threat viruses, and developed a knowledge-based computational hit discovery and experimental nomination strategy. This strategy was used to identify compounds with BSA activity and we report the results of this experimental effort herein. We also applied this multi-faceted cheminformatics mining approach to build a database of helicase inhibitors and select viral helicase inhibitors that underwent experimental testing the results of which are reported here. The knowledge-based curation and database generation also enabled us to build a high quality predictive Quantitative Structure Activity Relationship (QSAR) model which we used for virtual screening of compounds, some of which we nominated as inhibitors of Marburg Virus and report the initial experimental result here. The experimental nomination strategies in this dissertation aim to revise and reduce the time and cost of wet lab antiviral research, providing a cost-effective strategy to combat the lack of funding and interest in viral diseases after the initial emergence event.