Cheminformatic Approaches to Decipher Natural Product –Target – Disease Associations

Cancer is the second leading cause of death in the United States, with many of those deaths attributed to lung cancer. NSCLC accounts for nearly 85% of all lung cancer cases, making NSCLC a leading cause of cancer-related death in the United States. A chemistry-driven de novo discovery strategy utilizing cheminformatics recently identified ikarugamycin (IKA) as a potent and selective inhibitor of cellular proliferation amongst NSCLC cell lines. However, a detailed characterization of IKA and several analogs has yet to be performed within the context of NSCLC. This work aimed to further investigate the relationship between IKA analogs and the effect that structural diversity may have on the potency and selectivity of its’ antiproliferative properties concerning NSCLC. The chemical characterization and biological cytotoxicity profiling of IKA and its’ analogs against several NSCLC cell lines will drive the development of IKA towards clinical relevance. Biological evaluation of several IKA analogs revealed that the double bond within the 5-6-5 ring moiety is crucial to selective antiproliferative activity against NSCLC HCC44, H23, and Calu-1. All selective analogs exhibited an IC50 value within the 0.09-1.00 M range. Along with the strong toxicity trends we have already observed, IKA also appears to be interacting with a novel biological target outside of the commonly acted on genes (i.e., EGFR, ALK, BRAF, ROS1). This work also explored strategies to embed synthetic handles for future click pull-down experiments for target identification.Concurrently, this thesis contains work on the utilization of the NP Atlas as a compound repository for virtual drug screening – a first of its kind. To address the emerging viral epidemic of SARS-CoV-2, I employed the open-source Chemprop algorithm to train a Directed-Message Passing Neural Network (DMPNN) to identify key chemical descriptors that can be attributed to disrupting the viral replication mechanism. I utilized open-source screening data provided by the National Center for Biotechnology Information (NCBI) to train the DMPNN and subsequently utilized the trained neural network to score compounds found within the NP Atlas. Through this process, I was able to identify and validate the targeted interaction between closthioamide and the main protease of SARS-CoV-2. This work serves as a proof of principle for adjacent computational drug discovery strategies that may help scientist prioritize their efforts and lower the cost of resources necessary to screen libraries that are greater than 24,000 molecules