Half a million people develop drug resistant tuberculosis (TB) each year. Cases of drug resistant TB often result in poorer outcomes both healthwise and economically for patients, and many populations lack access to the resources needed to treat resistant TB. Increased antibiotic resistance has resulted in an urgent need to develop new, cost-effective drugs that are effective against Mycobacterium tuberculosis, the bacteria responsible for TB. In my research, I am using deep learning methods to design peptides that bind to the enzyme ClpP, a vital protease and known antibiotic target in M. tuberculosis. A class of drugs called Acyldepsipeptides (ADEP) have been shown to bind to ClpP and cause cell death in M. tuberculosis by preventing the formation of the ClpP complex with necessary ATPases, resulting in significantly lower proteolytic activity. We used the structure of ADEP as a basis for the peptide design and employed Rosetta, a macromolecular prediction and design software, to generate cyclic peptides bound to ClpP. I then used a sequence based deep learning tool to generate multiple sequences for each backbone design and computationally validated the resulting structures with AlphaFold, a highly accurate, machine learning based structure prediction tool. The structure of the ClpP binding interface resulted in it being a difficult target to design for with current deep learning methods. One peptide binder was predicted to bind to ClpP in our preliminary design rounds. We will chemically synthesize this binder and test it against ClpP in an enzyme inhibition assay. If the binder inhibits ClpP, it can serve as a basis for an effective and low cost drug that targets the ClpP enzyme in drug resistant TB. We will also expand and refine our design pipeline to produce more binder designs that can serve as viable drug candidates.

While vaccines help prevent infection from SARS-CoV-2, therapeutic drugs remain necessary to treat people who are infected. In my research, I develop peptide-based therapeutics, which are safe and readily bioavailable due to their low immunogenic response and low production cost. I help design peptide inhibitors against the Mpro enzyme, the main protease present in SARS-CoV-2. Mpro normally cleaves the virus’ pp1a and pp1ab viral polyproteins, which are important for viral replication and transcription. By creating a competitive inhibitor that binds to Mpro, I can prevent it from activating these proteins and thereby prevent viral proliferation. We began with computational design of peptide inhibitors using the stub-extension approach on the Rosetta Macromolecular Modeling suite, a peptide design computational platform. An effective drug both finds the active site of Mpro and binds to it more strongly than the natural substrate. First, our group started with selecting the most critical residues (stubs) that form well-characterized interactions in the active site. We then used Rosetta to add amino acids and cyclize the peptides. After generating cyclic backbones, we designed amino acid side chains with high shape and chemical complementarity to the active site. After filtering, we proceeded with the 50 best 7- to 11-mer peptide candidates. Afterwards, I used solid phase peptide synthesis to chemically synthesize these peptide designs, followed by cyclization and purification with High-Performance Liquid Chromatography. I helped to test their effectiveness in mass spectrometry-based inhibition assays. Overall, we have identified four promising peptides, and our most promising peptide is gzm1_1, which has an IC50 value (the concentration of peptide required to inhibit 50 percent of Mpro) of 0.076 µM. Through refining the structure of these inhibitors even more, I can improve their inhibitory effectiveness, enabling them to serve as the basis of an effective medication for people infected with SARS-CoV-2.