No description available.

No description available.

There exist many real-world applications such as recommendation systems, document retrieval, and computational biology where the correct ordering of instances is of equal or greater importance than predicting the exact value of some discrete or continuous outcome. Learning-to-Rank (LTR) refers to a group of algorithms that apply machine learning techniques to tackle these ranking problems. Despite their empirical success, most existing LTR models are not built to be robust to errors in labeling or annotation, distributional data shift, or adversarial data perturbations. To fill this gap, we develop four LTR frameworks that are robust to various types of perturbations. First, Pairwise Elastic Net Regression Ranking (PENRR) is an elastic-net-based regression method for drug sensitivity prediction. PENRR infers robust predictors of drug responses from patient genomic information. The special design of this model (comparing each drug with other drugs in the same cell line and comparing that drug with itself in other cell lines) significantly enhances the accuracy of the drug prediction model under limited data. This approach is also able to solve the problem of fitting on the insensitive drugs that is commonly encountered in regression-based models. Second, Regression-based Ranking by Pairwise Cluster Comparisons (RRPCC) is a ridge-regression-based method for ranking clusters of similar protein complex conformations generated by an underlying docking program (i.e., ClusPro). Rather than using regression to predict scores, which would equally penalize deviations for either low-quality and high-quality clusters, we seek to predict the difference of scores for any pair of clusters corresponding to the same complex. RRPCC combines these pairwise assessments to form a ranked list of clusters, from higher to lower quality. We apply RRPCC to clusters produced by the automated docking server ClusPro and, depending on the training/validation strategy, we show. improvement by 24%–100% in ranking acceptable or better quality clusters first, and by 15%–100% in ranking medium or better quality clusters first. Third, Distributionally Robust Multi-Output Regression Ranking (DRMRR) is a listwise LTR model that induces robustness into LTR problems using the Distributionally Robust Optimization framework. Contrasting to existing methods, the scoring function of DRMRR was designed as a multivariate mapping from a feature vector to a vector of deviation scores, which captures local context information and cross-document interactions. DRMRR employs ranking metrics (i.e., NDCG) in its output. Particularly, we used the notion of position deviation to define a vector of relevance score instead of a scalar one. We then adopted the DRO framework to minimize a worst-case expected multi-output loss function over a probabilistic ambiguity set that is defined by the Wasserstein metric. We also presented an equivalent convex reformulation of the DRO problem, which is shown to be tighter than the ones proposed by the previous studies. Fourth, Inversion Transformer-based Neural Ranking (ITNR) is a Transformer-based model to predict drug responses using RNAseq gene expression profiles, drug descriptors, and drug fingerprints. It utilizes a Context-Aware-Transformer architecture as its scoring function that ensures the modeling of inter-item dependencies. We also introduced a new loss function using the concept of Inversion and approximate permutation matrices. The accuracy and robustness of these LTR models are verified through three medical applications, namely cluster ranking in protein-protein docking, medical document retrieval, and drug response prediction.

Investigating interactions between proteins and small molecules at an atomic scale is fundamental towards understanding biological processes and designing novel candidates during the pre-clinical stages of drug discovery. By optimizing the methods used to study these interactions in terms of accuracy and computational cost, we can accelerate this aspect of biological research and contribute more readily to therapeutic design. While biological assays and other experimental techniques are invaluable in quantitatively determining in vitro and in vivo inhibition activity, as well as validating computational predictions, there is an inherent benefit in the possible throughput provided by molecular dynamics (MD) simulations and related computational methods. These calculations provide researchers with unparalleled access to large amounts of all-atom sampling of biological systems, including non-physical pathways and other enhanced sampling methods. This dissertation presents research into advancing the application of expanded ensemble and other simulation-based methods of ligand design towards reliable and efficient absolute free energy of binding calculations on the scale of hundreds to thousands of small molecule ligands. This culminates in a combined workflow that allows for an automated approach to the force-field parameterization of custom systems, simulation preparation, optimization of the restraint and sampling protocols, production free energy simulations, and analysis that has facilitated the computation of absolute binding free energy predictions. Specifically highlighted is our ongoing effort to discover novel inhibitors of the main protease (Mpro) of SARS-CoV-2 as well as participation in the SAMPL9 Host-Guest Challenge. / Chemistry

No description available.

No description available.

Through the SARS-CoV-2 pandemic, it has become clear that the development of antivirals is essential for the health and wellbeing of the population. In this study, novel active site protease inhibitors against SARS-CoV-2 were tested for their inhibitory activity against the viral 3-Chymotrypsin like protease through the means of FRET based enzymatic assays. Additionally, Compound 104 targeting the NS2B-NS3 protease was tested against Zika virus through yield reduction assays as a means to assess whether these assays are suitable for the assessment of peptide hybrid compounds in Zika virus.Novel compounds against SARS-CoV-2 were screened and five of the selected six active compounds were found to inhibit the viral protease at a half-maximal inhibitory concentration (IC50) of below 0.075 µM.In Zika virus, the yield reduction assay was assessed and it was found that under the conditions tested, this assay is not suitable for the assessment of peptide hybrid compounds in Zika virus.The active novel compounds against SARS-CoV-2 should be taken for further assessment in cell based assays as the next step of development. Compound 104 should be assessed under different experimental conditions to identify whether different conditions can make this assay suitable for the intended use.

Protein-protein interactions (PPI) represent a challenging target class in contemporary small molecule drug discovery. The difficulty arises because PPI sites are structurally and physicochemically different from conventional drug binding sites. Moreover, we currently lack a good understanding of the druggability of PPI targets: that is, how the structure and properties of a PPI interface site relates to the properties of small molecules that can bind to that site with high affinity. Efforts to achieve potent drug-like small molecule inhibitors of PPI interfaces, involving a wide range targets, historically have largely been unsuccessful, leading to the conclusion that new inhibitor chemotypes are needed to inhibit this class of target. In this thesis, I describe the application of two approaches to identify inhibitors of the PPI interface between Kelch-like ECH associated protein 1 (KEAP1) and Nuclear factor (erythroid-derived 2)-like 2 (Nrf2): (i) screening a library of synthetic macrocycles, and (ii) fragment-based lead discovery. I validate and characterize the hit compounds obtained. In the case of the fragment hits, I investigate what features of the compounds are required for binding to the target (Chapter Two). In parallel, I investigate the structure of the hot spot ensemble at the KEAP1/Nrf2 binding interface using three complementary methods: alanine scanning mutagenesis, fragment screening, and in silico probe mapping using the FTMap algorithm (Chapter Three). This analysis brings insight into the druggability of KEAP1, and advances our understanding of the utility and limitations of those three widely used methods for characterizing the hot spot ensembles at PPI interfaces (Chapter Three). Finally, to gain additional insight into the energetics of KEAP1/Nrf2 binding, I probe the additivity of combinations of alanine mutants (Chapter Four). I use the results to propose a quantitative approach to categorizing the various degrees of additivity that can be observed at PPI interfaces, and discuss the possible structural basis for these behaviors. The model potentially provides a more general framework for understanding the binding energetics at PPI interfaces using combinations of mutations.

Selective molecular recognition between proteins is a fundamental event in biological processes that governs cellular growth, function, survival, and differentiation. The immune system, for example, is a complex network of cellular processes regulated by protein-protein interactions (PPIs) between cells, receptors, and secreted molecules. Generating and maintaining an appropriate immune response and regulation requires coordination across many cell types and components, while dysregulation of these interactions can lead to disease. A major obstacle in small molecule therapy development towards these PPIs is their restriction to small protein-protein interfaces and a well-defined hydrophobic pocket. Most PPIs have large contact surface areas and lack traditional binding pockets making them historically challenging for the development of potent small-molecule modulators. To address this limitation, we utilized two binding-based approaches, a unique peptidomimetic fragment library and high-throughput small-molecule microarrays to design and discover molecules that target three important immunoregulatory PPIs: the DQ8-insulin complex, the KEAP1-Nrf2 complex, and the IL-4/IL-4R receptor complex. Many autoimmune diseases involve the ternary PPI complex between immunogenic peptides presented to T cell receptors through the major histocompatibility complex (MHC). Inhibiting this interaction may provide a therapeutic approach for delaying or preventing disease. To target type 1 diabetes, we developed a unique library consisting of 125 fragment-sized molecules that mimic glutamic acid and tyrosine residues from the immunogenic insulin B:9-23 peptide responsible for CD4+ T cell activation. Screening of our library after generation of the MHC class II protein responsible for insulin B:9-23 presentation, DQ8, has resulted in identification of 15 lead fragment compounds to date. Application of our fragment library towards pharmaceutically validated target for inflammation and neurogenerative diseases, the Kelch like ECH-associated protein 1 (KEAP1) and nuclear factor erythroid 2 like 2 (Nrf2), resulted in a 30% hit rate. These are promising results for the further development of selective compounds to inhibit these interactions. For treating inflammatory diseases, such as asthma or cancer, we report the identification of a first-in-class small molecule inhibitor to the cytokine Interleukin-4 (IL-4). The PPI between IL-4 and its receptor complex (IL-4Rα) contains no conventional binding pockets and binding is driven through clusters of complementary residues. Through the combination of small-molecule microarrays and cell-based assays we identified the lead compound, Nico-52, with micromolar inhibitory potency and micromolar affinity. A library of 60 analogs of Nico-52 was synthesized and preliminary structure activity relationships suggest amenability of the p-fluorophenyl substituent and importance of the diol substituent to retain binding potency. These studies resulted in development of a more potent inhibitor to IL-4 with a p-aniline substituent, which could be developed into a targeting ligand to deliver additional therapeutic payloads to an IL-4 enriched microenvironment. In summary, we have developed a peptidomimetic small molecule fragment library as a toolkit for screening against challenging PPI targets with applications towards type 1 diabetes and developed a first-in-class small molecule inhibitor towards IL-4 with applications towards inflammatory diseases. / 2025-09-21T00:00:00Z

Investigating novel treatment approaches to combat Clostridioides difficile Rusha Pal ABSTRACT Clostridioides difficile is the leading cause of antibiotic-induced diarrhea and colitis in hospitals and communities worldwide. The enteric pathogen, classified to be an "urgent threat" by the United States Center for Disease Control and Prevention (CDC), capitalizes on disrupted intestinal microbiome to establish infection with disease symptoms ranging from mild diarrhea to potentially fatal conditions. Disruption of the intestinal microbiome, caused mostly by antibiotic use, enables C. difficile to colonize and proliferate within the host. Paradoxically, antibiotics are used to treat C. difficile infection. These antibiotics decimate the gut microbial community further, thus priming the gastrointestinal tract to become more prone to recurrence of infection. To tackle this clinical setback, we utilized a combination of traditional and non-traditional drug discovery approaches and identified chemical entities and targeted treatment options effective against this toxin-producing intestinal pathogen. Herein, we exploited the strategy of high-throughput screening to identify leads that harbor anticlostridial activity. Our primary phenotypic screen of FDA-approved drugs and natural product libraries led to the identification of novel molecules that were further characterized for their anticlostridial efficacy both in vitro and in vivo. The most potent scaffolds identified were those of mitomycin C, mithramycin A, aureomycin, NP-003875, NAT13-338148, NAT18-355531, and NAT18-355768. Of these, mithramycin A, aureomycin, and NP-003875 were also found to harbor anti-virulence properties as they inhibited toxin production by the pathogen. Furthermore, natural product NP-003875 could confer protection to 100% of the infected mice from clinical manifestations of the disease in a primary infection model of C. difficile. Our final approach has been to develop targeted therapeutics called peptide nucleic acids (PNAs). PNAs are antisense agents capable of inhibiting gene expression in bacteria. In this study, antisense inhibition of the RNA polymerase  subunit gene (rpoA) of C. difficile was found to be bactericidal for the pathogen and could also inhibit the expression of its virulence factors. Additionally, antisense inhibition of the C. difficile rpoA gene was found to be non-deleterious for the tested commensal microflora strains. Given their intriguing anticlostridial properties, it can be concluded that our research opened exciting possibilities that can be further evaluated to uncover new treatments for CDI. / Doctor of Philosophy / Investigating novel treatment approaches to combat Clostridioides difficile Rusha Pal LAYMAN LANGUAGE ABSTRACT Clostridioides difficile is a prominent human pathogen that can colonize the gut and cause fatal infections. C. difficile is the most common cause of microbial healthcare-associated infection and results in substantial morbidity and mortality. The "most urgent worldwide public health threat" label has been assigned to C. difficile by the United States Centers for Disease Control and Prevention (CDC). There is a pressing need to develop new classes of antibiotics with improved efficacy to treat C. difficile infections (CDI). To address the need for novel strategies to combat the growing problem of CDI, we screened FDA-approved drugs and natural products library in search of novel drugs that possess potent and specific anticlostridial activity. Several promising hits were identified and evaluated successfully both in vitro and in vivo. The most potent and novel hits that displayed exceptional activity were mitomycin C, mithramycin A, aureomycin, NP-003875, NAT13-338148, NAT18-355531, and NAT18-355768. Furthermore, a murine model of C. difficile infection revealed that compound NP-003875 conferred 100% protection to the infected mice from clinical manifestations of the disease. Interestingly, these compounds were non-toxic to the gut microflora and human cells. Our final approach has been to develop non-traditional therapeutics to target specific genes in C. difficile. These novel therapeutics are called peptide nucleic acids (PNA). Herein, we designed a PNA targeting RNA polymerase  subunit gene (rpoA) of C. difficile. The designed PNA could successfully inhibit the growth of the pathogen and expression of its virulence factors. In conclusion, our research opened exciting possibilities that can be further evaluated to uncover new treatments for CDI.