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.

No description available.

The human gut microbiota contributes to host metabolic processes. Diverse microbial metabolic enzymes can affect therapeutic agents, resulting in chemical modifications that alter drug efficacy and toxicology. These interactions may result in ineffective treatments and dose-limiting side effects, as shown by bacterial modifications of the cardiac drug digoxin and chemotherapy drug irinotecan, respectively. Yet, few drug-microbiome interactions have been characterized. Here, a platform is developed to screen for drug-microbiome interactions, validated by the isolation of a gut bacterium capable of inactivating the antineoplastic drug doxorubicin. Two hundred gut strains isolated from a healthy patient fecal sample were cultured in the presence of antibiotic and antineoplastic drugs to enrich for resistance and possible inactivation. Raoultella planticola was identified for its ability to inactivate doxorubicin anaerobically through whole cell and crude lysate assays. This activity was also observed in other Enterobacteriaceae and resulted in doxorubicin inactivation by the removal of its daunosamine sugar, likely mediated by a molybdopterin-dependent enzyme. Other potential drug-microbiome interactions were identified in this screen and can be analyzed further. This platform enables the identification of drug-microbiome interactions that can be used to study drug pharmacology, improve the efficacy of therapeutic treatments, and advance personalized medicine. / Thesis / Bachelor of Science (BSc) / The collection of microbes in the human intestinal tract, referred to as the gut microbiome, can modify therapeutic agents and change the efficacy of drug treatments. Identifying these interactions between drugs and the microbiome will help the study of drug metabolism, provide explanations for treatment failure, and enable more personalized health care. For this project, a platform was developed to isolate gut bacteria from human fecal samples and characterize bacteria that are capable of inactivating various antibiotics and anticancer drugs. Through this platform, the gut bacterium Raoultella planticola was found to inactivate doxorubicin, a commonly used anticancer drug. These results suggest that doxorubicin may be inactivated in the gut and demonstrates how this platform can be used to identify drug-microbiome interactions.

The world of precision medicine was revolutionized by the discovery of CRISPR-Cas systems. In particular, the capabilities of the programmable nuclease Cas9 and its derivatives have unlocked a world in which applied genome engineering to cure human disease is a reality being pursued in patient clinical trials. Gene editing via the induction of programmable, site-specific double strand breaks (DSBs) has been revolutionary for the precision medicine field. However, there are many safety concerns centered on the induction of DSBs causing potential undesirable on- and off-target consequences, particularly for in vivo CRISPR applications. To circumvent these warranted concerns, many groups have attempted to repurpose recombinases or engineer new fusion systems to perform programmable genome engineering without the induction of DSBs. This dissertation will first highlight the development of recombinases for programmable DNA insertions over the course of decades, including efforts to evolve novel DNA recognition sequences, efforts to tether recombinases to programmable DNA-binding proteins, and the recent discovery of naturally occurring RNA-guided DNA transposition systems. This dissertation will then highlight the development of CRISPR-associated transposases (CASTs) as DSB-independent programmable mammalian gene editing tools capable of integrating large DNA cargos, as well as the future directions that may further enhance CAST activity in human cells. The works in this dissertation detail the initial efforts to engineer and optimize a new class of genome manipulation tools that were previously absent from the gene editing toolkit.

This dissertation examines the challenges and opportunities of analyzing distinct sources of mental health data in the age of precision medicine and big data. The focus lies on two areas: leveraging real-time Ecological Momentary Assessment (EMA) data to understand individual-level variations in mental disorders, especially depression; and the integration of data from randomized clinical trials (RCTs) to assess treatment efficacy, with an application to schizophrenia. The multifaceted and heterogeneous nature of mental disorders calls for nuanced and personalized assessment methods. In the first part of this dissertation, through our proposed machine learning method, the Heterogeneous-Dynamics Restricted Boltzmann Machine (HDRBM), we examine symptom-level variations beyond the traditional one-size-fits-all summary scores and learn the heterogeneous group dynamics. We demonstrate the effectiveness of our approach on simulated and real-world EMA data sets. We show that by incorporating covariates, HDRBM can improve accuracy and interpretability, explore the underlying drivers of the group dynamics of participants, and serve as a generative model for EMA studies. In the second part of the dissertation, we present the challenges of integrating multiple randomized clinical trials (RCTs) in mental health research, proposing data fusion as a means to integrate individual patient data across similar studies to enhance statistical power. The dissertation introduces novel estimators tailored for multi-study, multi-outcome fused datasets, aiming for the optimization of health outcomes for each treatment. The method also addresses the utilization of similar trials with different outcome follow- up measurements, serving as proxies for unobserved outcomes. An application is provided on cognitive remediation (CR) therapy’s efficacy using the NIMH Database of Cognitive Training and Remediation Studies (DCTRS) as a resource, emphasizing the importance of leveraging surrogate outcomes in clinical trials.

No description available.

No description available.

No description available.

A major physiological risk factor of temporomandibular disorders (TMD) is sensitization of peripheral and central nervous system pain processing pathways. Calcium channel, voltage-dependent, alpha-2/delta subunit-1 (CACNA2D1) has a crucial role in relaying nociceptive information in the spinal dorsal horn. Up-regulation of CACNA2D1 results in abnormal excitatory synapse formation and enhanced presynaptic excitatory neurotransmitter release. Blocking CACNA2D1 with gabapentinoid-class drugs relieves orofacial hypersensitivity. Drs. Foley, Horton, and Sciote previously reported that in a small sample group (n=12), CACNA2D1 expression was greater in males than females, but increased in women with TMD. The objectives of this study are to corroborate these data and investigate expression patterns of other ion channel and conducting system genes. Additionally, since the null polymorphism ACTN3-577XX associates with muscle fiber microdamage during eccentric contraction, we tested for possible gene associations with ACTN3-R577XX genotypes. Masseter muscle samples came from human subjects (n=23 male; 48 female) with malocclusions undergoing orthognathic surgery. This population had skeletal disharmony of the jaws and thus was prone to eccentric contraction. Three males and eighteen females were diagnosed with localized masticatory myalgia. Muscle total RNA was isolated and CACNA2D1, CACNA1S, GABARAP, and TRPM7 expression was quantified using RT-PCR. Expression of these genes were compared based on TMD status and various characteristics that may influence TMD including: sex, age, facial symmetry, sagittal dimension, vertical dimension, ACTN3-577 genotype and fiber type. CACNA2D1 expression differed significantly between sexes, overall (p<0.02), and without TMD (p=0.001). Women with (n=13) and without (n=23) TMD differed significantly (p<0.03). CACNA2D1 expression was also significantly higher (p=0.031) in subjects below age 25. Similarly, GABARAP expression was significantly higher (p=0.001) for patients younger than 25 and for patients less than or equal to age 18 (p=0.013). Otherwise, CACNA1S, TRPM7 and GABARAP differences were not significant. GABARAP expression differed, but not significantly by sex and for the ACTN3-577XX-null genotype. In a population of malocclusion patients, masseter muscle CACNA2D1 expression is significantly higher than CACNA1S, TRPM7, and GABARAP. CACNA2D1 expression is greater in males than females without TMD. However, CACNA2D1 expression increases significantly in females with TMD-associated myalgia. This may support evidence for calcium channel regulation of nociception differences seen between sexes in TMD. It was also found that expression of CACNA2D1 and GABARAP is significantly higher in younger subjects. Additionally, observations presented here suggest potential influence of ACTN3-null condition on function of GABARAP. / Oral Biology

Personalized medicine is the idea that every patient can be treated in a unique manner, tailored specifically to his or her individual needs. Traditionally the field of personalized medicine has focused on using genetic information to determine medical treatment. However, humans are not only the sum of their genetic parts. All people exist within the context of their environment, their experiences, and their relationships. While the connection between this greater context and medical treatment may not be immediately obvious, it exists. If we are to truly tailor medical care, it must occur in a holistic manner, combining both genetics and social context. A thorough understanding of the way that they interact, as well as the individual limitations of both, is the best way to offer individualized care to all patients. / Urban Bioethics