Search

Squamous cell carcinoma of the bronchus is the second most common and fatal subtype of lung cancer. In the process of squamous carcinogenesis, the normal bronchial epithelium undergoes a series of histologic transformations known as the metaplasia-dysplasia-carcinoma sequence. These intermediate histologic patterns are called premalignant lesions, and occur prior to the development of cancer. Compared to early stage cancer, survival following resection of premalignant lesions approaches 100%, highlighting the promise of lung cancer interception. However, because of our lack of understanding of the molecular events during squamous carcinogenesis, we are currently unable to predict which lesions will progress to cancer, and we do not have molecular targets for noninvasive treatment. The work in this thesis seeks to improve our understanding of the changes associated with grades of premalignant histology and progression at the level of single cells. I analyzed single cell RNA sequencing data from a cohort of 41 lesions from 26 patients, encompassing the normal-appearing bronchus, premalignant lesions, and early stage carcinoma. I described histology-associated changes in basal cells. Basal cells from low grade lesions expressed genes related to the maintenance of the normal epithelium, while basal cells from high grade lesions expressed genes related to the cell cycle and detoxification of the airway from smoking toxicants. Secondly, I identified a high grade lesion undergoing the epithelial-to-mesenchymal transition. These cells transitioned from a high grade basal cell state, lost their expression of basal cell markers, and expressed canonical EMT genes, including SPARC, FN1, and MMP2. Finally, I identified shifts in T cell subtypes and widespread expression of exhaustion markers PD-1, CTLA4, LAG3, and TIGIT co-occurring with high grade basal cells. Secondly, I leveraged our single cell data to identify gene modules associated with histology and progression in bulk RNA sequencing data. I identified a module of genes expressed in B and dendritic cells involved in antigen presentation through the MHC II pathway whose expression was decreased in progressive lesions. I also identified a module of stromal-expressed genes that were less expressed in progressive lesions, which had previously been unidentified. Associations between module expression and progression were validated in a second data set. This work improves our understanding of the signaling and interactions between cell types associated with histology and progression of premalignant lesions. These findings may be used to improve our prognostication and treatment of premalignant lesions.

Infectious diseases have impacted human populations throughout history, with increasing frequency and devastation as societies started living in larger groups in closer proximity. Well-known historical plagues are numerous, and their effects on human population size, disease resistance, and immunity can still be observed today. With the discovery of “The Germ Theory of Disease” in the 1800’s, new understanding and subsequent technologies promised a new age free of infectious disease. However, while significant advancements have been made, the long-drawn-out war on existing and novel pathogens continues even to this day with wide-ranging effects. As the COVID-19 pandemic proved, a broad understanding of potential pathogens and population-level effects is still critically important. While infectious diseases have traditionally held significant focus for pathogens’ impact on human health, recent research has unveiled pathogenic connections to a previously unrelated category of illnesses: non-communicable diseases (NCDs). When an infectious agent plays a causal role in NCDs, its impact on human populations is amplified. Several examples are known, including certain strains of human papilloma virus and cervical cancer, hepatitis C and different liver diseases, and others. However, many more are suspected or have limited epidemiological evidence. As these diseases cost billions of dollars to economies and significant suffering, innovation in discovering pathogen-NCD pairs, coupled with better public health policies and new vaccine technologies, are critically important. A broader understanding of the role that pathogens play in the etiology of NCDs would fill a critical gap in disease research and human health.This dissertation employs an innovative approach to identify new pathogen-NCDs pairs and provide additional evidence for suspected associations using electronic health records. Additionally, it introduces novel genomic evaluation techniques that can help provide additional evidence for associations in silico. Chapter 1 is a review of pathogens, infectious diseases versus NCDs, trends in causality of mortality through time, and historical examples of successful strategies combating pathogens that cause NCDs. Chapter 2 covers the main survey analyses that were completed, wherein association was estimated between multiple NCDs and multiple pathogens simultaneously. Data extracted from The UK Biobank were first assayed and significant results were validated in an independent dataset from TriNetX. Chapter 3 uses genomics data to identify additional evidence via two orthogonal analyses for two exemplar associations found in Chapter 2. The first orthogonal analysis involves looking at the levels of viral gene expression in disease-specific case/healthy control human RNA-seq data, to determine if viral gene expression is higher in cases. In the second orthogonal analysis, genes that are differentially expressed upon pathogen infection are assessed to determine if they have a significant overlap with genomic risk loci identified for the diseases of interest. Chapter 4 reviews the results presented in both Chapters 2 and 3, then focuses on both successful advances in limiting pathogen caused NCDs and highlights some limitations and possible fixes that were encountered throughout the project. Finally, recommendations for future work building on this study are given, which would enable researchers to expand this work into additional pathogens and/or NCDs.

Lens epithelial explants serve as a valuable in vitro model for studying cellular processes related to lens development and differentiation. Despite significant research, key mechanisms underlying lens fiber cell differentiation and related signaling pathways remain unclear. This dissertation aims to address this gap by investigating the roles of key genes, transcription factors, and microRNAs in lens development and fiber cell differentiation through multiple studies involving RNA-sequencing, ATAC-sequencing, and other molecular biology techniques.Chapter 2 focuses on the influence of vitreous humor on lens epithelial explants, revealing that it increases chromatin accessibility and upregulates genes related to lens fiber cell differentiation while downregulating those associated with lens epithelial cells. The study's unbiased analysis indicated that RUNX, SOX, and TEAD transcription factors might drive these gene expression changes, providing a basis for further exploration.Chapter 3 investigates the role of Fgfrs and Pten in lens fiber cell differentiation and immune responses using RNA-sequencing on explants lacking Fgfrs, Pten, or both. The results show that the loss of Fgfr signaling impairs vitreous-induced fiber differentiation and immune responses, while the loss of Pten can partially rescue these effects. Gene set enrichment analysis suggested that PDGFR-signaling might mediate this rescue, which was confirmed with immunohistochemistry showing beta crystallin expression, indicating fiber cell differentiation.Chapter 4 explores the functional roles of specific microRNAs in lens development. A comprehensive analysis of miRNA transcripts revealed that the loss of miR-26 leads to postnatal cataracts and significant changes in gene expression, with abnormal increases in genes related to neural development, inflammation, and epithelial-to-mesenchymal transition. This demonstrates that miR-26 is crucial for normal lens development and cataract prevention.Chapter 5 examines mutations in FOXE3, a key transcription factor in the lens epithelium. Using CRISPR/Cas9 genome editing, the study generated three novel mutant alleles, uncovering varying degrees of lens-related abnormalities, including cataracts and anterior segment dysgenesis. RNA-seq analysis of homozygous mutants revealed a significant downregulation of genes associated with lens fiber cell differentiation and an upregulation of genes linked to neural and retinal differentiation.Together, these chapters highlight critical aspects of lens development and fiber cell differentiation, demonstrating the complex interplay between key signaling pathways, transcription factors, and microRNAs. The findings suggest future research directions, such as exploring the role of PDGFR in fiber cell differentiation and investigating chromatin architecture changes using ATAC-sequencing, to further understand the mechanisms driving lens development and differentiation.

ETV1 is a member of the ETS transcription factor family that has been implicated in the development of the nervous system, electrical signaling in the heart, and cancer. Despite being implicated in normal development and pathological states, little is known about the genes regulated by ETV1. As a transcription factor, ETV1 binds to specific sequences of DNA, and this study sought to identify putative binding sites for ETV1 in the promoters and regulatory units of more than 300 genes identified in an expression screen to be differentially expressed in ETV1 knockout tissue. We found ETV1 the number of binding sites does not correlate with the degree of differential gene expression. Tools developed in this study make it possible to efficiently identify transcription factor binding sites in conserved regulatory units.

In the United States, approximately 9,000 people are diagnosed with germ cell tumors in 2022. Testicular GCTs (TGCTs) are the most common malignancy among men between ages 14 and 40, and the incidence has been increasing since the mid-20th century. GCTs are highly sensitive to a regimen of systemic platinum-based chemotherapy; however, better risk stratification and more refined management guidelines are required to minimize unnecessary interventions and their associated adverse effects. Furthermore, platinum-refractory and resistant disease remains a significant issue in 5-10% of patients, for which treatment options are limited, and survival rate is poor. Given its diagnosis at earlier ages, lethal TGCTs comprise the most life years lost for non-pediatric malignancies. Macroscopic and histopathologic evaluation of primary biopsies remains the foundation for clinical management of patients with TGCTs. TGCTs are classified according to their histological composition as interpreted by hematoxylin and eosin (H&E)-staining. Accurate histological classification of patient biopsies is paramount, as these subtypes differ in clinical and biologic behavior and, therefore, affect treatment and prognosis. However, this task may be complicated by the presence of multiple subtypes within a single tumor, and challenges for manual pathology interpretation of whole slide images to determine these features. Characterization and the clinical significance of TGCT heterogeneity remains poorly understood. Even further, attempts to identify histological features of TGCTs for early clinical distinction of high risk and chemo-resistant tumors have yielded sparse insights and remain heavily debated. Further, compared to other solid tumors in adults, TGCTs have very limited model systems and few somatic mutations, with a relatively low mutation rate ranging from 0.5 to 1 mutation per megabase. Also, recurrent mutations in known driver genes are infrequent and discordant between studies. Despite low point mutational loads, TGCTs are defined by chromosomal instability and extensive copy number changes, including highly recurrent chromosomal arm-level amplifications and reciprocal deletions that is particularly enriched in TGCTs compared to other malignancies; however, the transcriptional and epigenetic programs that not only underlie TGCT pathogenesis, but also distinguish TGCT subtypes remain unknown. In this dissertation, I use computational and deep learning approaches in interpreting H&E-stained whole slide images, single-cell RNA-sequencing, and orthogonal epigenetic profiling data to both investigate transcriptional and epigenetic modes of TGCT subtype pathogenesis and evolution as well as also guide the discovery of histopathologic biomarkers and clinically actionable targets to inform treatment, better risk stratification, and more refined management guidelines.

Gene expression in higher eukaryotic cells involves carefully orchestrated interactions between RNA-binding proteins (RBPs) and RNAs. Understanding the rules that govern the interactions between RBPs and RNAs is a central question in biology, and increasingly, in medicine. This thesis investigates the cellular RNA- RBP interactions in two ways. First, we develop colocalization CLIP, a method that combines CrossLinking and ImmunoPrecipitation (CLIP) with proximity labeling, to explore in-depth the subcellular RNA interactions of the RNA-binding protein HuR. Using this new method, we uncover HuR’s unique binding preferences in the nucleus, cytoplasm, and stress granule, during mock and arsenite stress conditions. Second, using available datasets that describe RBP binding in vitro and in cells, we devise two quantitative metrics to directly compare RBP binding behaviors: inherent specificity and mutational sensitivity. These new metrics provide a new quantitative framework to characterize the binding behaviors of an RBP.

Alzheimer's Disease (AD) is a complex neurodegenerative disorder with a significant genetic component. This dissertation investigated the genetic architecture of AD, focusing on how genetic variants alter transcriptional regulation.The study first explored common variants on the understudied X chromosome. We extended the PrediXcan method, optimized X chromosome gene expression prediction modeling, and identified ARMCX6, a female-specific putative AD causal gene in the brain cortex.Next, we focused on low-frequency autosomal variants. We developed a novel scan test leveraging chromosome topology to detect disease-associated regulatory variants in transcription factor binding sites. An empirical expression level simulator assessed test performance and confirmed the method's reliability and robustness.In conclusion, this dissertation contributes to understanding the role of genetic variants in AD. The findings emphasize considering sex-specific effects when analyzing the X chromosome and rare non-coding variants. The developed methods provide valuable tools for the genetics research community to identify and characterize disease-associated variants.

Alzheimer disease (AD) affects more than 6 million individuals in the US, and accountsfor 60-80% of dementia cases. AD demonstrates mixed pathologies of amyloid beta andtau tangle accumulation, leading to symptoms such as cognitive impairment andmemory loss. Taking advantage of a closed, isolated population, the Midwestern Amish,variants both in protection against AD as well as risk for AD were explored. Thecomparative reduction in incidence of AD within the Amish versus a general Europeanpopulation led to the examination cognitive preservation, rather than AD, as thediagnosis of interest. A genome-wide association study (GWAS) was performed on theAmish to identify protective loci for AD via the GENESIS R Package. PC-AiR and PC-Relatewere used for principal component analysis to assist in correcting for the highly relatedsample. Linkage analyses across the entire genome, once again for cognitivepreservation, were used to supplement and corroborate GWAS findings. These analyseswere initially completed in MERLIN, but certain regions demonstrating repeated16significance were run through MORGAN, an MCMC linkage analysis software. Theseanalyses highlighted one region on chromosome 2p11.2-13.1 that demonstrated aprotective effect against AD. This region was previously associated with risk for AD inthe Midwestern Amish, potentially indicating both protective and risk variants withinthis locus. Further evaluations of genetic loci implicated in AD took advantage of thisunique sample population to examine a potential risk factor. The MGMT locus waspreviously associated with increased risk for AD in the Hutterites, who are anotherclosed, isolated population of European decent. In the Midwestern Amish, one SNP inthe MGMT locus was significantly associated with AD diagnosis. However, this SNP wasa novel result, distinct from those detected in a previous study and not in linkagedisequilibrium with previously associated MGMT SNPs for AD. We argue that this is anew association for the MGMT region, not previously detected, possibly indicatingpopulation-specific variants impacting AD risk. Given these findings, this Amish population demonstrates a unique opportunity for identifying new variants for bothprotection against and risk for AD, and merits continued investigation.

Malaria remains a persistent challenge to human health despite extensive elimination efforts, and its infection mechanism(s) remains inadequately understood. Research has predominantly focused on symptomatic cases, in places such as sub-Saharan Africa, where most of the infections are asymptomatic. In this study, we conducted Genome-Wide Association Studies within the sub-Saharan African population to assess genetic factors associated with asymptomatic infection. Our investigation revealed novel human genetic variants with plausible biological implications. Additionally, we conducted the first Genome-Wide Interaction Study examining the interplay between cardiometabolic variables—Systolic Blood Pressure (SBP), Diastolic Blood Pressure (DBP), and Body Mass Index (BMI)—unveiling loci that interact with these parameters to affect risk of infection. Our findings offer insights into asymptomatic malaria, potentially shedding light on the early stages of disease development and contributing to a deeper understanding of its pathogenesis.

Understanding the viral infection and lifecycle of Severe Fever with Thrombocytopenia Syndrome Virus (SFTSV) and Kaposi’s Sarcoma-Associated Herpesvirus (KSHV) is important in improving disease outcomes and reducing viral prevalence. In our SFTSV study, we discovered specific cytokine profiles associated with the severity of clinical symptoms. We used single cell RNA sequencing (scRNAseq) on patient blood samples to identify a unique expansion of the B cell population in SFTSV-induced fatal cases which indicated that plasma B cells are a primary reservoir of SFSTV replication. These findings present a potential method of reducing the severity of SFTSV infection, especially in aged patients who are more susceptible to adverse outcomes.In our KSHV study, we developed a novel oral 3D infection model and demonstrated that KSHV can only infect exposed basal epithelial cells in the oral epithelia. We used scRNAseq to show that keratinocyte differentiation and cell death pathways were affected by KSHV infection, suggesting that epithelial differentiation could contribute to KSHV reactivation through changes in epigenetic regulation. In addition, we found a unique population of infected cells with limited early lytic gene expression and a unique gene expression profile, which we termed latent-2 cells. These findings demonstrate that our in vitro 3D epithelial ALI culture model should be a valuable tool to further understand oral KSHV infection for the development of future anti-viral therapeutics to block KSHV transmission.

In this study, we present CADRE (CAncer Drug Response nEtwork), a comparatively straightforward and naive computational model for synergy prediction that can outperform state-of-the-art machine learning algorithms across diverse cancer panels. Furthermore, we explore the intrinsic suitability of these data sets for synergy prediction and investigate the generalizability of models trained on these data to more intricate cancer systems, grounded in a comprehensive understanding of the underlying biological mechanisms. It is noteworthy that our models, as well as previously established models, exhibit diminished performance when applied to data sets from more heterogeneous cancer panels, including leukemia, and brain cancer. Additionally, we delve into the unique characteristics of melanoma as a subject of study in comparison to other cancer cohorts with analogous data, offering a plausible biological rationale for this disparity.

The American pika (Ochotona princeps) is an herbivorous mammal that inhabits rocky, mountainous regions across the western United States. Although they share a common ancestor with rabbits and other species in the order Lagomorpha, American pikas have a specialized diet due to an inability to migrate from their narrow habitat range. Gene families are made up of genes similar in sequence and function among species that share a common ancestor. Increases in gene copy number due to random duplication results in gene family expansion, whereas gene deletion results in family contraction. Evolutionary divergence can result in functional and genetic differences in the way that pikas and other lagomorphs digest their food with the help of enzymes. We hypothesized that American pikas have undergone lineage-specific expansions or contractions in gene families encoding enzymes (particularly digestive enzymes), allowing the species to digest available food in their narrow and changing habitat. Using the computational tool, OrthoFinder, protein sequences of the American pika were compared to seven distantly related taxa to identify gene families. CAFE5 analysis identified copy number evolution compared to the most recent common ancestor of all eight species. Functional enrichment analysis with PANTHER showed gene families related to digestive enzymes are significantly expanded in the American pika compared to other species. This indicates that protease digestive enzymes are more highly expressed in the American pika, possibly contributing to their metabolism of plants that inhabit their habitat.

CpG sites occur at an irregularly high frequency at regions of the genome called CpG islands; in normal samples, CpG islands are typically unmethylated. When a number of these CpG islands are instead hypermethylated in a cancer sample the sample is referred to as possessing CpG island methylator phenotype (CIMP). CIMP has been well-defined (and was first reported) in colorectal cancers, but research has also been performed on numerous cancer types such as gliomas, melanomas, and leukemia. The goal of the work presented here is to analyze the differences in gene mutation and expression between samples with CIMP (CIMP+) and samples without CIMP (CIMP-) across multiple cancer types. First we perform a coverage analysis to find the sets of genes and mutations which best correlate with CIMP+ samples while minimizing correlation with CIMP- samples. The cancer types used in these analyses are colorectal cancer (COADREAD), stomach cancer (STAD), and uterine corpus endometrial carcinoma (UCEC). The results from each tumor type are combined to create a synthesized set of mutations and mutated genes which correlate with CIMP+ samples. We then perform differential expression analysis on RNA-Seq data from The Cancer Genome Atlas (TCGA) to determine genes that are differentially expressed in CIMP+ samples when compared to CIMP- samples. We find 890 mutations and mutated genes with strong positive correlation to CIMP, many of which are also differentially expressed between CIMP+ and CIMP- samples. Among these mutated and differentially expressed genes are numerous genes in the MSigDB KRAS down-signaling gene set. This analysis furthers our understanding of CIMP and cancer in general and may give more insight into the differences between the genomic characteristics of CIMP-positive and CIMP-negative cancers.

This study aims to incorporate some advantages of computed tomographic data into the chest X-ray deep lung segmentation paradigm. We do this by training a deep convolutional neural network on chest radiographs (a.k.a. X-rays) with manually drawn ground truth and an identical network on radiographs digitally reconstructed from computed tomographic data with ground truth generated for the given computed tomographic image using an automated morphological 3D lung segmentation algorithm. The resulting twin-network ensemble generates pairs of lung image segmentation labels for chest X-rays: 1) a “traditional” segmentation of the lungs encompassing the apparently low-density tissue and 2) a novel, “full” lung segmentation encompassing an expanded view of the lungs’ positionin a chest X-ray including those regions obscured by the heart, ribs, and viscera, in essence, a 2D projection of any portion of the 3D lung. These networks perform consistently, with mean Intersection-Over-Union scores of > 90% and > 95%, respectively, across five trials.By subjective analysis, the proposed lung segmentation approach shows satisfactoryability to generalize onto genuine check X-ray images. The proposed technique’s high performanceand robustness establish a precedent for applying computed tomographic data to automatic chest X-ray segmentation and present an opportunity to further refine existingcomputer-aided detection and diagnostic tools by considering the full lung.

Injury to contractile organs such as the heart, vasculature, urinary bladder and gut can stimulate a pathological response that results in loss of normal contractility. PDGF and TGF are among the most well studied initiators of the injury response and have been shown to induce aberrant contraction in mechanically active cells of hollow organs including smooth muscle cells (SMC) and fibroblasts. However, the mechanisms driving contractile alterations downstream of PDGF and TGF in SMC and fibroblasts are incompletely understood, limiting therapeutic interventions. To identify potential molecular targets, we have leveraged the analysis of publicly available data, comparing transcriptomic changes in mechanically active cells stimulated with PDGF and TGF and identified a shared molecular profile regulated by MYC and members of the AP-1 transcription factor complex. We also analyzed data sets from SMC and fibroblasts treated in the presence or absence of the MYC inhibitor JQ1. This analysis revealed a unique set of cytoskeleton-associated genes that were sensitive to MYC inhibition. JQ1 was also able to attenuate TGF and PDGF induced changes to the cytoskeleton and contraction of smooth muscle cells and fibroblasts in vitro. These findings identify MYC as a key driver of aberrant cytoskeletal and contractile changes in fibroblasts and SMC, and suggest that JQ1 could be used to restore normal contractile function in hollow organs.

This thesis presents three self-contained chapters: a global detection test in the mixed membership Stochastic Block Model (SBM), a community estimation algorithm in the dynamic degree-corrected mixed membership SBM, and a network-based study of the transcriptional control of cellular signal sensing. Detecting the presence of structured communities is one of the most fundamental problems of statistical network analysis. In Chapter 1, we introduce a degree- and cycle count-based test statistic for global testing in the mixed-membership SBM, a common model for social networks. We derive its asymptotic null distribution and show that it is optimal for all choices of model parameters. Studying the evolving structure of complex dynamic networks is becoming increasingly popular. In Chapter 2, we propose a spectral algorithm for dynamic node embedding and mixed membership estimation. We establish explicit error rates under smoothness assumptions on the temporal evolution of mixed memberships and potentially severe degree heterogeneity, showing that our method is rate optimal across a broad parameter range. We showcase its effectiveness on a trade network and a human contact network. Network methods provide critical insights in many scientific disciplines, such as genomics. In Chapter 3, we report evidence of an important layer of transcriptional control of cellular signal sensing. Taking adhesion receptors as an example, we apply diverse network and statistical approaches to characterize the link between chromatin organization in the cell nucleus, gene co-regulation, and receptor proteins clustering.

DNA mutations accumulate in somatic cells throughout our lifetime, making the tissues in our body patchwork of mutant clones that evolve over time. This phenomenon, referred to as “somatic evolution”, underlies the initiation and progression of cancer as well as other age-related diseases. Limitations in existing analytical tools have largely constrained our understanding of somatic genome evolution in cancer and normal tissues. Although recent advances in single-cell DNA sequencing have enabled the characterization of genomic variations at the level of individual cells, they do not produce phenotypic measurements needed for assessing variant impact. Detecting genetic aberrations from single-cell RNA sequencing (scRNA-seq) data has the potential to overcome this limitation by coupling genetic and transcriptomic measurements within the same cell. In the first part of this dissertation, I describe a novel computational method, Numbat, that utilizes population-based haplotype phasing to achieve reliable detection of chromosomal alterations and clonal phylogeny from scRNA-seq data. Analysis of tumor samples from a wide range of cancer types shows that Numbat can accurately reconstruct the tumor copy number profile, precisely identify malignant cells in the tumor microenvironment, and resolve genetic subpopulations with transcriptional signatures relevant to tumor progression and therapy resistance. Although aneuploidy is a defining feature of cancer cells, its extent and consequences in normal tissues remain largely unknown. In the second part of this dissertation, I describe an analysis of 16,672 RNA-seq samples from 948 healthy individuals in the GTEx project using an extension of the Numbat algorithm, representing the first large-scale pan-tissue survey of mosaic chromosomal alterations (mCAs) in the human body. We found that approximately a quarter of healthy individuals carry a clonally-expanded mCA in at least one tissue, with incidence strongly correlated with age. The prevalence and genome-wide patterns of mCAs vary considerably across tissue types, suggesting tissue-specific mutagenic exposure and selection pressures. The mCA landscapes in normal adrenal and pituitary glands resemble those in tumors arising from these tissues, whereas the same is not true for the esophagus and skin. Finally, scRNA-seq analysis of a normal adult adrenal gland harboring mCAs revealed gene expression changes related to cellular proliferation and hormone secretion. Together, these findings show a widespread age-dependent emergence of mCAs across normal human tissues with intricate connections to tumorigenesis.

Treatment combinations are used broadly in clinical settings and are used to treat a variety of diseases and disorders. Despite the potential power of combination therapies to improve the lives of more patients, the complete potential of the space remains untapped. High-throughput screening methods allow for many experiments to be conducted in parallel, often testing hundreds of thousands or millions of single compounds. These approaches also enable researchers to test large volumes of compound combinations, but profiling many compound combinations is extremely resource demanding. This need has motivated many studies to computationally predict synergistic combinations using inputs such as compound targets, implicated pathways, or single-treatment transcriptional responses. Though some of these models have been successful, they are often generated with datasets intended for other applications, leaving ample room for improvement. To our knowledge there is currently no systemic dataset of transcriptional response intended to investigate the mechanisms of compound synergy. A dataset of this type would allow for an improvement of synergy prediction from phenotypic readouts in terms of model validation as well as the development of entirely novel models that may be retroactively applied to currently existing datasets. The Connectivity Map, and its underlying assay, the L1000 gene expression assay, is a powerful resource for high throughput small molecule screening and hypothesis generation (Subramanian et al., 2017). Though the assay has been used to profile single compounds extensively, there has yet been no large-scale effort to generate data using compound combinations. The L1000 platform relies strongly on comparisons to previously generated reference signatures (in L1000 and in other experimental contexts) and there is currently insufficient combinatorial reference data. We hope to enable future researchers to make use of those reference data generated to interrogate novel combinations.to predict synergies in novel disease areas. To address these shortcomings in both the synergistic research space as well as the L1000 platform reference dataset, we developed a framework to combine a cell-viability readout with an L1000-based gene expression readout on a diverse set of compound combinations. This allows us to better understand the molecular mechanisms of combinations shown previously to synergistically kill cancer cells. We have also developed a pilot workflow to carefully calibrate future experiments conducted with compound combinations, as well as identified several potential pitfalls to be avoided by future researchers generating combinatorial data on the L1000 platform. These data may subsequently be used to predict additional synergistic combinations and serve as a dataset to inform future experiments to be conducted on L1000 and other high-throughput screening platforms.

Protein folding and assembly is aided by molecular chaperones that prevent aggregation of unfolded or partially folded intermediates and guide them to their native folded state. Chaperones engage clients through cycles of binding and release from aggregation-prone, non-native regions. ATP-dependent chaperones utilize ATP hydrolysis to drive chaperone conformational changes between low and high affinity states, while ATP-independent chaperones are regulated by diverse mechanisms. Zinc-finger protein 1 (Zpr1) is an essential ATP-independent chaperone dedicated to the biogenesis of eukaryotic translation elongation factor 1A (eEF1A), a highly abundant GTP-binding protein. How Zpr1-mediated folding is regulated to ensure rapid Zpr1 recycling remains an unanswered question. We identified the highly conserved altered inheritance of mitochondria 29 (Aim29) as an eEF1A biogenesis factor. Structural modeling using AlphaFold-Multimer suggested that Aim29 senses the GTP-bound conformation of eEF1A folding intermediates bound to Zpr1. We validated this prediction using yeast genetics, cell biological and biochemical reconstitution approaches, and uncovered that Aim29 sensing of GTP-bound eEF1A coupled to a GTP hydrolysis event facilitates eEF1A exit from the folding cycle and allows for Zpr1 recycling. Our work reveals that a bespoke ATP-independent chaperone system has mechanistic similarity to ATPase chaperones, but unexpectedly relies on client GTP hydrolysis to regulate the chaperone-client interaction. Next, we attempted to use AlphaFold-Multimer to identify additional chaperones or biogenesis factors that haven’t yet been uncovered. We optimized an Alphafold-based pipeline for screening a protein of interest against the entire yeast proteome to identify high-confidence interactors. The success of this screening approach is highlighted by two examples. First, we identified the previously uncharacterized but conserved eukaryotic protein Ypl225w as an eEF1A chaperone candidate and subsequent work by another graduate student in the lab revealed that Ypl225w was a ribosome-associating chaperone that mediates GTP-driven vectorial folding of nascent eEF1A. We also applied this pipeline to an essential eukaryotic GTPase of unknown function, Drosophila melanogaster Misato-Like protein (Dml1). The top interactors for Dml1 were subunits of the chaperonin-containing T-complex (CCT), which we validated experimentally. Acute depletion of Dml1 lead to a decrease in levels of assembled CCT and accumulation of monomers. This observation, along with structural modeling and other preliminary results suggest that Dml1 could be an assembly chaperone for CCT.

Host responses to synthetic phenethylamines have been linked with alterations in the gut microbial profile. A novel gut commensal, Proteus mirabilis (P. mirabilis) strain Roy1, was discovered to be enriched in the cecal contents of a rat treated with the synthetic phenethylamine, 3,4-methylenedioxymethamphetamine (MDMA). The genomic DNA of this bacterium was subjected to Illumina sequencing, revealing a draft genome sequence of approximately 3.9 Mbps and comprises 3,602 coding sequences (CDSs). To identify functionally significant genes, a comparative genome analysis was conducted between P. mirabilis Roy1 and thirty different P. mirabilis genomes available in the NCBI database. The pan-genome comparisons indicated that P.mirabilis Roy1 shared approximately 75% of its core genome with the P. mirabilis referencestrains. Furthermore, the comparative analysis revealed unique genes contributing to its successful colonization and proliferation in the gastro-intestinal tract (GIT). Rats subjected to MDMA treatment exhibited diarrheal symptoms. The strain-specific genes detected in P. mirabilis Roy1 might have aided in bypassing the intestinal emptying process. The unique gene repertoire includes genes involved in inter/intraspecies competition, intestinal adherence, and pathogenicity. Interestingly, these gene sets closely resembled genes found in various Escherichia coli (E. coli) strains, suggesting that they may have been acquired from the rat’s gut contents. The direct contributions of P. mirabilis Roy1 in driving phenethylamine-induced responses require further experimental validations. Phage therapy has emerged as a treatment option for bacterial infections. However, its efficacy against drug-induced responses has not been addressed. Therefore, phages specific to P. mirabilis strain Roy1 were isolated. For orally administered phages to be therapeutically effective, they need to endure the passage through gastric fluid, including surviving at low pH levels. Thus, acidic pH-resistant phages specific to P. mirabilis Roy1 were also determined, results showed three isolated phages exhibited remarkable resilience at pH levels as low as two, indicating them to be valuable tools in this context.Additionally, bidirectional fecal microbial transplantations (FMTs) were also attempted to better understand the link between gut dysbiosis and altered phenethylamine-induced responses. Remarkably, an exchange in drug-induced sensitive and tolerant phenotypes was noted between methylone-treated and non-treated animals, post-FMT completion. The alterations in drug-induced responses were found to be correlated with significant shifts in the relative abundances of Erysipelotrichales, Alphaproteobacteria and Gammaproteobacteria classes before and after FMT completion. Collectively, these findings provide new concepts and insights into the ‘gut-brain axis’ dialog, particularly in the context of synthetic phenethylamine intake.

The information-containing nature of the DNA molecule has been long known and observed. One technique for quantifying the relationships existing within the information contained in DNA sequences is an entity from information theory known as the average mutual information (AMI) profile. This investigation sought to use principally the AMI profile along with a few other metrics to explore the structure of the information contained in DNA sequences. Treating DNA sequences as an information source, several computational methods were employed to model their information structure. Maximum likelihood and maximum a posteriori estimators were used to predict missing bases in DNA sequences. Other novel prediction methods based upon the AMI profile and its ability to evaluate the predictability of DNA bases were also developed and tested for accuracy. The AMI profile was also adjusted to account for the triplet-code nature of DNA sequences. Additionally, machine-learning techniques such as neural networks, support vector machines, and principal component analysis were used to classify different regions of DNA sequences using the AMI profile and to compare coding versus noncoding regions. Finally, the analysis considered the relative frequency of groups of bases (known as k-mers) in DNA sequences. Arithmetic coding was explored as a way to effect the compression of DNA sequences modeled upon the relative frequency of the appearance of k-mers. It was concluded that biological information stored in DNA is complex, yet this investigation provided methods to elucidate some of the character of the information structure of DNA sequences. Advisor: Khalid Sayood

Contemporary and historical changes in marine environmental conditions influence biodiversity drastically. In marine invertebrates with highly dispersive planktonic larvae, diversification is facilitated or constrained by tectonic plate movements altering geography, prevailing current patterns, and climate. The decapod crustacean infraorder Achelata, which contains the spiny (fam. Palinuridae) and slipper (fam. Scyllaridae) lobsters, is characterized by a long-lived phyllosoma larvae stage. Panulirus is the most economically significant and speciose genus of spiny lobsters, with 20 recognized extant species and 5 subspecies distributed throughout shallow tropical and subtropical oceans worldwide. Diversification within these lineages has been proposed to be primarily driven by geological changes causing dispersal and vicariance events that affect population distribution and influence the settlement and dispersal potential of long-lived larvae by oceanic currents. To reveal phylogenetic relationships within the Achelata, maximum likelihood (ML) and Bayesian inference (BI) morphological and molecular phylogenies were constructed. Next, a historical biogeographic scenario for the Achelata was modeled on the time-calibrated molecular ML phylogeny. The Achelata and the Palinuridae emerged in the Tethys during the Triassic epoch. After the ancestor of Panulirus diverged during the late Jurassic (~161 Mya), Panulirus Lineages 1 and 2 originated in the Western and Indo-Pacific during the early (~133 Mya) and late (~85 Mya) Cretaceous. The genus Panulirus experienced the most diversification during the Eocene (~52 Mya), then continued to diversify until the Pliocene (~3 Mya), with the greatest concentration of Panulirus ancestral species existing in the Indo-West Pacific. Ultimately, Panulirus diversification occurred due to global dispersal and vicariance events from the late Jurassic to the Pliocene. As Panulirus species and populations diverged in response to geographic change over time, they colonized habitats with variable environmental conditions and different selective pressures. Regional environmental conditions are known to drive the evolution of morphological, behavioral, and physiological traits in marine crustaceans. Therefore, the Achelata were also expected to respond to selective pressures on a molecular scale. As such, environmental conditions are expected to prompt the adaptive evolution of mitochondrial protein-coding genes (PCGs), which are vital for energy production via the oxidative phosphorylation pathway. I examined the effect of temperature and low oxygen on the adaptive evolution of mitochondrial PCGs. In the Achelata, mitochondrial PCGs exhibited a strong negative selection background. Signatures of positive selective pressure were detected in mitochondrial PCGs in equatorial Panulirus spp. inhabiting tropical equatorial regions with relatively consistent high temperatures and in deepwater Scyllaridae inhabiting low-oxygen depths. These results suggest that, following the initial diversification of Panulirus spp. temperature and oxygen levels have influenced the evolution of the 13 mitochondrial PCGs in equatorial Panulirus spp. and deepwater Scyllaridae. In general, this study has demonstrated how environmental conditions drive diversification as well as the adaptive evolution of mitochondrial PCGs in the Achelata.

Genotyping animals is necessary for various field-based applications that require precise knowledge of the sampled individuals. Though feces are considered a low-quality source of host DNA, molecular techniques are increasingly prioritizing its usage for field-based noninvasive projects. Here, we describe a reproducible workflow to genotype individuals using a whole-genome sequencing approach with the portable, high throughput MinION MK1B and the BWA-GATK variant calling pipeline. After filtering, only 4 of the original 5,394 SNPs passed the filtering criteria, leading to an unsuccessful attempt to generate an informative multiloci SNP panel to confidently and accurately differentiate animals. In the filtered SNPs, 5 samples were entirely void of genotyping data. The majority of SNPs exhibited allelic dropout and a lack of called heterozygote genotypes, leading to the presumable false genotypes of the sampled individuals. On average, approximately 97% of the genome remained unsequenced, with only about one read covering each base in the mapped regions. Despite the limitations of employing a whole-genome sequencing approach to differentiate individuals with the MinION using feces, for species lacking known variants, this strategy may be an effective way to initially identify SNPs for subsequent resequencing and genotyping. Future studies are necessary to validate the authenticity of the identified SNPs and to assess their ability to discriminate individuals effectively with enrichment and targeted sequencing techniques.

Long non-coding RNAs (lncRNAs) play critical roles in diverse biological processes. The extensive availability of public RNA-Seq data offers valuable resources for identifying novel lncRNAs. Here, we introduce LncDC (Long non-coding RNA detection), a machine learning-based tool designed to detect lncRNAs from RNA-Seq data. LncDC utilizes an XGBoost model incorporating features derived from primary sequences, secondary structures, and translated proteins to differentiate between lncRNAs and mRNAs. Notably, sequence and secondary structure k-mer score features, along with various open reading frame-related features, contribute to the classification of lncRNAs and mRNAs. Benchmarking experiments have shown that LncDC surpasses six state-of-the-art tools in several performance metrics. Applying LncDC to 180 RNA-Seq datasets from osteosarcoma patients led to the discovery of 97 novel osteosarcoma-specific lncRNAs. Additionally, the role of lncRNA in Oryza sativa RNase P protein 30 (OsRpp30)-mediated disease resistance in rice remains largely unexplored. OsRpp30 is known as a positive regulator of rice immunity against various pathogens. To further understand this mechanism, we conducted RNA-Seq and small RNA-Seq profiling of lncRNAs, miRNAs, and mRNAs in wild type, OsRpp30 overexpression, and OsRpp30 knockout rice plants. Our comprehensive transcriptome analysis identified 91 differentially expressed lncRNAs, 1671 differentially expressed mRNAs, and 41 differentially expressed miRNAs across these rice lines. We also explored interactions between differentially expressed lncRNAs and mRNAs, uncovering 10 trans- and 27 cis-targeting pairs specific to the OsRpp30 overexpression and knockout conditions. Furthermore, we constructed a competing endogenous RNA network comprising differentially expressed lncRNAs, miRNAs, and mRNAs to elucidate their interactions in rice immunity. Our findings reveal that lncRNAs participate in OsRpp30-mediated disease resistance in rice by regulating genes involved in pathogen recognition, hormone signaling pathways, transcription factor activation, and other critical biological processes related to plant immunity. Overall, the development of the LncDC bioinformatic tool, coupled with our comprehensive analysis of the interactions among lncRNAs, miRNAs, and mRNAs within the context of rice immunity, significantly advances our understanding of the complex roles and functional dynamics of lncRNAs.

The complex interplay of epigenetic factors including chromatin remodeling, histone modifications, and transcription factor binding regulate gene expression and therefore give rise to the diversity of cells within the human body from a common genome. Transcription factors are master regulators of these gene programs and are an attractive lever to manipulate the 1000’s of regulatory elements at the end of a signal cascade with a single perturbation. Here, we screen transcription factor perturbations and characterize their ability to engineer epigenetic regulatory networks in the context of T cell exhaustion or embryonic stem cell differentiation. First, we develop a new in vitro model to study T cell exhaustion suitable for high-throughput screens. We use single-cell genomics to demonstrate our T cells recapitulate the exhaustion seen in human tumor infiltrating lymphocytes functionally, transcriptionally, and epigenetically. Second, we identify the transcription factor Ikaros as a regulator of T cell exhaustion with a CRISPR knockout screen. We determine that knockout of Ikaros is sufficient to prevent the establishment of exhaustion in our model and preserve effector function. We use the clinically approved Ikaros degrader iberdomide to phenocopy our knockout and characterize its genome-wide effects on the transcriptome and epigenome. Then, we determine its mechanism of action, demonstrating that Ikaros represses enhancers of key effector genes by recruiting nucleosomes and preventing the binding of the activation transcription factor families AP-1, NFAT, ETS, and NF-kB/REL. Third, we develop a new single-cell, multi-omic screening method termed Perturb SHARE-seq to enable high-throughput perturbations paired with rich readouts. We use this method both for CRISPR knockout screens in hematopoietic differentiation and transcription factor overexpression to guide differentiation of embryonic stem cells. Collectively, our results highlight the utility of transcription factor perturbations and necessity of single cell epigenomics for understanding and manipulating cell state. Our workflow can be generalized to map the epigenetic regulatory networks in a variety of biological contexts and perturb each network member to investigate its role in controlling gene regulation and cell state.

This study aimed to investigate the impact of beta blockers on cardiovascular patients with asthma. We hypothesized that beta blockers, especially non-selective ones, might pose greater health risks. After matching controls and treatment groups from the Cleveland Clinic Foundation (2010-2018) and the National Health and Nutrition Examination Survey (1999-2018) data, our analysis revealed that beta blockers were associated with a reduction in the number of glucocorticoid bursts (OR 0.776, 95% CI [0.733-0.822]), fewer asthma exacerbations (OR 0.796, 95% CI [0.648-0.978]), and less frequent ER visits (OR 0.425, 95% CI [0.298-0.607]). The non-selective ones alone did not yield statistically significant results. However, cardio-selective ones were associated with even fewer health risks, with odds ratios of 0.761 for glucocorticoid bursts, 0.766 for asthma exacerbation and 0.283 for ER visits. Therefore, it is suggested that beta blockers, particularly cardio-selective ones, could be more widely used in patients with both cardiovascular disease and asthma.

Antimicrobial resistance is one of the greatest threats to human health. The β-Lactams (including penicillins, cephalosporins, carbapenems, and monobactams) are safe, effective, widely-used antibiotics. Mimicking an essential peptide bond and inhibiting enzymes involved in peptidoglycan biosynthesis, β-lactams disrupt the cell wall, killing susceptible organisms. Bacteria have developed resistance mechanisms to β-lactams, of which β-lactamases are the most prevalent and problematic. β-Lactamases hydrolyze the namesake, four-membered, cyclic amide β-lactam ring essential to the activity of β-lactam antibiotics, rendering them ineffective. Unfortunately, β-lactamases are constantly evolving in clinics and the environment and substrate expansion means fewer antibiotics are available to treat a given infection. Pseudomonas aeruginosa is a commonly multi-drug resistant (MDR) pathogen and Pseudomonas-derived cephalosporinase (PDC) is a chromosomal β-lactamase and major resistance determinant in the species.Herein, we take a comprehensive and multidisciplinary approach to understanding and overcoming PDC-mediated antibiotic resistance. We begin by proposing a standardized numbering scheme for class C β-lactamases, seeking eliminate the confusion that commonly arises when different research groups refer to the same amino acids using different designations. Next, we microbiologically and biochemically characterize and elucidate the mechanism of two PDC variants associated with oxyimino-cephalosporin resistance: a tyrosine to histidine substitution at amino acid 221 (associated with ceftolozane-tazobactam and ceftazidime-avibactam resistance) and the deletion of threonine 289 and proline 290 (associated with cefepime resistance). These variants modify the Ω-loop and R2-loop (regions bounding opposite sides of the active site and playing a crucial role in substrate specificity), respectively. Both variants lead to substrate expansion through a kcat or k3 driven mechanism associated with enlargement of the active site. We also characterize and examine the mechanisms and active site interactions of experimental and commercial inhibitors. Finally, we use a bioinformatic approach to examine β-lactamase diversity in Acinetobacter baumannii and P. aeruginosa, revealing that a few alleles are extremely widespread, most are rare, and several occur in concerning combinations.Collectively, these results enhance scientific communication, provide insight into mechanisms of β-lactamase-mediated oxyimino-cephalosporin hydrolysis and inhibition, and suggest priority variants for both biochemical characterization and further research and development efforts toward new β-lactam antibiotics and β-lactamase inhibitors.

The clustering of single-cell RNA-sequencing data has been established as an important first step in single-cell gene expression data analysis for scientists to identify cell type based on RNA level expression. This is important because once a cell type has been identified, the phenotype association, as well as the spatiotemporal dynamics of specific cell types, can be characterized, which could lead to identifying cells associated with cancers and other diseases. However, the high-dimensionality of the data poses computational challenges, while drop-outs (genes that are not identified despite being expressed) hamper the reliability of inference. Since established knowledge on transcriptional regulatory networks provide information on the regulatory relationships between genes, we hypothesize that regulatory networks can help remedy missing data, while also reducing dimensionality. To test this hypothesis, we use a previously existing regulatory network, modern clustering methods, and network propagation together to help enhance clustering performance, which enhances accurate identification of cell types.

Cancer and autism spectrum disorder/developmental delay (ASD/DD) are two common clinical phenotypes in individuals with germline PTEN variants (PTEN hamartoma tumor syndrome, PHTS). However, the accurate prediction of these two disparate phenotypes (cancer vs. ASD/DD) for PHTS at the individual level remains unestablished, despite the availability of population level prevalence estimates.Burgeoning studies have shown that genomic and metabolomic factors may act as modifiers of ASD/DD versus cancer in PHTS. We also found that mitochondrial complex II variants occurring in 10% of PHTS individuals modify breast cancer risk and thyroid cancer histology. Therefore, mitochondrial pathways could act as important factors in PHTS phenotype development. As such, in the first project, we investigated the mtDNA landscape extracted from whole genome sequencing data from 498 PHTS individuals, including 164 with ASD/DD (PHTS-onlyASD/DD), 184 with cancer (PHTS-onlyCancer), 132 with neither ASD/DD nor cancer (PHTS-neither), and 18 with both ASD/DD and cancer (PHTS-ASDCancer). We demonstrate that PHTS-onlyASD/DD have significantly higher mtDNA copy number compared to the PHTS-onlyCancer group. Additionally, the PHTS-neither group has significantly higher mtDNA variant burden than the PHTSASDCancer group.1 Relatedly, maintenance of genome integrity is one of the key biological functions of PTEN; however no integrative studies have been conducted to quantify the DNA damage response (DDR) in individuals with PHTS and to relate the cellular phenotype to organismal phenotypes and genotypes. Therefore, in the second project, we used 43 PHTS patient-derived lymphoblastoid cell lines (LCLs) to investigate the associations between DDR and PTEN genotypes and/or the clinical (organismal) phenotypes of ASD/DD vs. cancer. We found that PTEN nonsense variants are associated with inefficient DNA damage repair. In contrast, we observed the fastest DNA damage repairing rates in LCLs from individuals with ASD/DD, irrespective of PTEN genotype. We applied the reactiondiffusion partial differential equation mathematical model for integrative analyses, a tumor cell growth model with a DNA damage term to accurately describe the DDR process in the LCLs. At lower DNA damage level, live cell density is the highest in PHTS-ASD/DD subgroup than in the other phenotypes, while when DNA damage increases, PHTSASD/DD live cell density was lower than in the PHTS-cancer and PHTS-neither subgroups.Collectively, our studies implicate that mtDNA and DDR may act as modifiers of PHTS phenotype manifestation, ASD/DD vs cancer.

In the pursuit of enhancing prognostic methodologies for glomerular diseases, this dissertation introduces a comprehensive approach termed "deep phenotyping," which synergistically integrates artificial intelligence (AI) capabilities with engineered techniques to revolutionize outcome prediction. The primary objective is to harness the potential of AI to perform rigorous quality control on data, identify distinct tissue compartments, and subsequently extract features for downstream analyses that identifies the image biomarkers with strong associations to disease outcomes.The research journey commences by establishing a comprehensive machine learning prognosis pipeline rooted in deep phenotyping principles. Tailored specifically to glomerular diseases and utilizing digital kidney pathology images, this pipeline addresses the need for robust quality control in deep learning processes. It encompasses a multifaceted exploration, encompassing the impact of image compression artifacts on deep learning performance, the revelation of batch effects caused by digital pathology-related artifacts in multi-site repositories, and the introduction of a novel computer-aided quantitative quality control pipeline.Following this foundation, the dissertation advances to the creation and evaluation of intricate deep learning segmentation models. These models are developed to accommodate a range of kidney histologic primitives and stains commonly used in kidney pathology. The refined segmentation toolkit leads to the extraction of features, with a particular focus on peritubular capillaries (PTCs) due to their implications in kidney disease outcomes. This hierarchical feature extraction spans from fundamental attributes to complex aspects of PTCs.The subsequent sections delve into the investigation of the independent prognostic value of cortical PTCs and their interplay with interstitial fibrosis and tubular atrophy (IFTA). Machine learning prognostic models are built upon this PTC/IFTA interplay and evaluated across multi-site datasets, culminating in validation on an independent testing cohort.In summary, this dissertation pioneers the fusion of deep phenotyping with computational pathology, offering a novel avenue for advanced prognostic biomarkers in glomerular diseases. By harnessing AI for quality control, precise segmentation, feature extraction and machine learning prognostic model building, this work promises to yield enhanced insights into disease outcomes and potentially inform targeted therapeutic strategies.

The elastin-like polypeptide (ELP) is a well-studied structural protein that is easily amenable to amino acid (AA) sequence modifications and has the potential to yield a wide variety of uses in bioengineering and biomedical applications. One unique property of ELP is the inclusion of intrinsically disordered domains (IDP) within the structure that allow for its diversity of physical properties. While it is generally understood that amino acid sequence dictates protein folding arrangements, the contributions of specific amino acid sequences to the intrinsic disorder of ELP has yet to be fully resolved. Therefore, identifying the contributions of specific amino acid sequences to the final physical properties of ELP at the sequence level is an important step in understanding the basis of ELP’s unique biophysical properties. The results of this study resolve the role of amino acid sequence and repeat composition on the disordered structure of ELP. Experimentally, an enhanced synthetic ELP[AV-60] was developed, produced, and purified, with novel photocrosslinking and a transition temperature of 37°C. Computationally, the conserved AA sequence of valine-proline (VP) in the endogenous ELN sequence and its major function in the structure of ELP was identified. Finally, we identified a specific AA monomeric unit, VPGXAG with potential for application in IDPs research.

Host-associated microbial communities evolve with their host. They depend on the host for energy substrates and stability and must adapt to host-specific selective pressures. Evidence for how complex communities respond to certain exposures, such as host diet or medication use, is inconsistent across populations (both within and between host species) due to biological (e.g., interindividual variation) and technical (e.g., varying sequencing and quantification methods) complexities, coupled with evolving methods for analysis. Thus, deriving generalizable conclusions for how gut communities respond to chemical perturbations, particularly host diet, remains a challenge. In Chapter 1 of this dissertation, I discuss efforts made to systematically compare the effects of dietary exposure on the healthy human gut microbiome in relation to diverse gut contexts (across diets, health and disease phenotypes, and host species). In Chapter 2, I share a large-scale characterization of the gut microbiomes of canine and feline companion animals, which will contribute to our understanding of how diet interacts with the gut microbiome in different hosts. Additionally, this work was the first to provide community-wide evidence of microbial transmission of commensals between humans and companion animals, for which the health impacts remain unknown. In Chapter 3, themes of the two preceding chapters come together as I explore how the gut microbiome contributes to obesity, a diet-linked phenotype, in the feline gut microbiome.

Gene regulation is driven by the interplay of numerous features including regulatory factors, genome packaging, and the transcriptional machinery. Established approaches to characterize these features on a genomic scale have largely relied on short-read sequencing, providing an averaged, aggregate view across a large population of cells and obfuscating the underlying single molecule dynamics at play. Meanwhile, complementary single-molecule approaches tend to be limited in scope and scale, even as they provide a more granular view. However, the recent development of long-read single-molecule footprinting approaches like Fiber-seq has provided the potential to bridge the genome-scale breadth of sequencing and the single-molecule resolution of microscopy. In this dissertation, I describe how we applied Fiber-seq to characterize two distinct realms of gene regulation—nuclear transcription and mitochondrial genome organization. First, we used Fiber-seq to visualize RNA polymerases within their native chromatin context at single-molecule and near single-nucleotide resolution along up to 30 kb chromatin fibers. We found that RNA Polymerase II (Pol II) pausing destabilized downstream nucleosomes, with frequently paused genes maintaining a short-term memory of these destabilized nucleosomes. Furthermore, we demonstrated pervasive direct coordination between nearby Pol II genes, Pol III genes, and transcribed enhancers. Overall, we illustrated that transcriptioniv initiation mediates both competition and coordination with nucleosomes and nearby transcriptional machinery along individual chromatin fibers. Second, we used Fiber-seq to measure the packaging of individual full-length mtDNAmolecules at nucleotide resolution, and found that, unlike the nuclear genome, human mtDNA largely undergoes all-or-none global compaction. In addition, we showed that the primary nucleoid-associated protein TFAM directly modulates the fraction of inaccessible nucleoids both in vivo and in vitro, acting consistently with a nucleation-and-spreading mechanism to coat and compact mitochondrial nucleoids. Together, these findings revealed the primary architecture of mtDNA packaging and regulation in human cells. In combination with our characterization of nuclear transcription, this work demonstrates the capability of Fiber-seq to capture the complexity of gene regulation in distinct cellular contexts.

Cancer incidence rates in 2024 are estimated to increase, according to the latest statistics from the American Cancer Society, with cancer-related death rates declining overall due to improved intervention and treatment strategies. However, these strategies remain only partially effective for some cancers due to their sub-clonal evolution and associated molecular heterogeneity with added complexities related to socioeconomic status. One such type comprises head and neck squamous cell carcinomas (HNSCCs) that arise from the mucosal epithelium in the upper aerodigestive tract with the oral cavity representing a major subsite that presents with oral squamous cell carcinoma (OSCC) pathology. The five-year survival rate for OSCC is ~66% for localized stages and reduces to ~35% upon locoregional spread and metastasis. OSCC etiology is associated with tobacco usage, excessive alcohol consumption, and the usage of carcinogens-containing substances. Other HNSCC subsites, such as oropharynx and nasopharynx are associated with human papillomavirus (HPV) and Epstein-Barr viruses infections, respectively. The HPV-negative OSCC subtype is highly heterogeneous in its molecular and cellular composition and remains understudied compared to other solid tumor types; hence, there is a need for early detection strategies in addition to effective targeted therapeutics for improving treatment outcomes. With the advances in computational methodologies to study human diseases, multi-omic data modalities are becoming a crucial tool to improve patient care and have already led to remarkable discoveries. In my thesis, bioinformatics, statistical, and machine learning methodologies were leveraged to study the transcriptome-wide changes across different modalities (bulk and single-cell) of pre-cancer oral conditions in humans and of treatment effects on OSCC tumors in murine models. In my first aim, the transcriptional changes of early oral lesions and their progression to OSCC were studied. OSCC arises from oral epithelial dysplasia through a series of clinical and histopathological changes. A subset of OSCC develops from oral leukoplakia, classified as “oral potentially malignant disorders” (OPMD), clinically defined as either localized homogenous leukoplakia (LL), erythroplakia, or proliferative verrucous leukoplakia (PVL), with a malignant transformation ranging between 15% to 90%. The association of leukoplakia with dysplasia is a strong predictor of its capacity to progress to OSCC. In addition, leukoplakia with hyperkeratosis/hyperplasia or hyperkeratosis non-reactive (HkNR) can develop into OSCC. While these lesions have been characterized clinically, a rigorous molecular characterization of these lesions and of the concomitant microbiota are lacking. In this project, I leveraged bulk RNA sequencing to study the molecular profiles of oral lesions at different stages of transformation, and to characterize their defining pathways and host/microbiome interactions. To this end, I used data collected from tissue biopsies isolated from a cohort of 66 human patients harboring distinct histopathology groups consisting of healthy oral mucosa, PMLs comprising HkNR and dysplasia, and OSCC. These samples were profiled using total RNA-sequencing technology, allowing us to study the global transcriptome, including host and microbiome of the oral mucosa from these groups. Our data revealed that PMLs were enriched in gene signatures associated with cellular plasticity, such as partial EMT (p-EMT) phenotypes, and with immune response. Integrated analyses of the host transcriptome and microbiome further highlighted a significant association between differential microbial abundance and PML pathway activity, suggesting a contribution of the oral microbiome to the evolution of PML to OSCC. Collectively, this study revealed molecular processes associated with PML progression that may help early diagnosis and disease interception at an early stage. My second aim was focused on elucidating the mechanisms of action of candidate therapeutics. As the efficacy rate of FDA-approved drugs for HNSCC tumors remains relatively low due to a combination of factors, including tumor heterogeneity, late diagnosis, and drug resistance, the need to develop targeted therapies focusing on selected cancer pathways represents a promising avenue. In previous studies, our laboratories have shown that pharmacological blockade of Wnt/β-catenin/CBP activity with small molecule inhibitors effectively abolished oncogenic cell phenotypes in OSCC. To determine the identity of additional cell subtypes contributing to OSCC progression to advanced disease, we induced oral tumors in immunocompetent mice using a tobacco-associated carcinogen, 4-Nitroquinoline 1-oxide (4-NQO), that causes DNA damage and recapitulates the genomic and histopathological changes observed with the progressive development of OSCC in human oral epithelia. Following 4-NQO treatment, mice were treated with E7386, an orally active small molecule modulator of β-catenin/CBP activity, at two concentrations — 25 mg/kg and 50 mg/kg. We generated a single-cell RNA sequencing dataset with ~50K cells to explore the E7386 treatment-associated effects on OSCC development and to better understand the underlying molecular and cellular changes that might provide novel insights into its mechanisms of action. Our analyses revealed shifts in cellular diversity between treatment groups, with the proportion of epithelial cells decreasing upon treatment — consistent with greatly diminished tumor volumes – while endothelial and fibroblast populations increased compared to the 4NQO control group, providing further evidence for the antitumor activity of E7386. In the immune compartment, we found enrichment of effector T-cell (CD8+ and CD4+) activity and a decrease in disease-associated neutrophil activity in the inhibitor-treated profiles. Additionally, epithelial sub-typing using curated markers of cell identities revealed a decrease in basal cancer stem-like cells (Krt5+, Krt14+) concomitant with an increase in cycling cells (Top2a+, Cdc20+). Finally, we identified a decrease in a stress cell phenotype associated with the AP-1 complex (Jun+, Fos+), and decreased cell plasticity in response to E7386, confirmed by functional validations through immunofluorescence staining experiments. Overall, this study provides cellular characterizations of murine oral tumors and presents further evidence of Wnt/β-catenin/CBP inhibition as a promising therapeutic strategy in OSCC. In summary, my studies have contributed to advancing early detection and treatment strategies and to unveil molecular mechanisms of treatment response in OSCC.

The human microbiome, a complex ecosystem of microorganisms inhabiting various body sites, plays a crucial role in the immune system and overall human health. A comprehensive understanding of the microbiome and its interactions with the host is essential for advancing scientific knowledge and potential therapeutic interventions. This work focuses on two aspects of the microbiome space: taxonomic profiling for the identification of resident microbes, and longitudinal analysis to unravel the dynamics of microbial communities over time. A fundamental step in microbiome analysis is taxonomic profiling: the identification of resident microbes in samples. Numerous tools have been developed to cater to different sequencing types (e.g., 16S versus WGS) and contexts. However, despite significant advances in the profiling field, further work is needed to establish optimal methods for metagenomic classification. To address this gap, we introduce MetaScope, a comprehensive R-based package for accurate microbial composition identification at a strain-level resolution within a sample. We have performed benchmarking against mock microbial communities to validate MetaScope's performance against popular competitors using 16S datasets. Microbial time-series data presents unique challenges, including intricate covariate dependencies and diverse longitudinal study designs. Existing methods often fall short in addressing these challenges, lacking versatility, data type specificity, or the ability to account for the compositional nature of the data. In response, this work introduces LegATo, an open-source suite comprising modeling, visualization, and statistical tools tailored for analyzing microbiome dynamics. LegATo, with its user-friendly interface, accommodates various study structures and incorporates Generalized Estimating Equation (GEE) models, Hotelling’s T-squared tests, and several visualization functions. This toolkit enables researchers to identify microbial taxa affected by perturbations over time, such as the onset of disease or lifestyle changes, and predict their effects on the composition or stability of commensal bacteria. To illustrate the practical application of LegATo, we present two case studies focusing on the nasopharyngeal microbiomes of Zambian infants exposed to HIV and experiencing fatal acute febrile illness. These applications showcase the efficacy of LegATo for unraveling the complex dynamics of microbial communities, providing insight into the impact of specific perturbations on the microbiome. In conclusion, this research contributes to the advancement of microbiome analysis by enhancing taxonomic profiling methodologies and addressing the challenges posed by longitudinal data. The presented tools, MetaScope and LegATo, provide valuable resources for researchers exploring the intricate interactions between the microbiome and host over time, paving the way for a deeper understanding of microbial dynamics and their implications for human health.

Viruses associated with human cancers, known as "tumor viruses," can induce cellular transformation or immortalization, representing a crucial step in cancer initiation. In past decades, the association between viruses and cancer has been a prominent focus in cancer research. Many tumor viruses, such as human papillomavirus (HPV) and hepatitis B virus (HBV), possess the capability to integrate their genomic DNA or RNA into the target host cell, whereas others, like hepatitis C virus (HCV), rarely integrate into the host genome. Recent studies propose that virus integration may introduce additional oncogenic mechanisms. For instance, in cervical and head and neck cancer, HPV integration directly influences cancer-related gene expression, leading to the generation of hybrid viral-host fusion transcripts. Therefore, detecting viral integration sites in the host genome is crucial for further understanding their oncogenic mechanisms in cancer development. HPV is a well-established driver of malignant transformation in various cancers. However, the impact of HPV integration into the human genome remains largely unresolved due to sample size limitations and existing informatics challenges in identifying viral-host breakpoints from low-read-coverage sequencing data, especially in the presence of complex structural variations around fusion points. In response to these challenges, we developed SearcHPV, a novel method using targeted capture sequencing (TCS) to identify and assemble HPV integration sites in the genome. Our analysis of three HPV+ models demonstrated that SearcHPV detected HPV-host integration sites with higher sensitivity and specificity than two other commonly used methods. Additionally, we validated the junction assembly of SearcHPV, aiding in the accurate identification of viral-host junction breakpoint sequences. Our findings indicated that viral integration occurs through diverse DNA repair mechanisms, including microhomology-mediated repair, etc. We expanded our study to 291 head and neck squamous cell carcinoma (HNSCC) patients, employing TCS, RNA-Seq, and nanopore sequencing. We devised a novel approach to locally resolve complex HPV integrations using nanopore sequencing. Using statistical models, we labeled complex structures as "Type2", characterized by multiple integrations clustered with high copy numbers, and less complex integrations as "Type1." We revealed that Type2 events exhibited significantly more non-canonical splicing sites and were more likely to be transcribed, suggesting a complex transcription pattern. Additionally, RNA expression levels of oncogenes around Type2 events were significantly higher than Type1, indicating potential differences in oncogenic mechanism alterations induced by different types of integrations. In a subset of 78 patients with recurrent or metastatic samples, we explored the heterogeneity of HPV integration structures, revealing unique HPV integrations and varying copy numbers in different tumor sites of the same patients. Using nanopore sequencing on one cell line with primary and recurrent samples, we uncovered potential clonal selections of HPV integrations during tumor progression. Our findings emphasize the heterogeneous and complex nature of HPV integration associated with genome rearrangement, potentially contributing to distinct tumorigenic consequences. We broadened our methodology to other viral-associated cancers and investigated 48 Mucoepidermoid carcinoma (MEC) patients with both TCS and RNA-Seq. We detected one patient with HPV integrated into 13 host genes and exhibiting high expression of HPV16 oncogenes E6 and E7. The genetic mechanisms of host genome integration were found to be similar to our previous findings in HNSCC. This study provided insights into the role of HPV in tumorigenesis of MEC.

Clinical decision support systems are computer-based systems developed with the goal of assisting health care providers in arduous clinical tasks or improving decision-making. In routine clinical care, medical practices tend to be dynamic and must account for diversity of data. In this thesis, we focus on developing innovative multimodal and multidomain AI models for clinical decision support, with a focus on applications with limited data availability. We start with a survey chapter followed by three case studies of multimodal/multidomain proof-of-concept Clinical Decision Support (CDS) models that accommodate limited data. Our research seeks to address questions regarding constructing machine-learning-based models that mimic real-world mental models and bridge domain gaps in cases of limited data. In the first chapter, we explore a survey of state-of-the-art methods in multimodal machine learning applied to biomedicine, highlighting how these models address five challenges of multimodal machine learning: representation, fusion, translation, alignment and co-learning. Next, we tackle a case study where we develop a low-parameter model to discriminate pseudoprogression and true progression in glioblastoma using a small sample of MRI images. Then, we develop a clinically-informed privileged learning model which leverages both routine clinical data and privileged information (CBCT and protein serum/saliva tests) to detect Temporomandibular Joint Osteoarthritis (TMJ OA). Finally, we present a case of domain generalization to allow a model trained on one Alcon SN60WF lens to predict post-operative refraction in patients implanted with other lenses in cataract surgery, with an attempt to adapt to other populations and “A-constants” as well. We present these three case studies as examples of informed models that accommodate diverse data types, as real-world clinical practice is intrinsically multimodal and multidomain. We hope these models provide inspiration for additional models outside of the provided use cases and assert that methodologies can be combined and adapted as needed.

End-to-end relation extraction (E2ERE) is a crucial task in natural language processing (NLP) that involves identifying and classifying semantic relationships between entities in text. This thesis compares three paradigms for end-to-end relation extraction (E2ERE) in biomedicine, focusing on rare diseases with discontinuous and nested entities. We evaluate Named Entity Recognition (NER) to Relation Extraction (RE) pipelines, sequence-to-sequence models, and generative pre-trained transformer (GPT) models using the RareDis information extraction dataset. Our findings indicate that pipeline models are the most effective, followed closely by sequence-to-sequence models. GPT models, despite having eight times as many parameters, perform worse than sequence-to-sequence models and significantly lag pipeline models. Our results also hold for a second E2ERE dataset for chemical-protein interactions.

Cancer is a genetic disease. All cancers are caused by and dependent on mutations. The widespread adoption of NGS in cancer research has revolutionized the field by allowing us to catalog the variants within and across cancer types. This knowledge was used to create a new taxonomy of cancers, where many tumors are classified by genetic alterations in addition to tissue of origin. Despite the role of NGS in identifying the mutations behind these new classifications, pre-genomic methods such as ISH and IHC are still commonly used in the clinic to identify relevant genetic alterations in tumors.About 20% of tumors worldwide are caused by viruses such as HPV, Epstein-Barr virus and Hepatitis B and C viruses. The virus may act as a mutagen by causing prolonged inflammation or by integration into the host genome. Viral infection can therefore be considered another type of genetic alteration used for the classification of tumors, and indeed some tumors are now classified by their causal viral infections. Epimutations are a type of genetic alteration affects epigenetic marks such as DNA methylation rather than the DNA sequence. Epimutations are the genetic lesions behind several disorders including cancer predisposition syndromes. Here I present my work developing an app that integrates with the UCSF 500 pipeline to detect known oncogenic viral sequences in DNA samples from tumors (Chapter 2). I also present my work analyzing long-read sequencing and methylation data of a patient with bilateral Wilms tumor to identify the epimutation causing their cancer predisposition syndrome previously missed by other technologies (Chapter 3). My work reflects the utility of adapting state-of-the-art sequencing technologies for diagnostic use in the clinic. H3K27M gliomas have a mutation in the genes encoding histone H3 that interferes with the deposition of the repressive epigenetic mark H3K27me3. This mark is important in regulating cell type development and EMT. In Chapter 4 I present my work using gene expression analysis to identify the role of EMT in the development and oncogenesis of H3K27M gliomas. Aberrant EMT may serve as a key dependency of these tumors.

With advances in mass spectrometry, there is increasing demand for more effective and adaptive approaches to systematically extract biological insights from glycomic and lipidomic data. This thesis explores the application of Markov modeling in understanding two key cellular processes, N-glycosylation and lipid biosynthesis.In Chapter 2, we demonstrate that a Markov model of N-glycosylation network successfully captures intricate glycosyltransferase interactions by reproducing a set of glycoprofiles from glycoengineered CHO cells producing erythropoietin (EPO). We further validate the model parsimony and accuracy by predicting the glycoprofiles of other glycoengineered drugs from the trained models and their wildtype glycoprofiles. The results attest to the model’s ability to learn glycosyltranferase activities and substrate specificities. To increase the impact of this approach, we also develop GlycoMME to allow broader access to this modeling pipeline, presenting a promising direction for rational glycoengineering. Chapter 3 extends the modeling approach to lipid biosynthesis, introducing the Lipid Synthesis Investigative Markov Model (LipidSIM). As a low-parameter, biologically interpretable framework, LipidSIM proves powerful in leveraging the interdependency in lipidomic data and extracts and quantifies perturbations to lipid biosynthesis reactions, generating hypotheses directly testable by transcriptomic data. This method is showcased in 5 different scenarios with 3 different lipidomic datasets, and the results substantiate LipidSIM as a valuable tool for extracting insights from high-dimensional lipidomic data of different types. In Chapter 4, a Taguchi design is used to systematically characterize the impact of 15 media supplements on potential cellular phenotypes for CHO cells, especially N-glycosylation. The analytical pipeline allows disentangling the impact of individual supplements at different concentration levels with minimal numbers of experimental configurations. This approach answers the demand to find more flexible strategies for controlling N-glycosylation beyond genetic engineering. When applied in conjunction with the Taguchi design, our modeling framework has the potential to facilitate rapid customization of media for the growing market of glycoprotein drugs. This thesis encapsulates the innovative applications of probabilistic modeling in accounting for the biological dependency underlying omics data, offering insights into intricate cellular processes and motivating further exploration in actionable modeling of biological systems.

Biomolecular signaling networks enable cells to decode signals from their environment into the appropriate responses needed for survival and maintaining biological function. Using principles derived from these systems, synthetic approaches such as optogenetics were developed to grant users with precise control over cellular activity. Nonetheless, it remains challenging to engineer synthetic signaling systems towards predefined dynamic behaviors. In this work, we present a quantitative framework that integrates multi-objective optimization, dynamic modeling, and biosensor imaging to streamline the design and implementation of synthetic signaling networks in cells. First, we formulated a general signaling model and applied an evolutionary algorithm to derive a network design capable of decoding two distinct frequency modulated stimuli into separate response channels. Based on this design, we implemented an optogenetic circuit to control protein interactions in cells according to the blue light input pattern. To aid this process, dimerization-dependent fluorescent proteins were used to characterize and select optogenetic components with the required dynamic characteristics. Lastly, we applied our frequency decoder system to regulate downstream biological outputs such as gene transcription, antigen presentation, and antigen triggered CAR-T killing in mice. This framework provides an integrative approach for developing synthetic signaling systems that reduces reliance on trial-and-error.

Chromatin encapsulates not only genomic sequence information but also its structural organization. Chromosome folding patterns may juxtapose distal functional elements of DNA in space and are thought to regulate gene expression. Directly visualizing chromosome architecture to study this grammar of chromosome folding has been but a longstanding ambition in the field of epigenetics. An emergent imaging technology called multiplexed DNA-FISH has made it possible to localize the positions of thousands of genomic loci with nanometer precision inside single cells. Given a cell’s ploidy, deriving chromatin conformations from this data initially assumed detecting a fixed number of bright signals and connecting them in sequential genomic order. But during cell cycle chromosomes copies are subject to change; it is also subject to copy number variations such as amplifications or deletions; it is subjected further still to technical noise such as false positives appearing as true signal, or signal dropout. In my thesis, I sought to develop a different framework that accounts for various sources of noise that have gone unappreciated in multiplexed DNA imaging, to show that by embracing the complexity of noise we can in fact uncover new biology previously unseen. First, I describe spatial genome alignment and its derivative polymer fiber karyotyping, methods for accurately resolving chromatin structures in a copy number agnostic fashion. We apply this method to multiplexed DNA-FISH data of mouse embryonic stem cells (mESC) and the mouse cortex. In dividing mESCs, we provide the first reconstructions of tightly intertwined sister chromatids decondensed in interphase – convoluted structures which previously could not be parsed by computer vision or by human eye. We go on to uncover unusual structures such as replicated homologs interacting in one single chromosome territory, suggestive of mitotic crossover. In the mouse cortex, we uncover tightly paired chromosomes in non-dividing neurons suggestive of sister chromatids in replication. Second, I summarize the field of spatial multi-omics, describing where spatial DNA imaging fits within the technological landscape, as well as describing other emerging technologies. I illustrate the principles of two main methods for achieving spatial resolution: imaging in situ and sequencing-based “release-and-capture” techniques, and further provide a broad survey of different multiplexing strategies for indexing cellular contents. I compare and contrast the two main families of spatial technologies, delineating how the distinct advantages of each lends them to different experimental applications. Given the new capabilities we and others uncovered using spatial multi-omics, such as detecting copy number variations and visualizing histone modifications, that extend beyond transcriptomic profiling, I provide a future perspective on how spatial multi-mics and the many channels of information it captures could inform future clinical applications.

Enhancers are genomic elements that activate time- and tissue-specific gene expression through binding of transcription factors (TFs). Enhancers are therefore critical for successful development and homeostasis. Although the majority of disease-associated genetic variants reside within predicted enhancers, we still have little understanding of how enhancer sequence encodes its function, or the types of changes to enhancer sequence that alter enhancer activity and contribute to phenotypes. We and others have previously shown that suboptimal TF binding site affinity within enhancers are important for maintaining transcription levels and tissue-specificity. Here, we find that single nucleotide variants (SNVs) that increase binding site affinity cause gain-of-function (GOF) gene expression and developmental defects in Ciona robusta and mice. In humans, optimizing SNVs across hundreds of enhancers for a variety of TFs are enriched for GOF gene expression. Thus, a mechanistic understanding of enhancers, namely the importance of suboptimal affinity sites and variants that violate this principle, can pinpoint causal enhancer variants. Beyond the affinity of binding sites, the organization of sites is another component of functional enhancers. The physical properties of TFs and the DNA to which they bind are thought to constrain the way in which binding sites within a functional enhancer can be organized. The interplay of binding site syntax (the order, orientation and spacing of sites) and the affinity of sites is known as enhancer grammar. To better understand how enhancer sequences encode their function, we conducted an exhaustive screen of all possible organization of the Ciona Otx-a enhancer. The Otx-a enhancer activates expression in neural cells and is regulated by a common developmental logic, a tissue determinant (GATA) and signaling pathway effector (ETS). The majority of the organizations are inert, illustrating the importance of syntax within enhancers. We identify syntax features enriched in active enhancers and use them to predict genomic enhancers. We find that considering both syntax and affinity gives the greatest accuracy of genomic enhancer predictions, highlighting the importance of grammar. We also use grammar to design tissue specific enhancers. Collectively, these studies increase our mechanistic understanding of how enhancer sequences encode their functions, and the types of changes to enhancers that can alter their function and may contribute to novel expression patterns and phenotypes.

Mosquitoes harbor and transmit a variety of pathogens which are dangerous to humans and incur a constant and significant public health cost. Gaining more detailed knowledge about these pathogens, their interactions with the mosquito, and their molecular biology and genetics will allow for new techniques to be developed to prevent harmful effects on humans. These pathogens vary in complexity from viruses like West Nile virus to eukaryotic apicomplexan parasites like Plasmodium falciparum, the human malaria parasite.Culex mosquitoes routinely transfer viruses like West Nile virus in the United States representing a predominant health threat there. These mosquitoes are also routinely infected by a wide variety of other viruses, which have received very little research attention, yet could affect the transmission of pathogens like West Nile virus and St. Louis encephalitis virus. Therefore, in the first chapter of this dissertation work, we performed small RNA sequencing to examine the full virome of field-caught Culex mosquitoes in multiple geographical regions of southern California. These data also allowed us to analyze the interactions between viruses and identify potential pairs of co-infecting or mutually excluding viruses, as well as closely look at the small RNA immune response of Culex mosquitoes against these viruses, expanding the known role of the antiviral piRNA response in Culex. We also identified mosquito miRNAs that may be involved in antiviral immunity or virus infection by comparing highly infected mosquito pools against lowly infected ones. Apart from viruses, single-celled eukaryotic parasites are also transmitted from mosquitoes to humans. One of the deadliest of these pathogens is the malaria parasite, Plasmodium falciparum. Important regulators that propel the life cycle of this parasite also represent potential drug targets, that when identified could lead to new treatment options to lessen the impact of malaria in Africa and across the globe. We performed a wide variety of experiments and studies in this vein. In the second chapter, we examine the role of long non-coding RNAs (lncRNAs) on parasite gene regulation and predict lncRNAs across the genome, categorizing their properties and their essentiality for parasite survival. We also use experimental techniques to determine the binding sites of several of these predicted lncRNAs, and look at the effect of one in particular, lncRNA-14, by knocking it out and using transcriptomic and phenotypic analysis. The third chapter, on the other hand, uses Plasmodium berghei, a mouse malaria parasite, as a model for human malaria and examines parasites proteins SMC2 and SMC4, the homologs of proteins that make up the condensin complex in model eukaryotes. We determine that, as in other eukaryotes, Plasmodium SMC2 and SMC4 form a condensin complex and are key in cell division, particularly in the mosquito stages of the parasite life cycle. ChIP-seq determined that these proteins bind at the centromeres, and transcriptomic and phenotypic analysis revealed the exact roles of these proteins and their importance. Finally, the remaining two chapters examine RNA-dependent and RNA-binding proteins in Plasmodium falciparum, key elements in post-transcriptional regulation, another important aspect of gene regulation in the life cycle. In the fourth chapter, we performed a screen for RNA-dependent proteins using the R-DeeP protocol, obtaining a list of likely proteins and examining RNA-dependent complexes. We also characterized one of the proteins found as an RNA-binding protein using various techniques including enhanced crosslinking and immunoprecipitation followed by high-throughput sequencing (eCLIP-seq). In the fifth chapter, this technique and others were also used to determine the binding sites and functions of two other predicted RNA-binding proteins with RAP (RNA-binding domain abundant in apicomplexans) domains. These RAP proteins were determined to regulate the parasite mitochondrial rRNAs (mitoribosome). All in all, this dissertation work reveals new insights into mosquito-associated pathogens, their interactions with mosquitoes and mosquito immunity, and important molecular components of their life cycles which could be targeted to reduce their impact on humans.

Studying the interactions between cancer and the immune system has led to the development of immunotherapeutics that can cure patients with historically untreatable tumors. However, the mechanisms of action and predictive biomarkers of these therapeutics -- immune checkpoint inhibitors, or ICIs -- are yet to be fully characterized. Discovering these mechanisms requires measuring and contrasting various immune celltypes’ responses to ICIs in different cancer contexts across many potentially relevant pathways, a task for which single cell transcriptomics sequencing (scRNAseq) is particularly well suited. In Chapter 1 we compared the immune response to ICIs of melanoma and basal cell carcinoma (BCC) measured via scRNAseq and found that the ratio of macrophages to memory B cells can predict response to ICIs. In Chapter 2 we compared resistance mechanisms of melanoma and BCC to pancreatic ductal adenocarcinoma (PDAC) which does not respond to ICIs. We determined that PDAC’s resistance to ICIs is explained by low expression of the target proteins. In Chapter 3, we studied why BCC doesn’t induce an immune response. We discovered that most BCCs downregulate antigen presentation genes and therefore avoid immune detection entirely. In Chapter 4, we pivoted to studying feature selection (a step in standard machine learning pipelines whereby uninformative gene expression is removed) in scRNAseq data analyses. We found that feature selection, despite being a step that has not been studied in great depth in the context of scRNAseq analyses, can be a crucial step in correctly stratifying different groups of cells. We then derived an analytical feature selection method that correctly ranks features in terms of their utility in separating different cell states and removes false positive features that would lead to spurious cell states. The findings of this thesis will lead to advances in both immuno-oncology and scRNAseq analyses.

Neurodegeneration is a key feature of several neurological disorders and is characterized by systemic loss of neural structures and functions in the central nervous system. A diverse panel of environmental and heritable factors contribute to the development of neurodegen- erative phenotypes. Alzheimer’s disease, the most prevalent neurodegenerative disorder, is a complex polygenic disease that progressively alters the cellular and molecular makeup of the brain. A comprehensive understanding of these changes are necessary for developing future interventions and therapeutics to reverse the progression of disease or preventing it from ever occurring. In this dissertation, I leveraged several cutting-edge -omics technologies to generate high-quality molecular maps of the cell states that are dysregulated in AD. Fur- thermore, in-depth analysis of these maps have yielded numerous insights into the cellular phase of disease, especially regarding the relationship between genetics and cellular states. First, I used single-nucleus epigenomic and transcriptomic sequencing in postmortem human cortical tissue from late-stage AD donors and cognitively normal controls to deeply charac- terize changes in the gene regulatory landscape throughout disease. Next, I applied spatial transcriptomics to investigate changes with respect to amyloid pathologies in early-stage and late-stage AD, as well as AD in Down Syndrome. These studies led to numerous find- ings implicating a role for oligodendrocytes in neurodegeneration, thus prompting the final study. I leveraged mouse experiments where we used single-nucleus epigenomics and tran- scriptomics to track the molecular changes involved in remyelination, and to understand how the dynamics of oligodendrogenesis change with aging. Altogether, this dissertation provides foundational knowledge of gene-regulatory circuitry in neurodegeneration with cellular and spatial resolution.

Mathematical models can be used to describe how a biological system changes over time and moreover the consequences of a perturbation arising in an abnormal state of that system. Technological advancements in experimentation and growing large-scale datasets drive the need for increased quantitative analysis of biological phenomena. However, a significant effort of data analysis of these datasets prioritizes recapitulating observations and forward predictions without focus on the underlying interpretability of these approaches. Mechanistic models aim to reproduce experimental observations while also providing understanding of the biological processes that generate these observations. In this dissertation, mathematical models were employed to understand how changes in cellular (Chapter 2) and population (Chapter 3) dynamics of hematopoietic cells emerge in different physiological and pathological states. In Chapter 2, we interrogated how macrophage polarization alters the stimulus-response activation dynamics of the transcription factor, NFκB. Statistical approaches were first applied to differentiate NFκB response dynamics across polarization states and define a landscape of macrophage functional states based on the timecourse data of this single analyte. Utilizing this data however, biochemical parameters of the NFκB signaling network, such as rates of receptor synthesis and proteasomal degradation, that define these different molecular network states could also be inferred by employing a mathematical model of this signaling system. In Chapter 3, we examined how inflammatory signals change the steady-state abundances of hematopoietic stem and progenitor cells (HSPC) in settings of myeloid bias. Beyond simply characterizing differences in abundances and gene expression from these data, employing a mathematical model of HSPC population dynamics allowed identification of likely dynamic rate perturbations. This model-aided analysis determined that differentiation bias of hematopoietic stem cells alone is an insufficient explanation for myeloid bias amongst multipotent progenitors and that proliferation of early myeloid-primed progenitors additionally contributes to myeloid bias with NFκB dysregulation, aging, and malignancy. Together these studies demonstrate the utility of mathematical modeling in the interpretation of biological datasets and additional insights could be gained through extension of these models and integration with other analyses approaches.

Cytokine signaling a core mechanism by which immune activity is regulated in both health anddisease. Cytokine-mediated signaling regulates the proliferation, differentiation, and activity of cells in both the innate and adaptive immune systems. Due to their powerful regulatory capacity, cytokines have been leveraged as immunotherapies in a wide range of disease indications; for example, interleukin-2 (IL-2) has been explored as a potential immunostimulant for the treatment for cancer, as well as an immunosuppressant for the treatment of autoimmune diseases. However, in many such cases, the pleiotropic nature of cytokine signaling has stymied the development of efficacious and safe therapies due to the induction of signaling in off-target populations. To overcome this limitation and bias cytokines towards signaling in target populations, engineered cytokines with a variety of alterations, such as mutations affecting their binding interactions with their cognate receptors, fusion to antibody fragments, or co-formulation with antibodies to that cytokine have been developed. However, without a quantitative model of signaling the effects of such mutations and alterations are often difficult to anticipate, leading to inefficient cytokine engineering efforts. To address this lack of quantitative understanding, we conducted a battery of computational studies. First, using a mechanistic binding model, we developed a general, quantitative understanding of the landscape of cell-selective cytokine signaling, and found that affinity, valency, and multi-specificity must be simultaneously optimized to engineer optimally selective cytokines. We then specifically studied the IL-2 signaling pathway, and used both ordinary differential equation models and our mechanistic binding model to study the signaling characteristics of wild-type and engineered IL-2 mutants. Leveraging our newfound quantitative of how affinity and valency interact to determine a cytokine’s selectivity profile both generally and in the specific context of IL-2, we developed affinity-optimized tetravalent IL-2 mutants with superior regulatory cell selectivity. Using these models of IL-2 signaling, we also elucidated the mechanism by which engineered antibody-IL-2 fusions induced regulatory cell-selective signaling and conferred protection against autoimmunity. In total, this body of work demonstrates the critical role that computational modeling plays in potentiating the engineering of superior cytokine-based immunotherapies.

Indole-3-acetic acid (IAA) is a naturally occurring heterocyclic compound that is best known for its role as a growth hormone (auxin) in plants but is produced by many other organisms from across the tree of life. This dissertation presents insights that I uncovered from the genomic analysis of the indole-3-acetic acid catabolic (iac) gene cluster in bacteria. This gene cluster confers the ability to enzymatically convert IAA to catechol, which if further degraded and channeled into central metabolism means that IAA can serve as a sole source of carbon, nitrogen, and energy. First identified and characterized by Leveau and Gerards in 2008 in a bacterial isolate from the pear phyllosphere, strain Pseudomonas putida 1290, the iac gene cluster was subsequently described and studied in other bacteria, including Enterobacter soli LF7, Acinetobacter baumannii ATCC 19606, Paraburkholderia phytofirmans PsJN, Caballeronia glathei DSM50014, and Pseudomonas composti LY1. I collectively refer to these strains and P. putida 1290 as the iac model strains.Chapter 1, and its addendum, summarize the published literature in regards to the iac cluster in P. putida 1290 and other iac model strains. I discuss the contribution of individual iac-encoded proteins and enzymes to the multi-step conversion of IAA to catechol. I examine the IAA-degrading model strains in the context of the environments from which they were originally isolated. I elaborate on scenarios in which iac genes may benefit bacteria, such as the exploitation of IAA as a source of carbon, nitrogen, and energy and the interference with IAA-dependent processes and functions in plants and other organisms. I also compare and contrast the iac clusters with each other in terms of gene synteny and with two other bacterial gene clusters (designated iaa and iad) as alternative bacterial solutions to the catabolism of IAA.Chapter 2 is an in-depth comparative genomics analysis aimed at cataloging the diversity of iac-harboring genomes in the bacterial tree of life. I identified iac gene clusters in 4% (4,620 out of 115,091) of the bacterial RefSeq genomes within the Genome Taxonomy Database (GTDB). These genomes belonged to Alphaproteobacteria, Gammaproteobacteria, and Actinobacteria. I also screened metagenomes in the IMG database, which revealed the presence of iac genes clusters in a diverse array of environments including not only plants, but also soils/sediments, surface water, humans, and animals. By analyzing the distribution and prevalence of iac gene clusters across iv the bacterial tree of life along with non-metric multidimensional scaling analysis of IacD protein sequences, I deduced patterns that suggest both vertical inheritance and horizontal transfer of iac genes. Certain functions, such as the catabolism of catechol, benzoic acid, phenylacetic acid, and malonic acid were found to be enriched in genomes with iac gene clusters, hinting at co-selection of these functions with the iac-encoded conversion of IAA into catechol.Chapter 3 builds on findings from Chapter 2 and further explores the evolutionary origins of the iac genes by using a large and diverse subset of 94 iac-harboring genomes identified in Chapter 2. By applying a pangenomics approach to the iac gene neighborhoods in these genomes, I was able to identify a core set of iac genes (iacA, iacB, iacC, iacD, iacE, and iacI). A concatenated phylogeny of the proteins encoded by these core genes revealed four distinct clades of iac genes that corresponded well with the iac gene cluster synteny in members from these clades. I also compared individual iac proteins with their broader protein families, revealing possible shared ancestors between iac- and iad-encoded proteins, IacA and an ethionamide sulfoxygenating enzyme (Rv3094c) from Mycobacterium tuberculosis, as well as IacH and the indole-3-acetamide hydrolase (AMI1) proteins in plants. I also used structural phylogenetics to discover that the IacB protein belongs to the Dimeric alpha-beta barrel (Dabb) protein superfamily. Furthermore, using ancestral state reconstruction in conjunction with the creation of time trees, I was able to infer that the iac gene cluster most likely diverged from its closest ancestral relative around the time that plants moved to the land. Chapter 4 uncovers the genome of P. putida 1290, which was the first bacterium for which the genetic underpinnings of IAA catabolism were discovered. Most notably, the iac gene neighborhood of P. putida 1290 exists as part of a genomic island, alongside many other genes that enable the catabolism of plant-derived compounds including methylamine, phenylacetaldehyde, and opines. I initially predicted that a gene encoding a methyl-accepting chemotaxis protein (MCP) located directly upstream of the iac cluster might be responsible for the previously observed chemotaxis of P. putida 1290 towards IAA. However, recent research from another lab proposed a different gene (deemed pcpI) for this function. An addendum to Chapter 4 details my bioinformatic analysis of the MCPs (including PcpI) encoded by P. putida 1290.Chapter 5 delves into a potential application of our biological understanding of the iac gene v cluster in pathogenic Acinetobacter species. I explore the use of the IAA-analog 5-chloroindoleacetic acid (5-Cl-IAA), in combination with IAA, to act as an antibiotic against Acinetobacter lactucae NRRL B-41902. I show that growth inhibition by 5-Cl-IAA is conditional on iac gene expression, dose-dependent, and bacteriostatic, not bactericidal. This foundational research paves the way for further studies and potential treatments, which may help in combating hospital-acquired infections caused by iac-harboring bacteria such as A. baumannii. Finally, Chapter 6 presents a Python package based on scripts originally used in Chapter 2 to identify the iac gene cluster in bacterial genomes from the RefSeq database. An example workflow is presented which outlines the use of this software to identify the phenylacetic acid (paa) gene cluster in Pseudomonas putida genomes. Using known proteins encoded by a particular gene cluster of interest, one can screen the vast array of bacterial genomes in the RefSeq database, setting the stage for performing downstream comparative and phylogenetic analyses as I have done in Chapters 2 and 3.

Studying structural variants (SV) in populations is crucial since they cause more diversity and have more effects on gene function than small variants. However, population scale studies are challenging since finding SV from inexpensive short-read sequencing(SRS) methods have a high false positive rate. Conversely, long-read sequencing(LRS) are more accurate for SV discovery but are expensive at a population scale. Here, I develop new unbiased techniques to study SV in populations that are more scalable than the state-of-the-art. I show their utility in creating an SV catalog for the cattle breed augmented with their allele frequency.