T-cell-based immunotherapies hold promise in treating cancer by leveraging the immune system’s recognition of cancer-specific antigens. However, their efficacy is often limited in tumors with few somatic mutations and significant intratumoral heterogeneity. Here, we introduce a previously uncharacterized class of tumor-wide and public neoantigens originating from RNA-splicing aberrations in diverse cancer types. Notably, we identified T-cell receptor clones capable of recognizing and targeting neoantigens derived from aberrant splicing in GNAS and RPL22. In multi-site-directed biopsies, we detected the tumor-wide expression of the GNAS neojunction within glioma, mesothelioma, prostate cancer, and liver cancer. Importantly, these neoantigens are endogenously generated and presented by tumor cells under physiological conditions and are sufficient to trigger cancer cell eradication by neoantigen-specific CD8+ T-cells. Moreover, our study resolves the complex interplay of dysregulated splicing factor expression in specific cancer subtypes, leading to recurrent patterns of neojunction upregulation. These findings establish a molecular basis for T-cell-based immunotherapies addressing the challenges of intratumoral heterogeneity.

Protein synthesis is the most energy expensive process in the cell and a crucial component of most biological processes. Dysregulation of proper mRNA translation can lead to many human diseases, ranging from neurodegeneration to cancer. Translation initiation is the first step in protein synthesis, during which the small subunit of the ribosome scans the 5’ untranslated region (5’UTR) of an mRNA to identify a start codon and commence translation elongation. By unwinding and modulating secondary structures and other RNA features present in the 5’UTR, RNA helicases can regulate ribosome scanning and start codon selection. This dissertation presents an approach to measure the effect of RNA helicases on mRNA translation initiation. 5’UTR luciferase reporters are transcribed in vitro and employed in either an in vitro assay or an cell-based assay. Either method can be used to investigate how the perturbation of a helicase, such as changes in protein levels or mutations, affects translation initiation at the 5’UTR level. These assays will provide insights on the role of helicases and other translational factors as regulators of the proteome both in physiological and diseased settings.This dissertation also explores the role and function of a specific RNA helicase of the DEAD-box family, DDX3X, which regulates the translation of human transcripts containing complex 5′ untranslated regions. We developed a reporter called HART that uses DDX3X-sensitive 5′UTRs to measure DDX3X-dependent translational activity. Through SHAPE-MaP analysis, we determined the secondary structure of these 5′UTRs and used HART to investigate the features that render them sensitive to DDX3X. Additionally, we investigated the association of DDX3X with the ribosome during scanning and identified residues 38-44 of DDX3X as potential mediators of its interaction with the translational machinery. HART revealed that both DDX3X 38-44ala and the helicase-defective mutant DDX3X R534H exhibit deficiencies in promoting translation on DDX3X-sensitive 5′UTRs. These findings suggest DDX3X plays a crucial role in unwinding RNA structures during translation initiation through its interaction with the translational machinery. We also establish HART as a robust, lentivirally encoded, and adaptable reporter of DDX3X function.

RNA switches are key regulators of gene expression in bacteria, yet very few have been described and characterized in Metazoa. Here, we present an integrative computational and experimental approach to systematically annotate functional RNA structural switches genome-wide. By applying this approach to the human transcriptome, we discovered 245 putative RNA switches. Among these, we further characterized and functionally dissected a previously unknown RNA switch in the 3’UTR of the RORC gene. In vivo DMS-MaPseq complemented by cryogenic electron microscopy confirmed the existence of this element as an ensemble of two alternative functional conformations. We then used a genome-wide CRISRPi screen to identify the trans factors that mediate gene expression through this RNA structural switch. We discovered that the nonsense-mediated mRNA decay pathway acts on this element in a conformation-specific manner. Our findings suggest that RNA structural switches may play an important, yet underappreciated role, in shaping the gene expression landscape of the cell in eukaryotes.RNA-binding proteins (RBPs) are multifunctional regulators of gene expression with complex and context-dependent mechanisms of action. The underlying regulatory grammar that underlies RBP-mediated functions remains largely unexplored. Here, we developed and applied a multi-omic data integration platform to systematically decipher the context-specific functions of RBPs. We used in vivo proximity-dependent biotinylation (BioID) analysis of 50 human RBPs to generate a comprehensive map of major RBP neighborhoods. In parallel, we took advantage of CRISPR-interference with single-cell RNA-seq read-out (CROP-seq) to effectively capture the overall transcriptomic response downstream of each RBP knockdown. By integrating these physical and functional interaction data with the ENCODE transcriptome-wide atlas of RBP binding from eCLIP assays, we generated a map of RBP functional interaction. The resulting map captures well-studied post-transcriptional pathways and reveals novel RPB-mediated regulatory functions. Here, we report the validation of context-specific functions for several RBPs using biochemical and genetic approaches. We showed that TAF15 controls splicing, translation, and stability for distinct and largely independent RNA regulons. Similarly, we demonstrated that ZNF800 and QKI play non-canonical roles in both transcriptional and post-transcriptional regulation. Taken together, our integrative map reveals that each RBP may participate in multiple post-transcriptional regulatory modules, each with their own target regulon and regulatory function. Deciphering these complex modes of function and interaction are the first step towards a better understanding of post-transcriptional control in health and disease.

Patients differ in their response to clinical interventions. Both genetics and the environment contribute to the phenotype of a patient after a chemical intervention such as a drug. Well-validated interventions have large population effect sizes, meaning they are valuable for an individual on average. However, for complex conditions such as autoimmune disease and cancer, the precise outcomes of chemical interventions remain difficult to predict for an individual patient. Here we present methods for predicting chemical response and apply these methods to understand biological mechanisms. We discover a new biomarker associated with psoriasis patient response to the drug ustekinumab and use this biomarker to stratify patient populations at various clinical endpoints. Moving beyond association analyses, we develop a machine learning model that integrates chemical structures and gene expression information to predict cellular responses to hundreds of chemicals, and apply methods to investigate how this model works. Finally, we improve model generalizability by pretraining a new model on massive amounts of gene expression data and then applying it to downstream prediction tasks. This work contributes models that predict biological responses to chemicals and methods to interpret how these models work. We anticipate this work will advance precision medicine by improving the ability to predict how patients respond to drugs. Simultaneously, this work provides an in silico platform for screening new biological models against a diverse set of chemical probes. Lastly, our pretrained model may find broad applications in phenotype prediction tasks ranging from disease risk modeling to drug response prediction.

In recent years, the complexity of chemotherapies and their interactions with the peripheral nervous system have come into focus but limitations in experimental models have remained a significant challenge in the field. As evidence, despite most chemotherapy drugs being around for decades, there are currently no therapies approved that target chemotherapy-peripheral nervous system interactions as an anti-neurotoxic agent. Human pluripotent stem cells offer an appealing model system that, unlike rodent models, are compatible with high throughput, high content applications; techniques that reflect modern drug discovery methodologies. Thus, utilizing the key advantages of stem cell-based models in tandem with the strengths of traditional animal models offers a complementary and interdisciplinary strategy to advance chemotherapy-peripheral nervous system research and drug discovery.

This dissertation begins with an overview of the current status of chemotherapy-peripheral nervous system research, describing examples of taxane chemotherapy-induced damage to the sensory nerves and platin chemotherapy-induced damage to the enteric nerves. Avenues where stem cell-based models may further advance the field are also presented. Based on this foundation, I established a stem cell-based model of chemotherapy-induced enteric neuropathy, focusing my efforts on the platin chemotherapy drug class as they are heavily prescribed and highly neurotoxic to the enteric neurons, which innervates and controls the gastrointestinal tract.

To uncover the mechanism of platin-induced gastrointestinal neurotoxicity, I leveraged my scalable stem cell-derived enteric neuron model, performing high throughput screens and transcriptomic analyses to reveal excitotoxicity through muscarinic cholinergic signaling as a key driver of platin-induced enteric neuropathy. Single nuclei transcriptomics identified inhibitory nitrergic neurons as selectively vulnerable to platins, which we validated through histological analysis of the enteric nervous system in platin chemotherapy-treated patients. Lastly, we found that dampening muscarinic signaling through either pharmacologic or genetic methods is sufficient to prevent platin-induced excitotoxicity in vitro and platin-induced constipation and degeneration of nitrergic neurons in mice.

Altogether, this work succeeds in defining the mechanism underlying platin-induced gastrointestinal neurotoxicity. Furthermore, it serves as an example and framework for how stem cell-based models of the peripheral nervous system can be utilized to rapidly deconvolute mechanisms of chemotherapy-induced neurotoxicity and power drug discovery pipelines to tackle numerous peripheral neuropathies that have been intractable to date.

Nearly all essential nuclear processes act on DNA packaged into series of nucleosomes termed chromatin fibers. However, our understanding of how these processes (e.g. DNA replication, RNA transcription, chromatin extrusion, nucleosome remodeling) actually occur on such fibers remains unresolved. Our current understanding of the beads-on-a-string arrangement of nucleosomes has been built largely on high-resolution sequence-agnostic imaging methods and sequence-resolved bulk biochemical techniques. To bridge the divide between these approaches, we present the single-molecule adenine methylated oligonucleosome sequencing assay (SAMOSA). SAMOSA is a high-throughput single-molecule sequencing method that combines adenine methyltransferase footprinting and single-molecule real-time DNA sequencing to natively and nondestructively measure nucleosome positions on individual chromatin fibers. SAMOSA data allows unbiased classification of single-molecular ’states’ of nucleosome occupancy on individual chromatin fibers. We leverage this to estimate nucleosome regularity and spacing on single chromatin fibers genome-wide, at predicted transcription factor binding motifs, and across human epigenomic domains. Our analyses suggest that chromatin is comprised of both regular and irregular single- molecular oligonucleosome patterns that differ subtly in their relative abundance across epigenomic domains. This irregularity is particularly striking in constitutive heterochromatin, which has typically been viewed as a conformationally static entity. Our proof-of-concept study provides a powerful new methodology for studying nucleosome organization at a previously intractable resolution and offers up new avenues for modeling and visualizing higher order chromatin structure. ATP-dependent chromatin remodelers are one of the major regulators of patterns within the epigenome.

As a follow-up from our proof-of-concept study, we developed SAMOSA-ChAAT, a massively multiplex single-molecule footprinting platform to map the primary structure of individual, precisely-reconstituted chromatin templates subjected to virtually any chromatin-associated reaction. As proof-of-concept, we apply SAMOSA-ChAAT to study ATP-dependent chromatin remodeling by the essential imitation switch (ISWI) ATPase SNF2h, whose mechanism-of-action remains contentious. Using our approach, we discover that SNF2h operates as a density-dependent, length-sensing nucleosome sliding enzyme, whose ability to decrease or increase DNA accessibility depends on single-fiber nucleosome density. We validate our in vitro findings with single-fiber accessibility measurements in vivo, finding that the regulatory ‘mode’ of SNF2h-containing complexes (i.e. ‘opening’ vs. ‘closing’ chromatin) is dictated by the underlying nucleosome-density of individual chromatin fibers: at canonically-defined heterochromatin, SNF2h generates evenly-spaced nucleosome arrays of multiple nucleosome repeat lengths; at SNF2h-dependent accessible sites, the enzyme slides nucleosomes to increase accessibility of motifs for the essential transcription factor CTCF. Our approach and data demonstrate, for the first time, how chromatin remodelers can effectively sense nucleosome density to induce diametrically-opposed regulatory effects within the nucleus. More generally, our novel approach promises molecularly-precise views of any of the essential processes shaping nuclear physiology.

This dissertation presents a suite of computational methods and theoretical frameworks that advance our understanding of functional genomics, particularly in the context of single-cell analysis and CRISPR screening. Through the development of novel algorithms and analytical approaches, this work addresses critical challenges in processing, analyzing, and interpreting complex genomic data.

The research encompasses several interconnected areas. First, it introduces innovative approaches for studying cis-regulatory elements through massively parallel reporter assays and CRISPR interference screens, revealing distinct transcriptional networks in dementia and identifying hundreds of functional regulatory variants. Second, it presents a systematic analysis of autism spectrum disorder (ASD) risk genes during cortical neurogenesis, uncovering convergent cellular phenotypes and implicating specific molecular pathways in neurodevelopment.

The dissertation also introduces several computational tools that significantly improve existing methods in genomic analysis. These include GIA (Genomic Interval Arithmetic), a high-performance toolkit for genomic interval analysis that achieves 2-20x speed improvements over existing tools; geomux, a novel algorithm for cell identity demultiplexing in single-cell experiments that demonstrates superior accuracy in low multiplicity of infection settings; and a comprehensive CRISPR screening analysis toolkit comprising sgcount, crispr-screen, and screenviz, which streamlines the analysis of CRISPR screen data through efficient processing, statistical analysis, and visualization.

Finally, the work develops a theoretical framework for modeling gene regulatory networks, progressing from linear to increasingly sophisticated non-linear models. This culminates in a Hill-function product model capable of capturing complex biological phenomena such as multiple stable states and oscillatory behavior, while maintaining mathematical rigor and biological plausibility.

Throughout this body of work, there is a consistent emphasis on developing methods that are not only powerful and flexible but also accessible to the broader scientific community. By prioritizing computational efficiency, mathematical rigor, and user-friendliness, this research aims to democratize advanced genomic analyses and accelerate discovery across the life sciences.