Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2019 / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 147-166). / Genomic contact frequency maps obtained from high throughput chromosome conformation capture technologies have revealed several organizing patterns of mammalian interphase chromosomes, including self-interacting topologically associating domains (TADs) which are believed to function as coherent gene regulatory neighborhoods. However, the mechanisms driving these patterns are still unknown. In this thesis, I describe and apply computational methods that test the predictions of a recently proposed loop extrusion model in the context of experimental perturbations of its key molecular players. In the first project I introduce a new data model, file format, and supporting software package to cope with the challenges of the increasing size and resolution of Hi-C datasets, including a parallel and scalable matrix balancing implementation. / In the second project, I show that depletion of the Structural Maintenance of Chromosomes (SMC) complex, cohesin, in non-cycling mouse liver cells completely eliminates the appearance of TADs in Hi-C maps while preserving genome compartmentalization. In the third project, I demonstrate that depletion of a closely related SMC complex, condensin II, which plays a major role in mitotic chromosome condensation but is also found in the nucleus in interphase, has no impact on gene expression or the maintenance of genome organization in non-dividing cells. In the final project, I compile further evidence for loop extrusion in interphase by employing a combination of polymer simulations and meta-analysis of several Hi-C studies that performed targeted perturbations to modulate the presence of cohesin and the insulator protein, CTCF, on chromatin. / Together, these projects show that rather than being folded in a hierarchical fashion, mammalian genomes in interphase are organized by at least two distinct and antagonistic processes: global compartmental segregation dependent on epigenetic state, and local compaction dependent on cohesin. The latter process is likely to be the dynamic extrusion of chromatin loops driven by a yet-to-be-characterized motor activity of cohesin complexes and limited by DNA-bound CTCF extrusion barriers. / by Nezar Abdennur. / Ph. D. / Ph.D. Massachusetts Institute of Technology, Computational and Systems Biology Program

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2019 / Cataloged from PDF version of thesis. / Includes bibliographical references. / The complex etiology of neurodegenerative diseases is not fully understood, and the characterization of cellular pathways that are dysfunctional in these diseases is key for therapeutic development. Chemical and genetic perturbagens can probe cellular pathways to shed insight about both disease etiology and potential therapeutic targets. We analyzed the functional effects of chemical perturbagens in neurodegenerative disease models as evidenced by changes in transcriptomic, metabolomic, epigenomic, and proteomic data ("multi-omics" data). Our studies revealed novel modes of action for small molecule compounds that promote survival in a model of Huntington's Disease, a fatal neurodegenerative disorder. Integration of our multi-omics data using an interpretable network approach revealed that the autophagy and bioenergetics cellular pathways are affected by different sets of compounds that promote survival. Using staining and western blot assays, we validated the effect on autophagy for one set of compounds and found that the compounds activate this pathway. Using a cellular bioenergetics assay, we found that a second set of compounds shifts the bioenergetic flux from mitochondrial respiration to glycolysis, validating our network results. In a second study related to Huntington's Disease, we analyzed the effects of two peripheral huntingtin gene silencing techniques in mouse liver. We show that the transcriptional and metabolomic changes associated with both genetic silencing methods converge on similar cellular pathways, such as the immune response and fatty acid metabolism. As a whole, this thesis presents new insights into the functional effects of perturbagens that could impact neurodegenerative disease pathology and drug discovery. / by Natasha L. Patel-Murray. / Ph. D. / Ph.D. Massachusetts Institute of Technology, Computational and Systems Biology Program

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2019 / Cataloged from PDF version of thesis. "June 2019." / Includes bibliographical references. / The development of high-throughput techniques to observe and perturb biological systems has led to remarkable progress in the last several decades. From the tremendous amounts of data being accumulated, new opportunities have emerged, including the possibility of finding latent patterns in high-dimensional variables that are reflective of underlying biological processes. While these methods have led to countless discoveries and innovations, it is clear there is much more we could learn by measuring and perturbing at far greater scales. Here, I advance methods to understand and utilize latent patterns in new types of high-dimensional data. I devise a method of analyzing networks of 'frequency interactions' in 16S/18S time series data, showing that these can be used to identify microbial communities and associated environmental factors. / Then, as part of a highly collaborative project, I show how latent patterns in single cell RNA-Seq can be used together with optimal transport analysis to identify cell types and cell type trajectories, regulatory pathways, and cell-cell interactions in a time-course of developmental reprogramming. I then step back to ask a fundamental question: how do we choose which observations and perturbations to make, and how many of each are necessary? I approach this question on the basis of the inherency of latent structure in biology, and on foundational mathematical results concerning the analysis of highly-structured data. I present the beginnings of a framework to formalize how random composite experiments can make biological discovery more efficient by leveraging latent patterns. I first show how to recover individual genomes using covariance patterns in a series of composite (meta-) genomic data. / I then describe how random composite measurements and compressed sensing can be used to make gene expression profiling more efficient. Finally, I apply this idea to in situ imaging transcriptomics, demonstrating how many individual gene images can be efficiently recovered from a small number of composite gene images. / by Brian Cleary. / Ph. D. / Ph.D. Massachusetts Institute of Technology, Computational and Systems Biology Program

Tuberculous pericarditis is an extra pulmonary form of tuberculosis (TB) which leads to a lifethreatening form of pericardial fibrosis in up to 25% of patients despite anti tuberculous therapy. The mechanisms leading to the fibrotic phenotype following infection are poorly understood. A proof of concept study revealed decreased levels of the antifibrotic N-acetylseryl-aspartyl-lysyl-proline or Ac-SDKP in tuberculous pericardial fluid as compared to control (non infectious) pericardial fluid. Ac-SDKP is a physiological peptide that is synthesised from its precursor protein thymosin β4 by the sequential action of meprin-α and prolyl oligopeptidase (POP) and is cleaved by angiotensin-1 converting enzyme (ACE). Importantly, a role of ACE and Ac-SDKP in the regulation of inflammation and fibrosis in multiple tissues and organs has been increasingly described in the literature. This has prompted interest in both the mechanisms of and potential for protective benefits of ACE inhibitors and Ac-SDKP analogue administration in fibrotic disease. The aim of this project was to investigate a) the molecular mechanisms of the antifibrotic effects of Ac-SDKP in the development of fibrosis, particularly in TB pericarditis, and b) the potential of ACEi and Ac-SDKP analogues in vitro in fibrosis prevention. Pericardial fluid and blood samples from patients with TB pericarditis or undergoing coronary artery bypass surgery (non-infectious controls) was used to investigate the metabolism of AcSDKP in the tuberculous pericardium. Ac-SDKP levels as measured by ELISA, were significantly decreased (2.3 fold) in TB pericardial fluid as compared to controls. This reduction in Ac-SDKP levels was accompanied by a local 28% increase in the enzymatic activity of ACE, but no change in POP enzyme activity levels, both of which were measured using fluorogenic assays. This suggests that an increase in ACE activity in the pericardium following infection by the mycobacterium leads to a reduction of the levels of the antifibrotic peptide which is likely to contribute to the pathophysiology of fibrosing pericarditis. A mass spectrometric (MS) approach was employed in order to identify proteins whose expression is modulated by the effect of Ac-SDKP in the proteome and secretome of a human lung fibroblast cell line (WI-38). Label free quantitative MS was employed to identify 114 and 44 differentially expressed proteins in Ac-SDKP fibroblast proteome and secretome respectively. Various extracellular matrix components and their related factors such as collagens, cytoskeletal proteins and inflammatory proteins, were identified among the differentially regulated proteins. Reactome pathway analysis confirmed the significant enrichment of Ac-SDKP-related extracellular matrix proteoglycans and extracellular matrix in the differentially expressed proteins of the secretome. Using the same cell line, the antifibrotic effects of Ac-SDKP analogues and ACE inhibitors were investigated through quantitative western blotting for transforming growth factor β (TGF-β) and Smad 3 levels, and using a hydroxyproline assay. Ac-SDKP prevented TGF-β and collagen expression through inhibition of Smad 3 phosphorylation. The Ac-SDψKP analogue (whereby the peptide bond between the aspartate and lysine is reduced) alone prevented TGF-β mediated collagen secretion. The combination of Ac-SDKP and the N domain-selective inhibitor RXP407, but not the non-selective lisinopril had an additive effect on the inhibition of collagen in fibroblasts. However, the antifibrotic effect of Ac-SDψKP was comparable to the combination of Ac-SDKP and RXP407 and was not improved with added ACE inhibition. Finally, the ACE signalling response to Ac-SDKP and the ACE inhibitors RXP407 and lisinopril was investigated using mass spectrometry and quantitative western blotting for phospho JNK and JNK. The ACE inhibitors as well as Ac-SDKP triggered the ACE signalling cascade to induce JNK phosphorylation. This highlights a potential new mechanism for the anti-inflammatory and antifibrotic effects of Ac-SDKP and the inhibitors. This thesis has demonstrated an altered metabolism of Ac-SDKP is associated with increased ACE activity in the tuberculous pericardium. It has also provided a deeper understanding of the antifibrotic action of the tetrapeptide, and in vitro evidence for the use of the analogue AcSDψKP and inhibtion of N domain catalytic activity for decreasing fibrosis. These findings form a solid basis for future in vivo pharmacological studies on the effects of Ac-SDKP analogues and ACE inhibitors in the prevention and management of fibrotic conditions. Importantly, these therapeutic options present an exciting avenue to follow in the prevention of fibrosing pericarditis in TB pericarditis.

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, February, 2021 / Cataloged from the official PDF of thesis. / Includes bibliographical references (pages 261-281). / Biological processes from differentiation to disease progression are governed by gene regulatory mechanisms. Currently large-scale omics and imaging data sets are being collected to characterize gene regulation at every level. Such data sets present new opportunities and challenges for extracting biological insights and elucidating the gene regulatory logic of cells. In this thesis, I present computational methods for the analysis and integration of various data types used for cell profiling. Specifically, I focus on analyzing and linking gene expression with the 3D organization of the genome. First, I describe methodologies for elucidating gene regulatory mechanisms by considering multiple data modalities. I design a computational framework for identifying colocalized and coregulated chromosome regions by integrating gene expression and epigenetic marks with 3D interactions using network analysis. / Then, I provide a general framework for data integration using autoencoders and apply it for the integration and translation between gene expression and chromatin images of naive T-cells. Second, I describe methods for analyzing single modalities such as contact frequency data, which measures the spatial organization of the genome, and gene expression data. Given the important role of the 3D genome organization in gene regulation, I present a methodology for reconstructing the 3D diploid conformation of the genome from contact frequency data. Given the ubiquity of gene expression data and the recent advances in single-cell RNA-sequencing technologies as well as the need for causal modeling of gene regulatory mechanisms, I then describe an algorithm as well as a software tool, difference causal inference (DCI), for learning causal gene regulatory networks from gene expression data. / DCI addresses the problem of directly learning differences between causal gene regulatory networks given gene expression data from two related conditions. Finally, I shift my focus from basic biology to drug discovery. Given the current COVID19 pandemic, I present a computational drug repurposing platform that enables the identification of FDA approved compounds for drug repurposing and investigation of potential causal drug mechanisms. This framework relies on identifying drugs that reverse the signature of the infection in the space learned by an autoencoder and then uses causal inference to identify putative drug mechanisms. / by Anastasiya Belyaeva. / Ph. D. / Ph.D. Massachusetts Institute of Technology, Computational and Systems Biology Program

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, February, 2021 / Cataloged from the official PDF of thesis. "February 2021." / Includes bibliographical references. / Tumor progression, from the single mutated cell to the advanced stages of cancer, represents an evolutionary process. During tumor progression, cancer cells acquire new genetic mutations, becoming more heterogeneous, leading to tumor progression and resistance to therapy. However, clear genetic drivers of progression, metastasis, and therapeutic resistance are identified in only a subset of tumors, pointing to non-genetic contributors to cancer progression. Also, somatic evolution in cancer is occurring at the level of the single cell. Therefore, the application of the single cell genomic method is crucial for deciphering phenotypic heterogeneity. Here, we profiled single cell transcriptomes from genetically engineered mouse lung tumors at seven stages spanning tumor progression from atypical adenomatous hyperplasia to lung adenocarcinoma. The diversity of transcriptional states spanned by tumor cells increased over time and was reproducible across tumors and mice, but was not explained by genomic copy number variation. Cancer cells progressively adopted alternate lineage identities, computationally predicted to be mediated through a common transitional, high-plasticity cell state (HPCS). HPCS cells prospectively isolated from mouse tumors had robust potential for phenotypic switching and tumor formation and were more chemoresistant in mice. Our study reveals transitions that connect cell states across tumor evolution and motivates therapeutic targeting of the HPCS. / by Nemanja Marjanovic. / Ph. D. / Ph.D. Massachusetts Institute of Technology, Computational and Systems Biology Program

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2019 / Cataloged from the PDF of thesis. / Includes bibliographical references (pages 87-90). / In this thesis I observe evolutionary signatures in coding regions to: (1) understand the sources of highly mutable coding regions in mammals; (2) to elucidate a new candidate function for a stop codon readthrough candidate gene, BRI3BP; and (3) to show how rapid sequence-based structure approximations can help predict the structural impact of amino-acid changes. (1) First, I searched for deviations from the evolutionary signatures of coding regions to recognize synonymous acceleration elements (SAEs) in protein coding genes. I showed that these are driven by an increased mutation rate, which persists in the human lineage, in otherwise evolutionarily-constrained protein-coding regions, providing an important resource to better characterize protein-coding constraint in mammals and within humans. (2) Second, I combined evolutionary signatures at the protein-coding and protein-folding level to characterize the functional implication of stop-codon readthrough in BRI3BP. I showed that this readthrough region has conserved spaced hydrophobic residues that pattern match to the -terminal helix forming a coiled-coil-like domain. This change alters BRI3BP function from pro-growth to pro-apoptotic, similarly to VEGF-A. This suggests that readthrough-triggered apoptosis may represent a general mechanism for limiting growth of cells with aberrant ribosomal termination. (3) Third, I used rapid protein-structure approximation of burial of residues based on protein sequence to predict the structural impact of amino acid alterations. I show that the prediction can be improved over using exclusively the hydrophobicity change of the residue. Overall my work demonstrates how evolutionary and structural signatures can be used to predict highly mutational gene regions, readthrough function and structural impact of mutation. / by Maxim Wolf. / Ph. D. / Ph.D. Massachusetts Institute of Technology, Computational and Systems Biology Program

Microbial communities (microbiomes) comprise a vast component of life on our planet. They are involved in many fundamental processes, ranging from balancing global biogeochemical cycles to influencing human health. Recently, advances in genome sequencing technologies have allowed us to explore the genetic diversity of microbiomes in high-throughput, cataloging hundreds of thousands of microbial species and millions of genes. As genomic data is accumulating, the challenge remains: to translate genome sequences into functional predictions of relevant phenotypes. A promising approach to address this challenge is the annotation of genomic data to a metabolic network (referred to as genome-scale metabolic model reconstruction), which can then be analyzed to simulate metabolic phenotypes. Although this approach has provided valuable insight into microbial phenotypes, there are many sources of uncertainty in both reconstruction and analysis of genome-scale metabolic networks that currently limit their application. The development of improved reconstruction and analysis methods, and additional sources of data, that further address this uncertainty would facilitate our understanding of microbial community function. The first section of this dissertation is a review that outlines the major uncertainties along a general pipeline for genome-scale metabolic model reconstruction and analysis, and highlights existing approaches for addressing them. An emphasis is placed on probabilistic and ensemble based methods that can be used to formally represent uncertainty and facilitate the crystallization of metabolic network knowledge. The second section of this dissertation introduces a new probabilistic genome-scale metabolic model analysis method, inspired by percolation theory, to quantify the biosynthetic capabilities of microbial organisms in uncertain environments. This method was applied to microbial organisms from the human oral microbiome, providing broad insight into the structure of this microbial community. The third section of this dissertation describes the development of an experimental device to facilitate the collection of data related to metabolic interactions between microbes. The data collected with this device was probabilistically integrated with a mechanistic metabolic model to gain quantitative insight into the syntrophic interaction between an engineered E. coli auxotroph pair. Together, the work described in this dissertation introduces several novel probabilistic methods for metabolic modeling of microbial communities, and sets the stage for future work that can further improve our understanding of these important biological systems. / 2021-09-28T00:00:00Z

Within the last raw years several areas thin the state of Utah have been and are being developed for the conservation of migratory birds. Of utmost importance in this respect is the management of these areas so as to obtain a sufficient supply of rood plants for use by water fowl. Previous work done at the Bear River Migratory Waterfowl Refuge shows that the productivity or aquatic rood plants, chiefly Potamogeton pectinates L. and Ruppis marittima L. varies with different localities and that these differences could not be attributed to chemical conditions of the water, i.e., dissolved oxygen, alkalinity, and hydrogen ion. Inasmuch as little was done on the soils. it was felt that soil characteristics might have some street on the productivity and distribution or the aquatic plants at this locality. Results from such a problem would undoubtedly aid in rut development or these habitats; hence the study seems opportune and might be or value.

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2019 / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 141-151). / Models of biochemical reaction networks in cells are important for advancing our understanding of complex biological systems and for designing functional synthetic biological circuits. However, most models are based on a deterministic digital framework that is largely incompatible with nonlinear dynamics, stochastics, high-order feedback, cross talk, loading, and resource consumption in biology. In contrast, analog circuit design is the nearly 100-year-old art of crafting and analyzing nonlinear, stochastic, coupled differential equations to perform a desired task, often to given speed, precision, input sensitivity, power, load, or part-count constraints and in the presence of noise or device mismatch. In this thesis, we develop a canonical analog circuit that maps a wide class of biological circuits, whether at the DNA, RNA, protein, or small-molecule levels to design schematics that represent their underlying dynamical differential equations exactly. / We then apply techniques from analog feedback circuit design to two concrete biological circuits to improve their feedback performance: 1) We show that a novel synthetic microbial operational amplifier (OpAmp) with three amplification stages based on DNA, RNA, and protein stages and a dominant time constant is capable of high open-loop gain, stable, and robust-and-precise closed-loop performance; 2) We show that a synthetic tissue-homeostasis stem-cell circuit with a novel incoherent feed-forward loop attenuates negative phase and thus improves its robustness and precision of response to cell death in Type I diabetes. We also show that our novel use of both asymmetric division and symmetric division of stem cells improves feedback-loop performance w.r.t transients and robustness. To illustrate scalability of our approach to large-scale and high-speed simulations of the future, we use digitally programmable analog microelectronic chips to run complex simulations in parallel. / We develop a mapping that converts our analog schematics to a corresponding configuration on these chips, and demonstrate how to optimize the parameters of the biological OpAmp for high gain and improved performance. Our work illustrates that synthetic analog feed back control in living cells is amenable to rigorous design, analysis, simulation and implementation with the tools of analog circuit design, and leads to novel and experimentally useful synthetic biological circuits. / by Jonathan, Jin Yuan, Teo. / Ph. D. / Ph.D. Massachusetts Institute of Technology, Computational and Systems Biology Program