Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2010. / Cataloged from PDF version of thesis. / Includes bibliographical references (p. 157-168). / The epidermal growth factor receptor (EGFR, also known as ErbB 1) is a prototypical receptor tyrosine kinase (RTK) that activates multi-kinase phosphorylation cascades to regulate diverse cellular processes, including proliferation, migration and differentiation. ErbB 1 heterooligomerizes with three close homologues: ErbB2, ErbB3 and ErbB4. ErbB1-3 receptors are frequently mutated, overexpressed or activated by autocrine or paracrine ligand production in solid tumors and have been the target of extensive drug discovery efforts. Multiple small molecule kinase inhibitors and therapeutic antibodies against ErbB receptors are in clinical use or development. Despite their importance as RTKs, oncogenes and drug targets, regulation of ErbB receptors by the interplay of conformational change, phosphorylation, phosphatases and receptor trafficking remains poorly understood, and the impact of these dynamics on physiological activity and cellular responses to anti-ErbB drugs is largely unknown. This thesis investigates the dynamic opposition of kinases and phosphatases within the ErbB pathway. By standard biochemical analysis, ErbB receptors and downstream proteins appear to become phosphorylated and then dephosphorylated in approximately 30 minutes. However, pulse chase experiments where cells are exposed to ligand and then to small molecule kinase inhibitors reveal that individual proteins must in fact cycle rapidly between being phosphorylated and dephosphorylated in seconds. We construct a succession of differential equation-based models of varying biochemical resolution, each model appropriate for analyzing a different aspect of ErbB regulation, to help interpret the data and gain quantitative insight into receptor and drug biology. Rapid phosphorylation and dephosphorylation of receptors has important implications for the assembly dynamics of signalosomes. We find that signals are rapidly propagated through some downstream pathways but slowly through others, resulting in prolonged activation in the absence of upstream signal. We show that fast phosphorylation/dephosphorylation may provide cells with the flexibility necessary to rapidly detect and respond to changes in their extracellular environment. These fast dynamics also play a crucial role in determining the response to ErbB 1-targeting cancer therapies, which we find to vary significantly between drugs with different mechanisms of action. We show that treatment with one class of these drugs results in sustained signaling, instead of inhibition, and thus may actually promote tumor proliferation or invasion. Our work may help explain why certain drugs have been more effective in patients than others and suggests new approaches for evaluating biochemical signaling networks and targeted therapeutics. / By Laura B. Kleiman. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2012. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged from student submitted PDF version of thesis. / Includes bibliographical references. / Understanding the effectiveness of a drug therapy on halting disease progression is an essential aspect of cancer biology. Conventional assays that study cell behavior after a drug intervention report the average response of a cell population which can mask the heterogeneity and dynamics of seemingly identical cells. Recently, many single-cell techniques have been developed, but there are currently no methods that can fully characterize the long-term effects of drug treatment on cancer cell growth. To accomplish such, we developed an instrument to measure single-cell growth before and after drug treatment. In order to achieve femtogram-level mass resolution, we employed the suspended microchannel resonator (SMR), a vacuum-packaged cantilever with an embedded channel. Here, we describe three implementations that involve different technologies (optical trap, mechanical trap, and dynamic ow trapping) to capture a cell for repeated measurements and to perform drug delivery. Applying the technique we developed based on the dynamic ow trapping, we were able to monitor one or more generations of a cancer cell before and after drug treatment. We investigated the growth of mouse leukemia cells in response to drugs that inhibit the mammalian target of rapamycin (mTOR) pathway, induce apoptosis, or prevent translational activity directly at the ribosome. Our method was able to discern a particular growth signature for each drug investigated and to discover a new phenotype in cells following mTOR inhibition. Furthermore, our data demonstrates that the instantaneous growth rate changes following a drug treatment could potentially predict the long-term inhibitory effect on cellular biogenesis and mass accumulation. / by Yaochung Weng. / Ph.D.

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2016. / Cataloged from PDF version of thesis. / Includes bibliographical references. / A protein's function in the cell depends on its structure, which in turn depends on the intracellular environment. Stress like heat shock or nutrient starvation can alter intracellular conditions, leading to protein misfolding - i.e. the inability of a protein to reach or maintain its native conformation. Since many proteins interact with each other, protein misfolding and cellular stress response must be examined both on the scale of individual protein conformational changes and on a more global level, where interaction patterns can reveal larger-scale protein responses to cellular stress. On the individual scale, one example of a protein particularly susceptible to misfolding is the human von Hippel-Lindau (VHL) tumor suppressor. When expressed in the absence of its cofactors, VHL cannot fold correctly and is quickly degraded by the cell's quality control machinery. Here, I present a biophysical characterization of a VHL mutation that confers increased resistance to misfolding. Mathematical modeling provides an explanation for this mutant's increased stability in the cell by predicting how its cofactor and chaperone interaction sites are buried or exposed in the protein's predicted conformation. On a more global level, a budding yeast cell undergoing glucose deprivation both acidifies its cytosol and exhibits widespread protein clustering. By employing a proteome-wide computational assay, I examine how this drop in pH could lead to the formation of higher order protein structures. This modeling framework also provides a rationale for why these two related phenotypes might be beneficial, since protein clustering can help regulate relevant metabolic pathways and provide protection from protein misfolding and/or degradation. / by Kelly Paige Brock. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2018. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged from student-submitted PDF version of thesis. / Includes bibliographical references (pages 205-226). / RNA not only shuttles information between DNA and proteins but also carries out many other essential cellular functions. Nearly all steps of an RNA's life cycle are controlled by approximately one thousand RNA binding proteins (RBPs) that direct RNA splicing, cleavage and polyadenylation, localization, translation, and degradation. Despite the central role of RBPs in RNA processing and gene expression, they have been less well studied than DNA binding proteins, in part due to the historical dearth of technologies to probe RBP binding and activity in a high-throughput, comprehensive manner. In this thesis, I describe the affinity landscapes of the largest set of human RBPs to date elucidated through a highthroughput version of RNA Bind-N-Seq (RBNS), an unbiased in vitro assay that determines the primary sequence, secondary structure, and contextual preferences of an RBP. The 78 RBPs bound an unexpectedly low diversity of RNA motifs, implying convergence of binding specificity toward a small set of RNA motifs characterized by low compositional complexity. Offsetting the low diversity of sequence motifs, extensive preferences for contextual features beyond short linear motifs were observed, including bipartite motifs, flanking nucleotide content, and preference for or against RNA structure. These features likely refine which motif occurrences are selected in cells, enabling RBPs that bind the same linear motif to act on distinct subsets of transcripts. Additionally, RBNS data is integrated with complementary in vivo binding sites from enhanced crosslinking and immunoprecipitation (eCLIP) and functional (RNAi/RNA-seq) data produced through collaborative efforts with the ENCODE consortium. These data enable creation of "RNA maps" of RBP activity in pre-mRNA splicing and gene expression levels, either with (eCLIP) or without (RBNS) crosslinking-based assays. The mapping and characterization of RNA elements recognized by over 200 human RBPs is also presented in two human cell lines, K562 and HepG2 cells. Together, these novel data augment the catalog of functional elements encoded in the human genome to include those that act at the RNA level and provide a basis for how RBPs select their RNA targets, a fundamental requirement in more fully understanding RNA processing mechanisms and outcomes. / by Peter Freese. / Ph. D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2011. / Cataloged from PDF version of thesis. / Includes bibliographical references. / Cellular signaling and regulatory networks underlie fundamental biological processes such as growth, differentiation, and response to the environment. Although there are now various high-throughput methods for studying these processes, knowledge of them remains fragmentary. Typically, the majority of hits identified by transcriptional, proteomic, and genetic assays lie outside of the expected pathways. In addition, not all components in the regulatory networks can be exposed in one experiment because of systematic biases in the assays. These unexpected and hidden components of the cellular response are often the most interesting, because they can provide new insights into biological processes and potentially reveal new therapeutic approaches. However, they are also the most difficult to interpret. We present a technique, based on the Steiner tree problem, that uses a probabilistic protein-protein interaction network and high confidence measurement and prediction of protein-DNA interactions, to determine how these hits are organized into functionally coherent pathways, revealing many components of the cellular response that are not readily apparent in the original data. We report the results of applying this method to (1) phosphoproteomic and transcriptional data from the pheromone response in yeast, and (2) phosphoproteomic, DNaseI hypersensitivity sequencing and mRNA profiling data from the U87MG glioblastoma cell lines over-expressing the variant III mutant of the epidermal growth factor receptor (EGFRvIII). In both cases the method identifies changes in diverse cellular processes that extend far beyond the expected pathways. Analysis of the EGFRVIII network connectivity property and transcriptional regulators that link observed changes in protein phosphorylation and differential expression suggest a few intriguing hypotheses that may lead to improved therapeutic strategy for glioblastoma. / by Shao-shan Carol Huang. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2011. / Cataloged from PDF version of thesis. / Includes bibliographical references. / An ability to control the degree of heterogeneity in cellular phenotypes may be important for cell populations to survive uncertain and ever-changing environments or make cell-fate decisions in response to external stimuli. Cells may control the degree of gene expression heterogeneity and ultimately levels of phenotypic heterogeneity by modulating promoter switching dynamics. In this thesis, I investigated various mechanisms by which heterogeneity in the expression of FLO 11 in S. cerevisiae could be generated and controlled. First, we show that two copies of the FLOJ1 locus in S. cerevisiae switch between a silenced and competent promoter state in a random and independent fashion, implying that the molecular event leading to the transition occurs in cis. Through further quantification of the effect of trans regulators on both the slow epigenetic transitions between a silenced and competent promoter state and the fast promoter transitions associated with conventional regulation of FLO11, we found different classes of regulators affect epigenetic, conventional, or both forms of regulation. Distributing kinetic control of epigenetic silencing and conventional gene activation offers cells flexibility in shaping the distribution of gene expression and phenotype within a population. Next, we demonstrate how multiple molecular events occurring at a gene's promoter could lead to an overall slow step in cis. At the FLO] 1 promoter, we show that at least two pathways that recruit histone deacetylases to the promoter and in vivo association between the region -1.2 kb from the ATG start site of the FLO11 ORF and the core promoter region are all required for a stable silenced state. To generate bimodal gene expression, the activator Msnlp forms an alternate looped conformation, where the core promoter associates with the non-coding RNA PWR1's promoter and terminator regions, located at -2.1 kb and -3.0 kb from the ATG start site of the FLO]1 ORF respectively. Formation of the active looped conformation is required for Msnlp's ability to stabilize the competent state without destabilizing the silenced state and generate a bimodal response. Our results support a model where multiple stochastic steps at the promoter are required to transition between the silenced and active states, leading to an overall slow step in cis. Finally, preliminary investigations of heterozygous diploids revealed possible transvection occurring at FLO] 1, where a silenced allele of FLO 11 appeared to transfer silencing factors to a desilenced FLO11 allele on the homologous chromosome. These observations suggest a new mechanism through which heterogeneity in FL011 expression could be further controlled, in addition to the molecular events at the FL011 promoter we elucidated previously. / by Leah M. Octavio. / Ph.D.

Thesis (S.M.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2013. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged from student-submitted PDF version of thesis. / Includes bibliographical references (p. 41-44). / Understanding the cell as a system has become one of the foremost challenges in the post-genomic era. As a result of advances in high-throughput (HTP) methodologies, we have seen a rapid growth in new types of data at the whole-genome scale. Over the last decade, HTP experimental techniques such as yeast two-hybrid assays and co-affinity purification couple with mass spectrometry have generated large amounts of data on protein-protein interactions (PPI) for many organisms. We focus on the sub-domain of systems biology related to understanding the interactions between proteins that ultimately drive all cellular processes. Representing PPIs as a protein interaction network has proved to be a powerful tool for understanding PPIs at the systems level. In this representation, each node represents a protein and each edge between two nodes represents a physical interaction between the corresponding two proteins. With this abstraction, we present algorithms for the prediction and analysis of such PPI networks as well as web servers and databases for their public availability: 1. In many organisms, the coverage of experimental determined PPI data remains relatively noisy and limited. Given two protein sequences, we describe an algorithm, called Struct2Net, to predict if two proteins physically interact, using insights from structural biology and logistic regression. Furthermore, we create a community-wide web-resource that predicts interactions between any protein sequence pair and provides proteome-wide pre-computed PPI predictions for Homo sapiens, Drosophila melanogaster, and Saccharomyces cerevisiae. 2. Comparative analysis of PPI networks across organisms can provide valuable insights into evolutionary conservation. We describe an algorithm, called IsoRank, for global alignment of multiple PPI networks. The algorithm first constructs an eigenvalue problem that models the network and sequence similarity constraints. The solution of the problem describes a k partite graph that is further processed to find the alignments. Furthermore, we create a communitywide web database, called IsoBase, that provides network alignments and orthology mappings for the most commonly studied eukaryotic model organisms: Homo sapiens, Mus musculus, Drosophila melanogaster, Caenorhabditis elegans, and Saccharomyces cerevisiae. / by Daniel K. Park. / S.M.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2010. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged from student-submitted PDF version of thesis. / Includes bibliographical references (p. 218-228). / Understanding the microbial world is key to understanding global biogeochemistry, human health and disease, yet this world is largely inaccessible. Microbial genomes, an increasingly accessible data source, provide an ideal entry point. The genome sequences of different microbes may be compared using the tools of population genetics to infer important genetic changes allowing them to diversify ecologically and adapt to distinct ecological niches. Yet the toolkit of population genetics was developed largely with sexual eukaryotes in mind. In this work, I assess and develop tools for inferring natural selection in microbial genomes. Many tools rely on population genetics theory, and thus require defining distinct populations, or species, of bacteria. Because sex (recombination) is not required for reproduction, some bacteria recombine only rarely, while others are extremely promiscuous, exchanging genes across great genetic distances. This behavior poses a challenge for defining microbial population boundaries. This thesis begins with a discussion of how recombination and positive selection interact to promote ecological adaptation. I then describe a general pipeline for quantifying the impacts of mutation, recombination and selection on microbial genomes, and apply it to two closely related, yet ecologically distinct populations of Vibrio splendidus, each with its own microhabitat preference. I introduce a new tool, STARRInIGHTS, for inferring homologous recombination events. By assessing rates of recombination within and between ecological populations, I conclude that ecological differentiation is driven by small number of habitat-specific alleles, while most loci are shared freely across habitats. The remainder of the thesis focuses on lineage-specific changes in natural selection among anciently diverged species of gamma proteobacteria. I develop two new metrics, selective signatures and slow:fast, for detecting deviations from the expected rate of evolution in 'core' proteins (present in single copy in most species). Because they rely on empirical distributions of evolutionary rates across species, these methods should become increasingly powerful as more and more microbial genomes are sampled. Overall, the methods described here significantly expand the repertoire of tools available for microbial population genomics, both for investigating the process of ecological differentiation at the finest of time scales, and over billions of years of microbial evolution. / by B. Jesse Shapiro. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2013. / Cataloged from PDF version of thesis. / Includes bibliographical references. / The rapid development and applications of high throughput measurement techniques bring the biological sciences into a 'big data' era. The vast available data for enzyme and metabolite concentrations, fluxes, and kinetics under normal or perturbed conditions in biological networks provide unprecedented opportunities to understand the cell functions. On the other hand, it brings new challenges of handling, integrating, and interpreting the large amount of data to acquire novel biological knowledge. In this thesis, we address this problem with a new ordinary differential equation (ODE) model based on the mass-action rate law (MRL) of the biochemical reactions. It describes the detailed biochemical mechanisms of the enzyme reactions, and therefore reflects closely of how the enzymes work in the systems. Because the MRL models are constructed with elementary enzyme reaction steps, it is also much more flexible than the aggregated rate law (ARL) model to incorporate new enzyme interactions and regulations. Two versions of the MRL model ensembles for the central carbon metabolic network, which generates most of the precursors for the secondary metabolite, were constructed. The E. coli version contains the basic reactions in this network and was applied to optimize the aromatic amino acid production which requires fine-tuned flux partition between glycolysis pathway and the pentose phosphate pathway. The S. cerevisiae version is more sophisticated with the incorporated dynamics of the NAD/NADH and NADP/NADPH, as well as the automatic switch from aerobic to anaerobic condition. It was applied to maximize the ethanol production yield, for which the NAD/NADH ratio is a crucial regulating factor. In order to develop methodologies to understand the intrinsic network properties and optimize the network behavior, we further explored approaches for the identification of pathway bottlenecks. Four computational assays were studied, including metabolite accumulation, conditional Vmax, increased glucose input, and decreased E₀, which were applied to the ethanol model ensemble to discover their effectiveness in bottleneck identification in this network. The TDH reaction was detected as a major bottleneck restricting carbon flow towards the ethanol pathway and affecting NADH availability. To manipulate the network for desired production rates of target metabolites, we developed an optimization technique for mass-action rate law ODE models that allows parallel or sequential combinations of enzyme knock-out and over-/under-expression strategies to be conducted on the model. Many strategies were suggested to improve the aromatic amino acid production and help identify the two-direction flux feature of the pentose phosphate pathway. Strategies were also found to enhance the ethanol production yield above 95% of the theoretical yield. Although the two applications studied here are both in the field of metabolic engineering, it is anticipated that the mass-action rate law models for the central carbon metabolism can be extended to study the cancer metabolism. Preliminary studies show promising results for designing cancer clinical trial simulations with a combined model incorporating high level cancer progression and detailed cancer biochemical metabolism. / by Yuanyuan Cui. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2012. / Cataloged from PDF version of thesis. / Includes bibliographical references. / With the emergence of the 2009 A(H1N1) pandemic influenza virus and the rapid spread of drug resistance in recent years, the need to develop new anti-influenza drugs that can reduce the emergence of new resistant viruses is both urgent and important. This thesis explores the use of polymer-attached inhibitors as a new approach in the development of anti-influenza drugs, with particular focus on polymer-attached zanamivir (ZA). We have previously shown that covalently conjugating multiple copies of ZA via a flexible linker to poly-L-glutamine greatly enhances antiviral potency. In the first study, we have elucidated the mechanism of this phenomenon. Like ZA itself, the polymer-attached inhibitor binds specifically to viral neuraminidase and inhibits both its enzymatic activity and the release of newly synthesized virions from infected cells. In contrast to monomeric ZA, however, the polymer-attached drug also synergistically inhibits virus-endosome fusion, thus contributing to the dramatically increased antiviral potency. Next, we went on to investigate polymer-attached ZA's effect on the emergence of drug resistance. We found that viruses adapted rapidly to growing in high concentrations of monomeric ZA, whereas viral growth remained inhibited by low concentrations of polymer-attached ZA even after 23 passages in cell culture. Sequencing analysis established the emergence of an amino acid substitution known to confer ZA resistance (E119G in neuraminidase) after 8 passages of monomeric ZA selection. In contrast, virus grown in polymer-attached ZA remained free of substitutions in E119, and other known resistance-associated residues. We instead found novel substitutions in hemagglutinin (R220G, D241G) and neuraminidase (G111D), which emerged during passages 14-17. Importantly, although the drug-selected variants were resistant to monomeric ZA, the viruses remained susceptible to low pM concentrations of polymer-attached ZA itself. Taken together, these data demonstrate that attaching the drug to a polymeric chain (i) confers a new mechanism of antiviral action; (ii) significantly delays the emergence of drug resistance; and (iii) enhances potency against the selected ZA-resistant variants. The studies presented in this thesis provide further impetus for the use of polymer-attached inhibitors as influenza therapy, and as tools for better understanding of influenza biology. / by Chia Min Lee. / Ph.D.