<p> The Glucagon-like peptide 1 receptor (GLP-1R) belongs to the pharmaceutically important Class B family of G-protein coupled receptors (GPCRs) and its incretin peptide ligand GLP-1 analogs are adopted drugs for the treatment of type 2 diabetes (T2D). Despite remarkable anti-diabetic effects, Glucagon Like Peptide-1 (GLP-1) peptide-based drugs are limited by the need of injection or high cost oral formulation. On the other hand, developing non-peptide small molecule drugs targeting GLP-1R remains elusive likely due to the large nature of the orthosteric binding site on GLP-1R. A promising approach is to develop small molecule agonistic positive allosteric modulators (ago-PAMs) or positive allosteric modulators (PAMs) of GLP-1R by targeting the potential allosteric sites in the transmembrane (TM) domain of human GLP-1R. </p><p> As the first step of taking this approach, we constructed a three-dimensional structure model of the TM domain of human GLP-1R using homology modeling and conformational sampling techniques. Next, a potential allosteric binding site on the TM domain was predicted computationally. <i>In silico</i> screening of drug-like compounds against this predicted allosteric site has identified nine compounds as potential GLP-1R agonists. The independent agonistic activity of two compounds was subsequently confirmed using cyclic adenosine monophosphate (cAMP) response element (CRE)-based luciferase reporting system. One compound was also shown to stimulate insulin secretion through <i> in vitro</i> assay. In addition, this compound synergized with GLP-1 to activate human GLP-1R. </p><p> In 2017, the crystal structures of GLP-1R in its active state (PDB ID: 5VAI) became available. Hence, we have performed another round of <i> in silico</i> screening employing this structure. First, the potential ligand binding sites in 5VAI were identified using computational tools and <i> in silico</i> screening procedure as described above was carried out again. A new small 8 molecule with low molecular weight and logP was identified. <i> In vitro</i> studies of this compound confirmed that it acts as the ago-Positive Allosteric Modulator (PAM) of GLP-1R that improves GLP-1's affinity and efficacy towards GLP-1R. When used in combination with GLP-1, this compound improves insulin secretion than using GLP-1 alone. Site specific mutagenesis studies confirmed its binding site as predicted in the TM domain of GLP-1R. </p><p> Finally, this ago-PAM molecule was further optimized to improve its potency and specificity towards GLP-1R using structure-based optimization strategy and medicinal synthesis. The newly designed compound, whose molecular weight was less than the parental compound, was found to act as the PAM of GLP-1R and showed improvement in the specificity than the parental compound. Thus, this new compound could be further exploited in the drug development for T2D treatment. </p><p> These results demonstrated that allosteric regulation exists in GLP-1R and can be exploited for developing small molecule agonists. The success of this work will help pave the way for small molecule drug discovery targeting other Class B GPCRs through allosteric regulations.</p><p>

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2016. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 107-115). / Spatiotemporally coordinated cell behavior is observed during morphogenesis, in both embryonic development as well as tissue regeneration. An open question is how individual cells collectively generate force to achieve the correct tissue architecture. This thesis examines how the apical forces generated by Drosophila ventral furrow cells undergoing collective apical constriction are coordinated to fold the tissue. In Chapter 2, I investigate how discrete actomyosin contraction events are coordinated in time and between neighboring cells to yield tissue contraction and folding. I developed a computational pipeline to identify and classify contraction events from live images of ventral furrow formation. Using this framework, I found heterogeneity in contraction events, both in terms of contraction intensity as well as apical area behavior. I found that apical constricting cells transition in contractile behavior over time, from undergoing reversible contractions into a ratcheted state where contractions are irreversible. High expression of the transcription factor Twist is required for the transition into this irreversible, ratcheted state, which is associated with more neighboring contractions as well as cooperative interactions between neighbors. In Chapter 3, I examined how contractility is buffered against heterogeneity in cell apical area. I found that Cta-signaling is required to robustly maintain apical Factin cortex that can support contracting over larger apical distances. Without this buffering, apically larger cells progressively lose apical F-actin and exhibit delayed initiation of actomyosin contractions, leading to a lack of coordinated constriction. / by Shicong Xie. / Ph. D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2011. / Page 169 blank. Cataloged from PDF version of thesis. / Includes bibliographical references (p. 165-168). / Crystallization is a commonly used purification process in industrial practice. It usually begins with heterogeneous nucleation on a foreign surface. The complicated mechanism of heterogeneous nucleation is not well understood, but we hypothesize a possible correlation between binding affinity to a surface and nucleation enhancement. Amorphous polymers have been used in controlling crystallization. However, to our knowledge no attempt has been made to investigate the possibility of using binding affinity to help guide the selection of polymers promoting heterogeneous nucleation. This study investigated the possibility of using binding affinity of one molecule and many molecules to help guide the selection of these polymers. To measure the binding affinity of one molecule, we developed a two-step approach to compute the free energy of binding to a binding site, using a system of ethylene glycol, polyvinyl alcohol (PVA), and heavy water (D20). The first step of our approach uses Adsorption Locator to identify probable binding sites and molecular dynamics to screen for the best binding sites. The second step employs the Blue-Moon Ensemble method to compute the free energy of binding. We then applied our procedure to the systems of aspirin binding on the surfaces of four nonporous crosslinked polymers in ethanol-water 38 v%. These polymers are poly(4- acryloylmorpholine) (PAM), poly(2-carboxyethyl acrylate) (PCEA), poly(4-hydroxylbutyl acrylate) (PHBA), and polystyrene (PS), and they all are crosslinked with divinylbenzene (DVB). We developed an approach to construct these crosslinked polymers and built three independent surfaces for each polymer. We found the similarity between the trend of heterogeneous nucleation activity and that of the average free energies of binding to the best site of each polymer surface. To measure the binding affinity of many molecules, preferential interaction coefficient and the number of aspirin molecules associated with the area of the binding site was calculated. We found that there is also a similarity between the trend of heterogeneous nucleation activity and that of number of aspirin molecules associated with the area of the binding site (taken into account the effects of polar/apolar atom interactions between an aspirin and a polymer). These results suggest the possibility of using binding affinity, especially the free energy of binding to the best site and the number of nucleating molecule, to help guide the selection of polymers promoting heterogeneous nucleation. / by Surasak Chunsrivirot. / Ph.D.

Thesis: S.M., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2014. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 31-35). / Alternative splicing of mRNA transcripts is a significant step in the production of functioning protein. This process is a major source of molecular diversity, as numerous mRNA and protein products can arise from a single gene locus, and incorrect regulation has been implicated in numerous diseases. While many robust methods exist to study genome-wide single exon splicing patterns, no methodology has been established to accurately examine multiple events over a single isoform. Read sequencing technology has been the limiting factor; however, the recent development of real time, single molecule read sequencing provides an opportunity to characterize alternative splicing on the whole transcript level. We propose a computational approach to detect the splicing patterns of pairs of alternative exons in the same gene. Using a sequenced full-length cDNA library of human MCF-7 transcripts, we are able to evaluate 761 genes and identify three with evidence of non-random splicing of distinct nonadjacent alternative exons, all of which are frame-preserving and biased toward mutual inclusion. Characterizing their protein products reveals that the domain, secondary, and tertiary structures of the isoforms are not significantly affected. Low read coverage proves to be the greatest hindrance to a larger result set, but overall we provide a computational proof of concept for studying coordinated alternative splicing events on a transcriptomic scale. / by Asa Adadey. / S.M.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2012. / Cataloged from PDF version of thesis. / Includes bibliographical references. / Cells are constantly faced with the challenge of functioning reliably while being subject to unpredictable changes from within and outside. Here, I present two studies in which I analyze how biochemical circuits that regulate signaling and gene expression can generate robustness or phenotypic variability between otherwise identical yeast cells. Using the osmosensing signaling pathway which consists of a phosphorelay connected to a MAPK cascade, we predict signaling robustness to changes in kinetic rate constants by employing a computational sensitivity analysis. Consistent with the model predictions, we find that the input-output relation of signaling activation is severely impacted by protein coding sequence changes in the MAPK cascade genes, but not the phosphorelay genes. By decoupling the network into two separate modules, we show that an input-output analysis of each of the modules can generate the observed disparity in their tolerance to kinetic parameter variations. Our analysis suggests that the input-output relation of catalytic signaling pathways i.e. MAPK cascade are intrinsically sensitive to kinetic rate perturbations. By contrast, signaling governed by stoichiometric biochemical reactions i.e. phosphorelay exhibit robust input-output functions. We further find that cells challenged to alter their input-output function mostly recovered by gaining mutations in the MAPK cascade genes, which further supports our model. We next explore how HAC1 RNA splicing contributes to heterogeneity in the unfolded protein response (UPR). We adapt the single molecule FISH (sm-FISH) method to count endogenous spliced and unspliced HAC1 transcripts in single cells. We use a stochastic bursting-transcription-and-splicing model to determine the kinetic rates from the single cell measurements. We find that the cell-to-cell variability in the degree of splicing is tightly regulated in the presence of a UPR-inducing chemical agent, but is compromised under heat stress. By considering models including extrinsic noise at the splicing or transcriptional level, we show that the increased variability in the degree of splicing under heat stress can be generated by increased fluctuations in the splicing rate. Lastly, we present an approach using sm-FISH and protein synthesis inhibitors to measure translation and we show preliminary results suggesting its feasibility. / by Mei-Lyn Ong. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2011. / Cataloged from PDF version of thesis. / Includes bibliographical references. / Many organisms have evolved DNA damage response mechanisms to deal with the constant damage to DNA caused by endogenous and exogenous agents. These mechanisms activate cell cycle checkpoints to allow time for DNA repair or, in the case of severely damaged DNA, initiate cell death mechanisms to maintain genomic integrity. The cell's response to DNA damaging agents includes wide spread changes in the transcriptional state of the cell that have been implicated in cell death or survival decisions. However, we do not fully understand how the multiple and sometimes opposing transcriptional signals are interpreted to make these critical decisions. A computational and systems biology approach was taken to study the wide-spread transcriptional changes induced in human cell lines after exposure to a DNA damaging and chemotherapeutic agent, 1,3-bis-(2-chloroethyl)- 1 -nitrosourea (BCNU or carmustine). Cell lines with extreme sensitivity or resistance to BCNU were identified from a set of twenty four genetically diverse human lymphoblastoid cell lines using a high-throughput method that was developed as part of this thesis. This assay has broad applications and can be used to simultaneously screen multiple cell lines and drugs for accurate measurements of cell proliferation and survival after drug treatment. The assay has the advantage of having a large dynamic range that allows sensitivity measurements on a multi-log scale allowing better resolution of comparative sensitivities. Temporal transcription profiles were measured in cell lines with extreme BCNU sensitivity or resistance to generate a large transcription data set amenable to bioinformatics analysis. A transcriptional signature of 706 genes, differentially expressed between BCNU sensitive and resistant cell lines, was identified. Network and gene ontology enrichment identified these differentially expressed genes as being involved in key DNA damage response processes like apoptosis and mitosis. Experimental evidence showed that the transcription signature correlated with observed cellular phenotypes. Furthermore, the NF-Y transcription factor binding motif was enriched in the promoter region of 62 mitosis-related genes downregulated in BCNU sensitive but not resistant cell lines. Chromatin immunoprecipitation followed by sequencing (ChIP-seq) confirmed NF-Y occupancy in 54 of the 62 genes, thus implicating NF-Y as a possible regulator of the observed stalling of entry into mitosis. Using experimental and computational techniques we deciphered the functional importance of differential transcription between BCNU sensitive and resistant cell lines and identified NF-Y as an important factor in the transcriptional and phenotypic cell response to BCNU such as the control of entry into mitosis. / by Chandni Rajan Valiathan. / Ph.D.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Computational and Systems Biology Program, 2012. / Cataloged from PDF version of thesis. / Includes bibliographical references (p. 141-149). / The reconstruction of regulatory networks is one of the most challenging tasks in systems biology. Although some models for inferring regulatory networks can make useful predictions about the wiring and mechanisms of molecular interactions, these approaches are still limited and there is a strong need to develop increasingly universal and accurate approaches for network reconstruction. This problem is particularly challenging in mammals, due to the higher complexity of mammalian regulatory networks and limitations in experimental manipulation. In this thesis, I present three systematic approachs to reconstruct, analyse and refine models of gene regulation. In Chapter 1, I devise a method for deriving an observational model from temporal genomic profiles. I use it to choose targets for perturbation experiments in order to determine a network controlling the responses of mouse primary dendritic cells to stimulation with pathogen components. In Chapter 2, I introduce the algorithm Exigo, for identifying essential interactions in regulatory networks reconstructed from experimental data where regulators have been silenced, using a network reduction strategy. Exigo outperforms previous approaches on simulated data, uncovers the core network structure when applied to real networks derived from perturbation studies in mammals, and improves the performance of network inference methods. Lastly, I introduce in Chapter 3 an approach to learn a module network from multiple highthroughput assays. Analysis of a diffuse large B-cell lymphoma dataset identifies candidate regulator genes, microRNAs and copy number aberrations with biological, and possibly therapeutic, importance. / by Ana Paula Santos Botelho Oliveira Leite. / Ph.D.

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2016. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 119-121). / Cancer is a clonal evolutionary process. This results in complex clonal architecture and intratumoral heterogeneity in each patient. This also presents challenges for effective therapeutic intervention - with constant selective pressure to induce or select pre-existing resistant subclones toward drug resistance. Mathematical/computational modeling from population genetics, evolutionary dynamics, and engineering are being utilized to a greater extent in recent times to study tumor progression, intratumoral heterogeneity, drug resistance, and rational drug scheduling/combinations design. In this thesis we present several joint quantitative and experimental approaches for the rational design of drug combinations to tackle the issue of intratumoral heterogeneity and clonal evolution. Using a tractable experimental system with pre-defined tumor compositions, we derived computational approaches to rationally design drug combinations with the goal of minimizing a given heterogeneous tumor. We found that the best drug combinations can oftentimes be non-intuitive as they do not contain component drugs most effective for the individual subpopulations. This was the result of a need for combinatorial considerations on the effects of each drug on all subpopulations, hence at times leading to non-intuitive drug regimens. We validated our computational model predictions in vitro and in vivo in a preclinical model of Burkitt's lymphoma, with predictable evolutionary trajectories upon treatment. Next, we extended this methodology to study the effects of more complex tumor heterogeneity on combinatorial drug design, with similar conclusions. Sampling and statistical analyses over a range of tumor compositions can further inform effective drug combinations under some uncertainty in initial tumor heterogeneity. Moving beyond a model where we have control of initial tumor composition, we sought to examine collateral resistance and sensitivity during clonal evolution. Using a murine model of Ph+ acute lymphoblastic leukemia, we performed drug selection and pharmacological screen experiments. We observed important evolutionary processes of selection and drift in giving rise to resistance to clinically used BCR-ABL1 inhibitors. Remarkably, the resistant population also became hyper-sensitized to nonclassical BCR-ABL1 inhibitors at intermediate stages of the clonal evolution, in this so-called 'temporally collateral sensitivity'. Mathematical modeling and experimentation brought additional insight into the evolutionary dynamics and mechanism of action, with demonstrated in vivo efficacy. / by Boyang Zhao. / Ph. D.

Thesis: Ph. D., Massachusetts Institute of Technology, Computational and Systems Biology Program, 2015. / Cataloged from PDF version of thesis. / Includes bibliographical references (pages 114-124). / Protein aggregation underlies a number of human diseases. Most notably, it occurs widely in neurodegenerative diseases, including Alzheimer's and Parkinson's. At the molecular level, neurotoxicity is thought to originate from toxic gains of function in multimeric aggregates of proteins that are otherwise predominantly monomeric and disordered, fluctuating between a very large number of structurally dissimilar states on nano- and microsecond timescales. These proteins, termed Intrinsically Disordered Proteins (IDPs), are notoriously difficult to probe using traditional biophysical techniques. In order to obtain structural information pertaining to the aggregation of IDPs, it is often necessary to develop computational and modeling tools, both to leverage the full extent of the experimental data, and to generate testable predictions for future experiments. In this thesis, I present three separate computational studies studying the formation of multimeric aggregates in IDPs, spanning different aspects of the aggregation process, from early nucleation events to fibril elongation. In the first study, I present a conformational ensemble of a-synuclein, the culprit protein of Parkinson's disease, constructed using a Variational Bayesian Weighting algorithm in combination with NMR data collected by our collaborators. We find that the data fit a description in which the protein predominantly exists as a disordered monomer but contains small quantities of multimeric states containing both helical and strand-rich conformations. In the second study, I focus on the process of amyloid fibril elongation in the Amyloid-[beta] (A[beta]) peptide of Alzheimer's disease. I compute the free energy surface associated with the fibril elongation reaction, and find that elongation of both A[beta]40 and A[beta]42 experimental fibril structures occurs on a downhill free energy pathway, proceeding via an obligate, fibril-associated hairpin intermediate. The fibril-associated hairpin is significantly more stable (relative to the fibrillar, elongated state) in A[beta]42 compared with A[beta]40, suggesting a potential clinical target of interest. Finally, I present lengthy, all-atom molecular simulations that suggest that nucleation of the minimum aggregating fragment of c-synuclein proceeds via a helical intermediate, requiring a structural conversion into a strand-rich nucleating species via a stochastic process of individual helices unfolding and self-associating via backbone hydrogen bonds. / by Thomas Gurry. / 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. / High-throughput "omics" data are becoming commonplace in biological research and can provide important translational insights, but there is a need for well-crafted user-friendly tools for integrating and analyzing these data. In this thesis, I present versions 1 and 2 of Omics Integrator, a software tool designed to take advantage of the Prize-Collecting Steiner Forest algorithm from graph theory to provide users with high-confidence biological networks informed by their omics results. I show the results of using this flexible tool in several studies of Huntington's disease (HD), a fatal neurodegenerative disorder with no cure. By leveraging Omics Integrator on omics datasets from induced pluripotent stem cell (iPSC) derived models of HD, I discovered and highlighted several pathways that are altered in these cell line models, including neurodevelopment and glycolytic metabolism, which may lead to important therapeutic targets in the disease. Finally, I compare omics data derived from three iPSC-derived models differentiated towards a striatal neuron cell type using different protocols, and show that by performing this large comparative analysis I can implicate functions and pathways common to several models of HD. Future integrative and comparative studies like these will be made easier by the Omics Integrator tool. / by Amanda Joy Kedaigle. / Ph. D.