22689 research outputs found
Sort by
DESIGN, FABRICATION, AND EVALUATION OF FOLDABLE MICROELECTRODE ARRAYS: ELECTRODE PERFORMANCE, UV-TUNABLE POLYMER OPTIMIZATION, AND NEURAL STIMULATION IN BRAIN ORGANOIDS
No full textNeural organoids (NOs) have emerged as important tissue engineering models for neuroengineering and biocomputing. Establishing reliable relationships between stimulation and recording traces of electrical activity is essential to monitor the functionality of NOs, especially as it relates to realizing biocomputing paradigms such as reinforcement learning or stimulus discrimination. While researchers have demonstrated neuromodulation in NOs, they have largely used 2D microelectrode arrays (MEAs) with limited access to the entire contour of the 3D NOs. Here, we report neuromodulation using tiny mimics of macroscale EEG caps or shell MEAs. Specifically, we observe that stimulating current within a specific range (20 to 30 µA) induced a statistically significant increase in neuron firing rate when comparing the activity five seconds before and after stimulation. Such neuromodulatory behavior was observed using both three and 16-electrode shells
Mechanisms of Genetic Regulation of Transcription Through Multi-omics and Single Molecule Imaging
No full textTranscription of mRNA is a highly regulated process, involving a coordinated network of transcription factors recognizing DNA sequence motifs located in promoters and enhancers throughout the human genome, and a host of other regulatory proteins including ATP-dependent chromatin remodelers to restructure nucleosomes, histone modifiers, the general transcription factors, RNA Polymerase II enzyme and the Mediator complex which forms condensates to compartmentalize transcriptional machinery and facilitate transcriptional initiation and elongation. Transcriptome analysis from large human populations have identified genetic variants associated with steady-state mRNA levels, which show strong enrichment in diverse human traits and have provided important insights into the genetic architecture of complex diseases. However, existing studies conducted on bulk transcriptomes often fail to account for the high heterogeneity of human tissues. Moreover, there are limited studies which comprehensively explore the relationship between transcription, epigenetic states, and protein expression from diverse cell types and from individuals with diverse genetic ancestries. Finally, the unique roles of “pioneer” transcription factors in creating accessible chromatin primed for productive transcription have been underexplored in genomic studies. The mechanisms of pioneer factor engagement with DNA in different cellular environments and the implications in development and disease are similarly not completely understood.
Here, I leveraged multi-omic data collected from diverse human individuals in bulk tissues and single cells to systematically prioritize genetic variants associated with transcription, chromatin accessibility, methylation, and protein expression levels. I developed powerful and flexible machine learning frameworks to prioritize functional rare variants and performed extensive statistical analysis of molecular quantitative trait loci (QTL) to investigate context-specific effects of genetic regulation of transcription and its propagation across the regulatory cascade. I systematically analyzed human transcription factors in diverse tissues and revealed important regulatory functions of pioneer factors in shaping inter-individual variability of epigenetic states. Finally, for a prototypical human pioneer factor, NF-Y, I performed single molecule imaging in live cells followed by statistical characterization of its diffusive states to investigate generalizable target search mechanisms. Overall, this thesis provides new computational models and mechanistic insights into the genetic regulation of transcription and their important implications in human health and disease
ADMINISTRATIVE UNIT PROLIFERATION AND INTERCOMMUNAL CONFLICT IN SOUTH SUDAN
No full textThe central objective of this research is to examine the relationship between South Sudan’s recent
administrative redistricting and intercommunal conflict and the nature of that relationship.
Contrary to the more benign and even positive effects of decentralization present in the existing
literature, this research finds that boundary changes have been an important driver of instability
during the period under study (2014-2018) and have had a distinct, leading role in shaping and
driving conflict. This research finds that decentralization is not an inherent “good” and when
instrumentalized by a self-serving regime, unit proliferation can be a powerful “bad” in aggravating
inequalities, exclusion and conflict.
Via interview data, ACLED data and content analysis, this research finds support for the hypothesis
that South Sudan’s administrative redistricting increased and/or exacerbated intercommunal
conflict via a multi-step process. In this process, the ethnically-informed proliferation of states in
2015 led to increased scarcity of key political and physical resources, particularly in opposition
areas. This scarcity led to increased perceptions of insecurity and a perceived need to protect
against threats. In many cases this led to the arming, training and/or mobilization of local defense
forces and militia which ultimately contributed to a growing number of identity militia as well as an
increased rate of intercommunal conflict over time (a 67% increase between the initial 10 state
period and final 32 state period).
This research does not advance a monocausal argument to explain the increase in intercommunal
conflict in the targeted time period, nor does it find support for a singular mechanism explaining
boundary proliferation-related conflict. This research identifies multiple mechanisms which have contributed to increased insecurity including scarcity, uncertainty, threat perception and identity
mobilization. Drivers of conflict including resource scarcity, threat perception and ethnic solidarity
as well as opportunities including weak institutions, a lack of security forces/justice mechanisms
and porous boundaries, are identified as key elements of the mechanisms at play. This research
does not assign supremacy to one or a combination of these mechanisms and instead finds support
for the presence and utility of multiple mechanisms as explanations for the rising rates of
intercommunal violence
Ollier Disease and Maffucci Syndrome: Investigating the Function of Hif-1α Variants And Other Candidate Cancer-Related Gene Variants
No full textEnchondromas are benign cartilaginous tumors that arise from the medulla of bones. Patients with Ollier disease (OD, OMIM 166000) and Maffucci syndrome (MS, OMIM 6145692) have at least three enchondromas. In addition to enchondromas, patients with MS have vascular anomalies. OD and MS are cancer susceptibility syndromes, and approximately 50% of the patients develop a malignancy. Approximately 80% of the patients with OD and MS have a somatic, heterozygous, gain-of-function variant in IDH1 and IDH2 present in enchondromas, vascular anomalies, and chondrosarcomas. In addition, about 25% of the patients with OD and MS have a causative germline and early post-zygotic variants in one of seven HIF-1 pathway-related genes.
Here, I focus on a patient with a HIF-1α-p.Gln355His germline variant and a somatic IDH1-p.Arg132Cys variant identified in their enchondroma. To characterize the impact of the HIF-1α-p.Gln355His variant, I created a HIF-1α knockout HEK293T cell line to express a transient plasmid with the HIF-1α-p.Gln355His variant. Next, I investigated the impact of the co-occurrence of the IDH1-p.Arg132Cys variant on HIF-1α mRNA levels and HIF-1α transcriptional activity in the patient’s enchondroma cell line. HIF-1α mRNA levels remained the same in the IDH1-p.Arg132Cys positive enchondroma and the unaffected cartilage from the patient under normoxia and hypoxia; however, the IDH1-p.Arg132Cys positive enchondroma had a statistically significant increase in CA9 mRNA, suggesting an increase in HIF-1α transcriptional activity, under hypoxia compared to the unaffected cartilage from the patient.
Since approximately 75% of patients with OD and MS did not have a causative variant in one of the seven HIF-1-targeted genes, I performed a burden analysis focusing on cancer-associated genes to determine if any of these genes were more mutated among the patients compared to the controls. The cancer burden analysis identified two strong candidate causative genes, TCF3 and FGFR1, with a false discovery rate <0.05. We will further investigate the function of these variants to determine their impact in the development of enchondromas and vascular anomalies in patients with OD and MS
On Iteratively Reweighted Least Squares for Time Series Denoising
Full text linkThis thesis explores applications of structured-low rank approximation to the problem of time series denoising. Although some other methods will be covered, the main focus will be the development of an iterative-reweighted least squares (IRLS) algorithm to solve this problem via the minimization of a surrogate function for the rank.
A large part of the work contained in this thesis revolves around making the IRLS algorithm feasible for large scale data. Although the algorithm itself was already developed by other members of this group, the author of this thesis implemented optimized techniques for matrix operations to improve runtime by significant factors. Additionally, refined stopping criteria which promote solutions more desirable for our setting were explored and implemented.
The algorithm and implementation details provided in this thesis were developed with the intent of being applied to data generated by the Laser Interferometer Space Antenna (LISA). The gravitational waves detected by this satellite can be well approximated by sums of sinusoids, and as such our denoising analysis is restricted to this setting. We outline a procedure for denoising data of this form and show early analysis of its performance on synthetic data relative to baseline methods
Mainstreaming Personal Agency in Global Development: A roadmap for funders, implementers, and researchers to understand and leverage personal agency approaches for human flourishing
No full textThis roadmap provides background on the concept of personal agency and its need in development, and outlines agency’s potential, key challenges, and how agency-based approaches work. It draws on a growing body of evidence, including lessons from the Self-Empowerment and Equity for Change (SEE Change) Initiative, and actionable recommendations from
social psychologists and development economists around the world
TF-BindNet: A pilot study to decipher the uncharacterized binding preferences of transcription factors using a neural network-based algorithm
No full textTranscription factors (TFs) are a class of proteins defined by their ability to bind specific DNA sequences and thereby regulate transcription. Technologies such as protein binding microarray and chromatin immunoprecipitation sequencing have revealed the binding preferences of many TFs in mammalian genomes. These binding profiles enable applications like lab-based cell differentiation and gene regulatory network construction. However, the binding preferences of approximately 25% of the human TFs remain uncharacterized due to experimental limitations. The develpment of deep learning provides a novel tool to study protein, making it plausible to predict the binding preference of these uncharacterized TFs based on amino acid sequence. In this study, we build a neural network-based model to identify the strongly bound DNA 8-mers, and tested it on unseen TFs. We collected most of the published protein-binding microarray data in the train set, which consist of approximately 2000 transcription factors and over 30,000 8-mer DNA probes. To enhance the protein representation, we applied multiple encoders, including one-hot, protein language embedding and protein positional encoding. In this pilot study, we studied how can models learn the latent DNA-binding mechanisms of the TFs, and evaluated the model's generalizability on unseen TFs. These preliminary findings provide insight to our overall goal of identifying binding preferences as motifs of uncharacterized TFs in future large-scale studies
THE MILLENNIAL FISH A NOVEL: FIRST TWELVE CHAPTERS
No full textThe Millennial Fish is a coming-of-age novel that follows a young college student named Jonty, who tragically loses his brother in an accident. Struggling with an existential crisis and contemplating suicide, Jonty embarks on a road trip across the country with a recently dumped Hawaiian man named Hinano. Along the way, the people they meet and the experiences they share help redefine their lives and their ability to move forward
ACCELERATING THE EXPLORATION OF CARBOXYLIC ACID-ZEOLITE INTERACTIONS: USING CLASSICAL AND MACHINE LEARNING INTERATOMIC POTENTIALS
Full text linkThe upgrading of carboxylic acids over zeolitic catalysts is a promising path to a sustainable future. In this study, we systematically investigated the interactions between various carboxylic acids and Brønsted acid sites within zeolite frameworks. A combination of classical interatomic potentials simulations, machine learning interatomic potentials (MLIPs), and density functional theory (DFT) calculations was employed to predict adsorption configurations and energies. MLIPs trained using NequIP improved accuracy, although challenges remained in transferability. This work highlights both the potential and current limitations of computational modeling in catalyst design, suggesting future improvements through more transferable ML models
Mechanisms of growth abnormalities in two germline disorders of H3K27me3
No full textMendelian disorders of the epigenetic machinery (MDEMs) result from pathogenic germline variants in genes encoding writers, erasers, and readers of histone marks. This emerging class of rare disorders shares features such as growth abnormalities and intellectual disability or developmental delay; however, the molecular mechanisms underlying these common characteristics remain elusive. Here we investigate two disorders centered upon H3K27me3. Weaver syndrome (WS) is caused by variants in EZH2, a H3K27me3 methyltransferase (writer), whereas Kabuki syndrome 2 (KS2) is caused by variants in KDM6A, a demethylase (eraser) targeting the same mark.
We first modeled the skeletal overgrowth of WS in transgenic mice bearing the most common patient variant, EZH2 p.R684C. Compared to Ezh2 +/+ littermates, Ezh2 R684C/+ mice exhibited greater femoral cross-sectional tissue area, and increased mineral apposition rate (MAR) in vivo. Since MAR is a measure of osteoblast activity, we differentiated primary bone marrow mesenchymal stem cells (BM-MSCs) towards osteoblasts. Ezh2 R684C/+ cells indeed showed enhanced osteogenic potential by Alizarin red staining. We next performed RNA-seq, which indicated that Ezh2 R684C/+ cells had a collective dysregulation of osteoblast differentiation genes, as well as the BMP (bone morphogenetic protein) pathway. Finally, we demonstrated that an inhibitor targeting the opposing erasers of H3K27me3, KDM6A and KDM6B, was able to substantially reverse the effects of the R684C allele at both the phenotypic and transcriptional levels.
In contrast to WS, KS2 manifests with growth retardation. We described skeletal growth retardation in a mouse model of KS2, Kdm6a tm1d/+, which had shorter femurs and tibias than Kdm6a +/+ littermates as well as other abnormalities in bone parameters. Furthermore, Kdm6a tm1d/+ growth plates were shorter due to decreased hypertrophic chondrocyte size. To further characterize the disruption to chondrocyte function, we generated Kdm6a -/- ATDC5 cell lines and showed that Kdm6a -/- cells exhibited precocious differentiation. This was reminiscent of prior findings in a related disorder, Kabuki syndrome 1 (KS1), which shares a clinical phenotype with KS2 despite separate genetic etiologies. RNA-seq on both KS1 and KS2 chondrocytes revealed a similar transcriptomic signature between the disorders, shedding light on the common pathways between two highly similar MDEMs