80264 research outputs found
Sort by
COF and Effect: Covalent Organic Framework Applications Enabled by Scalable Synthesis
No full textCovalent Organic Frameworks (COFs) are promising materials due to their high porosity, tunable functionality, and ordered architecture, making them ideal for applications in catalysis, electronics, and environmental remediation, however insufficiencies in scalable synthesis limits their practical use. This work presents a scalable and rapid chemical vapor deposition (CVD) method to synthesize large-area, crystalline COF films, as well as two potential applications of COFs, both as organic semiconductors and photocatalytic adsorbents for PFAS remediation. Utilizing co-evaporation of monomers, we achieved high-quality, oriented COF films on Si/SiO₂ substrates within 20 minutes, producing films with hydrazone, imine, and ketoenamine linkages. Additionally, we synthesized nitrogenated holey graphene (C2N-h2D) films using CVD, demonstrating their semiconducting properties through comprehensive characterization, including Reflection Electron Energy Loss Spectroscopy (REELS), Ultraviolet Photoemission Spectroscopy (UPS), and Valence Band Spectroscopy (VBS). The synthesized C2N-h2D films exhibited smooth topology, high crystallinity, and an estimated bandgap of ~2.4 eV, highlighting their potential in electronic applications. Furthermore, we developed a novel COF-based method for the adsorption and photocatalytic degradation of per- and polyfluoroalkyl substances (PFAS) using a packed-bed column followed by UV-C irradiation named the “Texas Two-Step” method. This process enables efficient removal and mineralization of multiple EPA regulated PFAS. The CVD approach not only enhances film quality and scalability but also offers a promising route for the integration of COF materials into advanced functional devices and environmental remediation technologies
Efficient and Scalable Network Simulation Using GPUs
Full text linkNetwork simulation using Discrete Event Simulation (DES) tools is an effective way for network designers to study the effects and tradeoffs of various network designs with high-fidelity experimental results. However, current network simulators mostly target CPU platforms, and we observe these simulators do not scale well, thereby making it difficult for network designers to obtain insights at scale.
In this thesis, we discuss the design space for efficient network DES on GPUs. First, we carefully map the DES algorithm to the GPU hardware by leveraging memory coalesced data structures to optimize for high throughput operations. These data structures ensure GPU compute units fully utilize DRAM bandwidth. Second, we design a synchronization mechanism that allows the GPU to process events for many network components in parallel without CPU intervention during simulation execution. This helps the simulator more fully saturate GPU compute units and avoid CPU overheads. Third, we design a fully decentralized algorithm that allows individual components to process events without the need to pull events from a global priority queue, a key bottleneck in existing DES simulators.
We implement and evaluate nsX, a CUDA-based GPU network simulator built using DES principles. We demonstrate that nsX achieves three orders of magnitude greater simulation throughput than state-of-the-art CPU simulators at large scale, unlocking the ability to accurately simulate networks at the scale of modern data centers
A Way of Building (A Way of Life)
No full textThis thesis proposal, “A Way of Building (A Way of Life),” explores contemporary construction practices through the lens of collectivity, craft, and care, using the timber-framed barn as a symbol of enduring communal values. The project reimagines these barns, not as nostalgic relics, but as adaptable spaces for a new era where agriculture is no longer the primary function. By emphasizing material salvage and reuse, the proposal challenges conventional ideas of restoration and aims to transform these barns into vital community hubs.
Through interviews and site visits, the project examines the importance of salvaged materials in the local context, with a particular focus on the reuse of materials from those deteriorating barns. A catalog documents the process of material sourcing and assembly, highlighting the potential of these materials to form new architectural interventions. The proposal advocates for a flexible, adaptable design approach, where salvaged elements—such as timber, stone, and corrugated metal—are repurposed into functional spaces for workshops, markets, and community events.
Ultimately, the thesis proposes a model for reviving the barn typology, not just as an architectural form but as a means of fostering community resilience, hands-on craftsmanship, and sustainable building practices. This approach offers a new framework for communities facing similar challenges, combining preservation with innovation to create spaces of collective care
Buy-Side Agency Algorithms: Institutional Investor Trades and Market Quality around SEC Filings
Full text linkIn modern markets, institutional investors employ agency algorithms that strategically subdivide large orders to minimize the price-impact and signaling of their trades. As a consequence, the daily level of institutional market activity has become increasingly difficult to identify. In this paper, I exploit an intraday artifact caused by the repeated order submission of agency algorithms – Clock-Time Periodicity – to quantify their use and illuminate the trading behavior of the institutional investors that employ them. I show that agency algorithms are utilized on a daily basis and account for an average of 3.1% of trade volume. I show that the arrival of historically informative SEC filings increases the volume of institutional trading. Earnings Announcements and Regulation Fair Disclosure filings have the largest impact, with an increase in institutional trade volume of 13.9% and 2.56% respectively. I present evidence supporting two plausible motivations for the increased activity: a) Urgency and b) Liquidity. First, the increase in institutional volume is contrasted by a stark decrease in the number of limit-orders used to achieve it. An indication that agency algorithms not only trade more but do so more aggressively. Second, on filing dates an increase in institutional volume worsens trade conditions with magnitudes that rival the effect of the news itself. Lastly, the increase in the size of institutional trades is overshadowed by an even larger increase in the size of non-institutional trades. This suggests that the trading behavior of non-institutional traders provides a ‘smokescreen’ for agency algorithms to trade more aggressively while remaining stealthy
Relational Computation for Very Large-Scale Machine Learning
Full text linkIn mathematics, a tensor is an algebraic object that describes multilinear rela-
tionships among sets of algebraic entities associated with a vector space. From a
computational perspective, tensors are commonly represented as multi-dimensional
arrays—a format that plays a central role in machine learning. A widely used convention for expressing tensor operations is Einstein summation notation (EinSum), which
compactly encodes summation over indexed terms. This notational framework not
only streamlines the expression of complex tensor computations but also lends itself
to an alternative interpretation: a multi-dimensional array can be viewed as a map-
ping from a vector of integers (i.e., a primary key) to a real number. This perspective
aligns closely with the classical definition of a relational database relation. As a result, many numerical and machine learning computations in tensor calculus can be
reformulated as sequences of joins and aggregations over relational data. Executing
these computations within a relational database system offers several key advantages,
including automatic parallelization, distribution, and scalability. Moreover, relational
databases are particularly effective at handling sparsity, as they are designed to effi-
ciently represent and process cases where only a small subset of the possible primary
keys actually occur in the relation.
In this thesis, I propose an extension to Einstein notation called Upper-Case-
Lower-Case Einstein Notation—a simple yet expressive framework for describing tensor programs that interleave operations over sparse (relational) data with efficient
kernel calls over dense tensors. This notation enables the concise representation of
computations optimized for complex sparsity patterns. To support this notation, I
develop a compiler, SparseEinSum, which takes standard EinSum expressions as
input, transforms them into extended Upper-Case-Lower-Case Einstein Notation as
intermediate representation, and compiles them into tensor-relational algebra. The
compiler incorporates sparsity estimation and cost-based schema selection to guide
the transformation. The resulting programs can be executed on virtually any relational database system, leveraging arrays to manage dense tensors within a relational
execution model. Experiments across tensor computation benchmarks demonstrate
that the generated tensor-relational computations offer significant performance improvements.
To support automatic differentiation of relational computation compiled from
EinSum, I derive key rules that enable automatic differentiation for relational algebra.
I introduce functional relational algebra to build functions in the relational domain
and define relational analogs of partial derivatives, Jacobians, gradients, and a set
of relation-Jacobian product rules for core relational operators, including table scan,
selection, aggregation, and join. This functional framework builds the foundation for
differentiation in relational algebra. Then, I propose a relational algebra automatic
differentiation algorithm using an efficient, correctness-preserving implementation of
the relation-Jacobian product. Through extensive experiments, I show that executing
machine learning computations on top of a relational engine—augmented with relational algebra automatic differentiation algorithm—can scale efficiently to very large
datasets. The resulting system achieves performance competitive with specialized
distributed machine learning systems, while retaining the advantages of relational
query optimization
Computational Methods for Delineating Homology Groups in Bacterial Genomes
No full textAssuming a gene to be the basic evolutionary unit has been commonplace in bacterial genomics. For instance, when quantifying horizontal gene transfer, it is common practice to infer gene trees and reconcile them with species trees to account for recombination-based processes. Phylogenomic studies typically assume that each locus is homologous across its entire length. However, evolutionary processes such as horizontal gene transfer, gene gain, and gene loss can disrupt this homology by recombining only parts of genes, leading to gene fission or fusion. Traditional methods for identifying conserved loci in comparative genomics, whether through existing protein annotations or de novo whole-genome alignments, are vulnerable to these genetic events, potentially compromising this fundamental assumption.
This thesis makes three key contributions. First, I examine the limitations of using conserved marker genes as loci and propose a new concept—Maximal Homologous Group (MHG), defined as a maximal set of maximum-length sequences that are homologous across the entire sequence alignment. A graph-based algorithm is designed and implemented in an open-source package, MHG-Finder, to accurately identify these units directly from raw DNA sequences, eliminating the need for protein annotations. Second, to enhance efficiency in handling large datasets and to provide detailed insights into intermediate MHGs involving subsets of input genomes, I introduce a new method utilizing a guide tree to significantly improve scalability and provide more informative biological results, implemented as tMHG-Finder. I compare tMHG-Finder with state-of-the-art whole-genome aligners on simulated data, showcasing tMHG-Finder’s superior accuracy. Finally, by applying tMHG-Finder to a phylum of extreme-environment-resistant bacteria, I demonstrate its utility in revealing evolutionary information beyond MHGs and the phylogenetic tree.
As sequencing technology rapidly advances in throughput and availability, it is crucial to conduct bacterial phylogenomic studies not only to understand the evolutionary process of different bacteria but also to guide downstream analyses. This thesis defines the basic unit for such studies and offers computational methods to identify this unit, providing valuable insights for future research in bacterial evolution
Evolution-Inspired Design of Bacterial Biosensors
No full textBacterial signal transduction systems are critical tools for synthetic biology. However, while hundreds of thousands of systems have been identified in bacterial
genomes, only a small subset have been characterized. In this work, we develop
tools for scalable interrogation of bacterial signal transduction mechanisms at three key junctures: sensing, processing, and actuation. In the first portion of this work, we identify 13 novel antimicrobial peptide (AMP)-activators of the S. Typhimurium PhoPQ two component system (TCS). We characterize the activation profiles of a subset of these AMPs in S. Typhimurium and E. coli using a variety of PhoPQ responsive promoters. We show that PhoPQ homologs from extraintestinal pathogenic E. coli and K. pneumoniae, which occupy in vivo niches, exhibit distinct activation profiles, providing new insights into the specificities, mechanisms, and evolutionary dynamics of TCS-mediated peptide sensing in bacteria. In the second portion, we propose a novel workflow for the prediction of transcription factor (TF)
operator sites based on a modified phylogenetic footprinting framework. We increase operator prediction accuracy from 45% to 83% when compared to leading methods and use this framework to predict operator sites for thousands of bacterial TFs. We then use these predictions to interrogate structural features that confer DNA-binding specificity to the TetR family of TFs. Finally, in the third portion of this work, we develop new methods for analyzing signal transduction in bacterial TCSs. We characterize interactions between membrane-bound sensor histidine kinases (SKs) and their cognate response regulators (RRs) using the protein language model ESM2. We show that the model is able to identify highly coevolving residues at the SK-RR interface. We then train a dedicated pairing model to predict TCS interactions from their amino acid sequences. We experimentally test our models with the E. coli SK BasS and the S. PCC6803 RR CcaR. Using model-guided mutagenesis, we successfully engineer signal transduction between BasS and CcaR in vivo. As a whole, this work provides a framework for characterizing signal input, processing, and response of bacterial signal transduction systems at scale
Housing Quality Registry: Preliminary Data and Results on Housing Challenges Facing the Houston Area
No full textDespite its implications, there is relatively limited data about housing quality in the Houston area. To address this need, the Houston Population Research Center at the Kinder Institute for Urban Research at Rice University launched a first-of-its-kind study on the conditions of homes in the region. The study is ongoing and will continue to collect and expand the data available on this topic. This report provides an overview of information learned from almost 1,000 Houston-area respondents who have already participated in the study, and it provides preliminary evidence of housing quality conditions that will be updated and refined as further data collection and future research explore the issue. This initial report will focus on issues with the conditions and functionality of people’s homes, experiences with flooding and water damage, problems with pests in the home, and neighborhood safety and resources. Finally, the brief ends with a look at what people like about where they live, because even when considering the challenges people face regarding their housing quality, there are still aspects of their living space they like and appreciate
Estimate the Mechanisms of Helicase-Polymerase Coupling in the Organelle DNA Replication
No full textAll living organisms must duplicate their genomic DNA before cell division, and the replication process relies on a range of proteins centered around the cooperative actions of DNA Helicases (Hel) for separating the DNA strands and DNA Polymerases (Pol) for synthesizing new strands of DNA. This thesis estimates the structural and mechanistic coordination between helicases and DNA polymerases (Hel–Pol) with minimal organelle replisomes as model systems.
Recent structural studies of the T7 replisome reveal a T-shaped configuration formed by helicase and polymerase at the replication fork. In our study, the T7 Pol will utilize a positively charged cleft and a β-hairpin loop structure to stabilize its DNA binding when working with T7 Hel. We constructed T7 Pol variants with mutations in the β-hairpin loop and the positively charged cleft to evaluate their roles in replisome function. Mutations in both motifs of the T7 Pol significantly reduced strand-displacement synthesis and rolling-circle amplification. These findings support a model in which T7 polymerase plays a dominant role in coordinating Hel–Pol interactions and driving replisome progression.
Using techniques combining negative stain EM and biochemical assays, I compared the differences of Hel-Pol coupling in mitochondria and T7 replisomes. In mitochondria, Polγ and Twinkle helicase interact dynamically, and the C-terminal tail of Twinkle contributes to replisome stability through its DNA-binding function. Deletion of Twinkle’s C-terminal tail reduces DNA-binding affinity and duplex unwinding, impairing Twinkle-Polγ DNA synthesis and fork progression.
In the apicoplast, a unique organelle in the malaria parasite, apPol—a multitasking A-family DNA polymerase—has evolved to perform both high-fidelity replication and exonuclease proofreading. Structural data indicate that apPol exhibits significant movements in the finger domain and thumb domain between the binary and ternary complex.
Altogether, this work defines operational principles that establish Hel–Pol coordination in reduced organelle replisomes, providing a mechanistic basis for replication fidelity and processivity that informs both mitochondrial disease models and antimalarial drug strategies