Treasures @ UT Dallas
Not a member yet
7697 research outputs found
Sort by
Molecular Dynamics Simulations of Photoisomerizable Azo-phosphatidylcholine Lipid Bilayers
No full textLiposomes are phospholipid-based vesicles commonly used in drug delivery applications.
When modified with certain components, liposomes become sensitive to particular stimuli
such as heat or light, and release their contents upon stimulation. One particular class of
light-sensitive liposomes are azosomes, whose light-sensitivity is imbued by the incorporation
of a photoisomerizable phospholipid azo-PC. Azo-PC is a distearoylphosphatidylcholine
(DSPC) derivative containing an azobenzene moiety in one of the lipid tails and has the
ability to photoisomerize between cis-azo-PC and trans-azo-PC, corresponding to the isomeric
state of azobenzene. When the azosome is irradiated with UV light, trans-azo-PC isomerizes
to cis-azo-PC, the area per lipid increases more than 30 % and the azosome bilayer thickness
decreases by 20 %. This results in an increase in azosome diameter and leakiness and
ultimately a release of its encapsulated payload. Unique to photoisomerizable liposomes,
subsequent irradiation with blue light causes the reverse cis-to-trans isomerization to occur.
Consequently, the azosome diameter decreases, and payload release is arrested. The use of
light enables on-off payload release with spatiotemporal resolution to a high degree. This work
uses molecular dynamics simulations of pure azo-PC bilayers to explore the azosome structural
(area per lipid, bilayer thickness, NMR order parameter, and void volume) and mechanical
(area expansion modulus) changes that accompany azo-PC isomerization. Furthermore, this
work uses steered molecular dynamics simulations and constrained molecular dynamics to
determine the free energy profile of the translocation of a model drug (phenol) across a pure
azo-PC bilayer, and proposes a mechanism of release. Moreover, preliminary development of
an azo-PC model at the coarse-grained level compatible with the SPICA (surface property
fitting coarse graining) force field is presented. The work presented can be used as a foundation
for further studies into photoisomerizable liposomes for drug delivery
Grin and Bare It: Let Me Cry About White Supremacy
No full textThis MFA thesis is an articulation of a diptych of sculptures titled “Grin and Bare It.” Each
sculpture captures a moment of mutilation in my Black body where I disfigure, impale, and
display myself to survive my surroundings in two historically white spaces: the academy and the
gallery. Made from primarily wood and metal, the diptych presents a state grounded in
performance and another that suggests the backstage of one’s life. The vacillation between the
two sculptures, Talking out the side of my neck and Pins and needles, simulate my lived
experience in these spaces utilizing affect theory—a field that attempts to grasp at emotional
phenomena that occurs in the body before the mind can rationalize them—and the concept of
“minor feelings.
Essays in Health Economics
No full textThis dissertation concerns health economics, and the use of non-stationary data in empirical
study. The first chapter examines disability rates in the United States over time. Despite
general improvements in health, advances in medical technology, and the passage of the
Americans with Disabilities Act, the percentage of working-age adults receiving Social Security Disability Insurance (SSDI) increased nearly 150% from 1986 to 2013. In 2013, nearly
one in twenty working-age adults was out of the labor force and receiving SSDI. After 2013,
however, the increasing trend reversed, and the number of SSDI recipients has been persistently falling since. Given the non-stationary nature of the long-run trend in SSDI rates, I
use cointegration regression analysis and find that the percentage of the population ages 55
and 59 drives the long-run trend in national SSDI rates. This result allows me to forecast
future national and state SSDI rates. State SSDI rates follow the same common national
trend, but there is a large degree of idiosyncratic variation. In 2013, state SSDI rates ranged
from 2.75% to 8.60%. The results show median household income, age demographics, and
application rates explain 84.5% of the variation in SSDI rates between states.
In Chapter 2 I note pockets of counties in the United States with extraordinarily high
disability rates; in some such counties more than one in five working-age adults is disabled and
out of the labor force. I show that labor market conditions impact SSDI claiming behavior.
Specifically, I use a constructed instrumental variable regression to show that exposure to
negative county-level employment shocks leads to growth in future SSDI rates in that county.
I find that a 10% decrease in the employment growth rate leads to a 12.4% increase in SSDI
growth two years later. These results indicate SSDI not only explicitly insures workers
against disability, but also implicitly insures workers, particularly lower skilled workers,
against employment loss resulting from declining demand in shrinking industries.
Finally, in Chapter 3 my coauthors Peter C.B. Phillips and Donggyu Sul and I explore the
impact of Covid-19 vaccinations and mandates on state cumulative vaccination rates. Due to
the non-stationary nature of cumulative vaccination rates, the popular two-way fixed effects
(TWFE) regression may be problematic. We first show the pitfalls in two-way fixed effects
(TWFE) regressions when the outcome variables contain nonlinear and non-stationary trend
components. If a policy change shifts trend paths of outcome variables TWFE estimation
can distort results and invalidate inference. A robust solution is proposed by allowing for
dynamic club membership empirically using a relative convergence test procedure. The long-
run impact of a policy can thus be examined via its impact on convergence club membership.
We then apply this proposed dynamic convergence clustering to analyze Covid-19 vaccination incentives and mandates. We create a new weekly database to track individual state
vaccination policies and mandates over the period from March 2021 to February 2022. The
results reveal that Federal-level vaccine mandate announcements produced a merger of state
vaccination rates into a single convergence cluster
S-Adenosylmethionine as a Label-Free Probe for the Detection of Non-Covalent Interactions in the Active Site of Methyltransferase Enzymes
No full textS-adenosyl-L-methionine (AdoMet, or SAM) is a small biomolecule co-substrate used by enzymes
as a methyl donor in methylation reactions and applied in various biochemical processes.
Understanding how SAM interacts with enzyme active sites and the sub-molecular mechanistic
details of how noncovalent interactions increase the molecule’s reactivity and catalysis is highly
interesting across numerous biochemical and drug discovery communities. This interest arises
from the need for further physical insight into chemical catalysis and the role of many SAM-
binding proteins in human diseases such as cancer.
This study focuses on developing the ground vibrational states of SAM when bound to a well-
studied protein lysine methyltransferase, SET7/9, as detectors of noncovalent interactions in
solution that the molecule experiences due to the active site environment. While SAM binding to
SET7/9 is well studied using crystallographic methods, it is essential to have solution state
information to properly characterize the noncovalent interaction. Solid-state NMR spectroscopy
has also been conducted to detect unusual noncovalent interactions, such as the carbon tetrel
interaction to the reactive methyl group in high-resolution crystalline material, further orthogonal
physical evidence is necessary to validate the hypothesis of these interactions contributing to
binding. Moreover, frequencies observed in the SAM vibrational spectrum when bound within
active sites can provide experimental constraints for binding and kinetic isotope effect studies.
Toward this goal, frequencies measured in SAM’s infrared and Raman spectra must be assigned
to the motions of specific atoms via isotope incorporation at discrete positions. The incorporation
of isotopes into SAM’s structure can be accomplished via an established enzymatic synthesis using
isotopically labeled precursors. However, published protocols produced an intense and highly
variable IR signal, which overlapped with many of the signals from SAM rendering comparison
between isotopes challenging. We discovered this intense absorption came from co-purifying salts
and the SAM counterion, producing a strong, broad signal at ~1100 cm-1. To fix this, we developed
a revised enzymatic synthesis and purification protocol of SAM that mitigates the contaminating
salts. The optimized protocol was used to incorporate heavy atoms at various positions in SAM’s
molecular structure using commercially available isotopes in the precursors.
To use specific vibrations of SAM as probes of intermolecular interactions, the sensitive marker
frequencies observed in its spectrum were assigned to specific atoms in the molecular structure.
We present the first isotopically labeled methyl-d3-SAM, methyl-13C-SAM, and amino-15N-SAM
Raman spectra and assign the asymmetric stretching S-CH3 signal to 682 cm-1 using a combination
of isotope labeling, molecule mimics, and DFT calculations. The assigned S-CH3 stretch signal
can be used as a probe or marker band to detect non-canonical non-covalent interactions such as
carbon tetrel, C•••O, the CH•••O hydrogen bonding, or the Chalcogen S•••O interaction in the
enzyme active site. The result of this study lays the foundation for further exploration into the
application of SAM vibrations as label-free probes in sensing various non-canonical interactions
in SAM-dependent-enzyme active sites.
Insights gained from this study can be utilized in drug design efforts to address the specificity
challenges encountered by lead compounds targeting a single methyltransferase in this extensive
enzyme class. It also lays a strong analytical and physical foundation for future endeavors,
establishing SAM probe vibrations to enable investigations on various SAM binding proteins, and
addressing unresolved mechanistic and thermodynamic questions related to binding and catalysis
in enzyme active sites
Stochasticity and Spintronics for Bio-Inspired Computing and Hardware Security
No full textConventional CMOS scaling is rapidly slowing, galvanizing the search for alternative ar-
chitectures or technologies to complement the weaker aspects of CMOS. This dissertation
demonstrates that overlooked stochastic techniques and spintronic technologies provide sig-
nificant advantages over conventional CMOS in bio-inspired and hardware security applica-
tions: First, stochastic computing is experimentally demonstrated for ultra-efficient Bayesian
inference. Second, the inherent thermal stochasticity of Josephson junctions is harnessed for
efficient superconducting spiking neural networks. Third, the inherent stochastic switching
of magnetic tunnel junctions is experimentally demonstrated to enable analog-like Hebbian
learning in binary neural networks. Fourth, the isotropic stochasticity of nanomagnets is
identified as a self-destruction mechanism for physically secure logic locking. Fifth, the
highly-nonlinear dynamics of nanomagnet switching is harnessed for ultra-efficient reservoir
computing. Sixth, as this work is targeted toward moving these technologies toward commer-
cial use, a compact electrical model of magnetic domain wall motion is presented, enabling
efficient design of large-scale neural networks. While these techniques and devices are not
suitable for general purpose computing, they provide significant advantages over CMOS in
the bio-inspired and hardware security spaces, and should be pursued further for commercial
use
Mechanisms of Vagus Nerve Stimulation to Enhance Extinction Learning from Drug-Seeking Behavior
No full textSubstance use disorders are complex medical conditions characterized by an inability to refrain
from the seeking and consuming of drugs of abuse despite negative consequences. The pairing of
cues with drug taking creates powerful associations that can later lead to strong cravings and
relapse. One option is extinction learning where a new neutral association is paired with the
previously drug-paired cure to reduce the ability to elicit cravings. Unfortunately, extinction-based
therapies tend to fare poorly in the clinical setting with some attempts to enhance extinction
learning yielding mixed results. VNS is FDA approved for epilepsy, treatment-resistant
depression, obesity, and stroke rehabilitation. VNS has also been found to enhance memory
consolidation, and drive neuroplasticity. We have previously shown that VNS can enhance
extinction learning to reduce drug-seeking, but the mechanisms that drive these effects are still
unknown.
The work of this dissertation investigates a few of the potential mechanisms of how VNS can
enhance extinction learning to reduce drug-seeking. In Chapter 2 I describe experiments to test if
VNS affects cognition. We found that acute VNS can enhance short term memory, and cognition.
ix
In Chapter 3 I describe experiments investigating the underlying molecular mechanisms of VNS.
We examined the role of BDNF in VNS and found that it is necessary for VNS to enhance
extinction learning from drug-seeking. Additionally, we found that VNS can reverse glutamatergic
signaling deficits in the prefrontal cortex, and this too is mediated by BDNF signaling. In Chapter
4 I describe experiments that examine how VNS affects networks that regulate drug-seeking. We
found that VNS paired with extinction learning reduces activity in projections from the vHPC and
BLA to the IL, while activity from the BLA to PL is increased. Additionally, we found that VNS
increased PV interneuron activity in the PL, and decreased activity in the IL. These results start to
explore several mechanisms of how VNS can enhance extinction learning to reduce drug-seeking.
Understanding how VNS reduces drug-seeking is critical to for better implementation of VNS in
the clinical setting
Where Lies the Hope? Emergence of a Reimagined Cultural Identity After the 1985 Mexico City Earthquake
No full textIn 1985, an earthquake shook Mexico City, exposing inadequacies in the infrastructure of the
Federal District’s government agencies. Cultural change accompanied the aftermath of the
natural disaster and helped generate a new underlying concept of contemporary Mexican national
identity; recently released archival information from 1985 reveals cultural changes in grassroots
identities contributing to the nation's contemporary history. Mexicanidad, or “Mexican-ness,”
continues to clarify how historians and intellectuals retell earthquake events. Contemporary
intellectuals Carlos Fuentes, Carlos Monsivaís, and Elena Poniatowska reflected the mexicanidad
of Octavio Paz, transforming the documentation of the historical record as they experienced the
earthquake firsthand. Poetry, opera, comics, and celebrated icons that centered on the Mexico
City earthquake lent credence to various grassroots movements and even to propaganda by the de
la Madrid administration. The historical record reflected cultural topics popular in 1985 to
differing degrees. While poetry resonated with the generation of the 1968 student movement,
opera resonated with the upper classes. Remembrance comics attracted the low-literacy public
and mimicked traditional telenovelas. Celebrated icons of the era marked turning points in
history, signifying successes and failures of public support. The de la Madrid administration
supported babies pulled out of the earthquake rubble who are now 30 years old. Their story
highlighted what the presidential administration could accomplish. Also, seamstresses who lost
colleagues were granted union status to represent marginalized women. These cultural histories
have the potential to perpetuate a continued excitement of the significance of 1985, a gateway
moment between the single-party rule of the Partido Revolucionario Institucional (PRI)--who
maintained the office of the presidency since the Mexican Revolution of 1910--and the more free
and fair democratic elections of the year 2000, in which Vicente Fox and the Partido Acción
Nacional (PAN) changed single-party rule
Reduced Complexity Signal Subspace Algorithms With Applications in Wireless Ranging and Radio Astronomy
No full textThis dissertation presents reduced complexity signal subspace algorithms with applications in
radio frequency interference (RFI) mitigation in radio astronomy and wireless ranging. The
primary contributions focus on developing efficient model-based signal subspace and deep
learning methods for extracting either the signal of interest (SOI) or the relevant parameter
associated with SOI.
Toward RFI mitigation in radio astronomy, we devise signal subspace methods for the sce-
nario where the contaminating signal is much stronger than the SOI. This research introduces
two novel approaches for RFI mitigation in arrays of size M antennas: 1) a computationally
efficient Lanczos method based on the Quadratic Mean to Arithmetic Mean (QMAM) ap-
proach utilizing previously collected data under similar conditions, and 2) a reference-based
approach using celestial sources. The QMAM method employs the Lanczos algorithm to find
the Rayleigh-Ritz values of the covariance matrix (R ∈ CM ×M ), reducing the computational
complexity to O(dM 2), where d ≪ M . Numerical results from the Long Wavelength Array
(LWA-1) data demonstrates the efficacy of both methods in mitigating strong RFI, with the
QMAM-based approach offering significant computational efficiency.
Towards wireless ranging, we consider the scenario where the SOI is contaminated by a com-
parable or attenuated interference like multi-path signal. We use the Multi-Carrier Phase
Difference (MCPD) approach, which utilizes two-way channel frequency response (CFR)
measurements. Although MCPD approach comes with its benefits, it faces challenges due to
multi-path components, low number of snapshots, and practical model deficiencies. While
deep learning methods can help to address these challenges, they often require extensive
data and computational resources, and struggle to generalize across different environments.
This dissertation proposes a low-complexity deep learning model leveraging randomized low-
rank approximation (RLA) of the signal subspace from the two-way CFR covariance matrix.
The extracted signal subspace is fed into a Separable Convolutional Neural Network (CNN)
model, drastically reducing parameters, computational complexity and memory usage com-
pared to traditional methods like MUSIC or Support Vector Regression (SVR) while still
generalizing well to unseen environments. The proposed model achieves a root mean square
error (RMSE) of 0.5 m for ranges between 1 m and 7 m, outperforming MUSIC by 44% in
new environments without needing fine-tuning or additional data collection.
Additionally, a novel signal subspace decomposition (SSD) algorithm is developed for BLE
ranging in high multi-path environments. By integrating Fourier transform and RLA into
the SSD algorithm, computational complexity is minimized, making it suitable for embedded
devices. The enhanced SSD algorithm’s pseudo-spectrum features are used as input to a Long
Short-Term Memory (LSTM) recurrent neural network, forming a data-driven SSD-LSTM
wireless range estimator. Evaluations on real-world BLE data for both single- and multiple-
antenna scenarios show the proposed approach improves performance by over 37% compared
to existing methods like MUSIC and SVR, with reduced computational demands
Essays in Operations Management
No full textThis dissertation consists of three main chapters, which focus on resource allocation problems
that arise in the operations of crowdsourcing contest platforms and in the operations of
nonprofit organizations in the education sector.
In Chapter 2, we study the optimal contest design problem in crowdsourcing contest settings where contestants’ outputs are quantified using a pre-specified objective evaluation
metric, such as in data-science contests. We adopt the mechanism design framework (Myerson, 1981) to derive an optimal contest. Furthermore, we provide a practically convenient
implementation of the optimal contest.
In Chapter 3, to demonstrate the practical significance of the optimal contest designed
in Chapter 2, we compare the optimal contest with the most popular contest format in
practice, namely, the Winner-Takes-All (WTA) contest, on several metrics. Our numerical
investigation of the comprehensive testbed shows that the optimal contest delivers 23% (on
an average) additional utility to the contest designer than the WTA contest. Furthermore,
we show that the optimal contest is superior to the WTA contest on the expected best-output
metric. Conversely, we show that the WTA contest delivers higher welfare to the contestant
pool.
In Chapter 4, we study a resource allocation problem in nonprofit initiatives in which the
outcome to a beneficiary depends on their effort. For example, in education initiatives,
the lifetime outcomes for beneficiaries (students) depend strongly on the effort they exert
in their studies. In particular, we focus on initiatives that adopt a two-stage structure in
allocating resources to their beneficiaries, such as Tata Trusts - Karta Initiative’s Catalyst
Scholarship Program in India. We analyze the strategic role of an NPO’s resource allocation
strategy on the agent’s effort and their lifetime outcome. We show why an NPO benefits from
deliberately restricting resource access, even without resource scarcity. Finally, we analyze
the effect of resource scarcity and competition among the beneficiaries on their effort and
lifetime outcomes
The Effect of Private, Foreign and US Majority Ownership on Sporting and Financial Performance of English Soccer Clubs
No full textEuropean soccer clubs have been experiencing changes in their ownership structures in the last
two decades. This study analyzes the impact of different ownership structures on sporting and
financial performance with a scope on English Premier League for the seasons from 2007-2008
to 2021-2022. This study contributes to the literature, first by constructing a novel dataset of
ownership structures of Premier League Soccer Clubs for the seasons from 2007-2008 to
2021-2022. Second, the time scope of the previous literature is expanded, with the inclusion of
post-covid league seasons. The panel regression results show a negative effect of foreign private
majority ownership on sporting and financial performance. When US majority ownership and all
foreign majority ownership are compared, we see that US majority ownership does not perform
better in the league, but it does so financially. We also suggest evidence for the positive effect of
the seasons 2013-2014 and 2016-2017 on financial performance, pointing out respectively the
introduction of Premier League profit regulations, and the new TV broadcasting deal. The
negative effect of Covid shock in 2019-2020 is also present. In addition, the results support prior
literature by showing the positive and significant effect of payroll costs on sporting
performance. Lastly, the transfer expenses do not seem to improve sporting performance when
controlled for payroll costs