PhD Poster Competition, Art Exhibition, and Next Gen Startup Showcase - Research Days 2026 | The University of Texas at Dallas

PhD Poster Competition, Art Exhibition, and Next Gen Startup Showcase - Research Days 2026 | The University of Texas at Dallas
Skip to content
PhD Poster Competition, Art Exhibition, and Next Gen Startup Showcase
PhD Student Poster Competition, Art Exhibition, and Next Gen Startup Showcase
Wednesday, October 15
th
, from 9:30 AM to 3 PM
Session 1: 9:30 AM – 11:30 AM | Session 2: 1:30 PM – 3:30 PM
Join the Office of Research and Innovation as we host the PhD Student Poster Competition, Art Exhibition, and Next Gen Startup Showcase on Wednesday, October 15th at the UT Dallas Visitor Center.  Students will present their research during the poster sessions from 9:30 AM to 11:30 AM and 1:00 PM to 3:00 PM. For the first time, we will host a separate art showcase where nominated students can present their art. In this “people’s choice” competition, you will be able to vote for your favorite poster and artwork.  Winners will be announced during the awards ceremony immediately following the afternoon sessions.
Additional information can be found
here
Questions? Contact
Taylor Yarborough
for the Poster Competition and
DeMia Keppel
for the Art Showcase.
PhD Poster Students
Next Gen Startup Showcase
ART STUDENT PARTICIPANTS
Arman Eshtiaghi
Art
301
FORMULA
Art Description →
FORMULA is a multimedia art project in which Arman Eshtiaghi externalizes and visualizes his inner emotional landscape, mapping how feelings, identity, and memory intertwine. At its core, it seeks to render the invisible visible, the cycles of emotion, the flux of identity, and the often hidden spaces of the mind. Media and Structure: The project fuses animation, 3D modeling, typography, video art, layout design, and lighting and projection within an immersive gallery environment. A soundscape underpins the visual work, with the artist’s voice reading poems in the background, modulated to approximate brainwave frequencies during sleep. The central written piece, a poem titled Formula, was composed over six years and divided into four segments, each corresponding to one of four video and brain imagery works. Typography is used not just as text but as texture, vibration, and visual form, with some words legible and others intentionally obscured. Color acts as a structuring principle, with each poem and video section tied to a distinct color bar that guides the viewer through the emotional narrative. The exhibition also incorporates visuals produced by deep neural networks, representing thoughts and cognition as generative visual forms. Concepts and Themes: Emotion is portrayed as cyclical and looped, with videos evoking the ebb and flow of internal states, suggesting that emotional experiences rarely resolve linearly but return, evolve, and repeat. Identity and queerness are central, with FORMULA foregrounding plurality and fluidity, challenging static notions of gender and sexuality, and positioning identity as mutable and layered. Visibility and invisibility play a strong role, with imagery of masking and hidden text reflecting how identity can be constrained or policed. The fusion of 3D brain renderings with typographic elements suggests dialogue between emotion, thought, and language. Subconscious and dream spaces are evoked through audio tuned to sleep brainwave frequencies, pointing toward hidden realms where emotional truth resides. Purpose and Intention: FORMULA is a reclamation of voice. Eshtiaghi frames it as a defiant response to silencing and erasure, using the work to make visible those parts of self that are suppressed or denied. It creates a mirror not only for his own psyche but also for anyone who has felt unseen, unheard, or confined by external expectations. By weaving together poetry, visual art, technology, and sound, FORMULA invites the viewer to inhabit a porous understanding of the self, one in constant becoming, always hovering between clarity and mystery.
Jacqueline Hinojosa
Art
302
Ackerman Center and S/PN Gallery Digital Twin
Art Description →
This project is a recreation of the Ackerman Center and S/PN Gallery as immersive digital twins using 3D modeling, photogrammetry, and game engines. This allows users to explore the spaces virtually, making it possible to attend lectures and access historical information at the Ackerman Center. For the S/PN Gallery, it also serves as a living archive, preserving past exhibitions and giving remote audiences the ability to experience them as if they were physically present.
Joseph Gutierrez
Art
303
Jaxon of the Comanche
Art Description →
A playable cinematic experience that is meant to encourage the youth of the Comanche Nation to be more involved with preserving their heritage, culture, and language.
PHD STUDENT SESSION 1 PARTICIPANTS
Xingyu Chen
Poster
101
From Ambiguity to Precision: Clarifying Cell Identity in scRNA-seq atlases of C. elegans via Cell–Cell Interaction–Guided Learning
Abstract →
The invariant cell lineage of Caenorhabditis elegans has long served as a foundational model for understanding the molecular mechanisms of development. Accurate cell identity annotation is essential for building mechanistic models of early embryogenesis. In foundational single-cell RNA sequencing (scRNA‑seq) atlases of the early embryo, closely related sister cells often display near‑indistinguishable transcriptomes as cell numbers grow and lineage branches diversify, leading to pervasive merged labels (e.g., ABala/ABalp → “ABalx”) and limiting downstream inference of gene‑regulatory network and morphogenetic programs. To address this, we presented CHACAM (Cell–Cell Interaction–guided Hierarchical Attention–based Cell Allocation Model), a supervised framework that integrates gene expression, a curated C. elegans ligand–receptor (L–R) interaction database, and a high‑resolution physical contact map to disambiguate cell identities. CHACAM learns “interaction signatures” that distinguish a cell’s contacting neighbors from non‑neighbors and uses a triangulation strategy with third‑party reference cells. Applied to the early embryo atlases (Tintori et al., 2016; Cole et al., 2024), CHACAM achieves error‑free leave‑one‑embryo‑out allocations at all stages and resolves long‑standing cell ambiguities in multi lineages, especially in AB‑lineage. The curated C. elegans L–R database (currently 344 pairs spanning major developmental pathways) and the model’s attention weights highlight contact‑specific signals as top discriminants, providing interpretable hypotheses for early fate specification. Our results establish intercellular context as a critical determinant of identity at the onset of development and offer a generalizable strategy for high‑precision annotation in other systems.
Arowa Yasmeen
Poster
103
Structuring the Airspace to allow commercial UAV pre-flight path planning
Abstract →
Traditional Air Traffic Control systems cannot accommodate UAV operations in uncontrolled Class G airspace, creating challenges for high-density urban UAV traffic management. This research presents a comprehensive framework addressing two critical challenges: designing structured airspace infrastructure and developing real-time path-planning algorithms for safe, efficient UAV operations. Our approach introduces a modular system architecture with a Flight Scheduler coordinating airspace management, geofencing, weather monitoring, and safety components. The airspace design synthesizes Air-Matrix, Air-Network, Air Tubes, and Layer-based concepts, incorporating Vertiports, Vertical Corridors, Horizontal Corridors with Sky Lanes, and Elliptical Intersections for traffic flow management. Our use of temporal allocation of airspace subdivisions using strategically positioned waypoints enables greater efficiency than full-time route blocking while minimizing collision risks. We also introduce an approximate path-planning algorithm that optimizes travel time and energy consumption while satisfying temporal, spatial, and regulatory constraints through real-time multi-objective optimization. This research contributes scalable UAS Traffic Management solutions essential for integrating UAVs into everyday urban operations while maintaining safety and regulatory compliance.
Maryam Piroozzadeh
Poster
105
Fluorescence-Based Detection of PFOS Using a Naphthalenediimide-Linked Covalent Organic Framework
Abstract →
Sensitive detection methods for “PFAS” are critical for evaluating the safety level of our water sources. LC-MS-MS is the method we are currently using for the detection of PFAS, but it requires highly trained laboratory personnel, expensive and complex instrumentation, difficult sample preparation, and tedious data collection. Therefore, alternative detection method based on fluorescent sensing have intrinsic advantages like fast response time, easy operation, cost effectiveness, naked eye identification, and the potential to build hand-held sensor devices for onsite detection. ❖ We have designed a novel cationic covalent organic framework (c-COF) based on perylene diimide ❖ The hydrophobic functionality and cationic charge assisted for the detection of both light weight PFCAs and PFOS.
Zeinab Delaram
Poster
107
Mortality Prediction in Liver Transplantation
Abstract →
Liver transplantation (LT) remains the definitive therapy for end-stage liver disease, yet predicting post-transplant outcomes remains a critical challenge. Current systems, such as the Model for End-Stage Liver Disease (MELD) and the Donor Risk Index (DRI), provide baseline stratification but fail to fully capture longitudinal disease paths or complex donor–recipient interactions. The goal of this study is to introduce a comprehensive benchmarking framework that systematically compares traditional clinical scores, statistical survival models, and modern machine learning approaches within a unified LT cohort. Unlike prior investigations that assess models in isolation, our approach emphasizes direct head-to-head comparisons. By integrating longitudinal deep learning architectures, we move beyond static predictions to model temporal progression. Furthermore, our pipeline integrates explainability (e.g., SHAP values) and fairness analyses to enhance equity concerns in outcome prediction. This unified, clinically relevant framework aims to improve accuracy, interpretability, and fairness, thereby strengthening the potential for real-world adoption in transplant decision-making and long-term patient management. Future integration of AI-driven models with real-time patient data is anticipated to improve mortality prediction further, optimize treatment planning, and improve long-term patient outcomes.
Amber Hasan
Poster
109
New Mathematical Tools for Strengthening How We Handle Data
Abstract →
This research develops new mathematical tools for keeping information safe from mistakes using permutations. By applying six main techniques, we improve known results and prove that bigger and stronger lists of permutations are possible, leading to better ways of organizing and protecting data.
Bianca David
Poster
111
Electrochemical Profiling of for Systemic Inflammatory State Detection
Abstract →
This research aims to develop a portable biosensor device for quickly detecting vWFA2, a biomarker for inflammatory conditions. This sensor could dramatically change detection methods and lead us to improve the sensitivity of our tests to overcome the limitations of conventional detection methods. Our label-free biomolecular assay is constructed on an Au-ZnO electrode surface and uses electrochemical impedance spectroscopy (EIS) to measure the capacitive change in impedance, revealing the binding effects of the target vWFA2, to the capture probe. Our developed biosensor platform exhibits greater sensitivity and specificity, covering a wide dynamic range of 750–24,000 pg/mL and showing a strong correlation with inflammatory conditions. This sensor exhibited a greater accuracy ranging from 86–110% for the known spiked concentrations in nondiluted or modified plasma samples. This electrochemical sensor has the potential to advance point-of-care diagnostic methods due to its high sensitivity and rapid response time. The vision behind this research is to develop an electrochemical sensor that can rapidly and accurately detect disease states, thus creating a pivotal prognostic tool in inflammatory state treatment and ultimately mitigating severe mortality and morbidity.
Emmanuel Adehunoluwa
Poster
113
Appropriately Targeted Plasticity Through Vagus Nerve Stimulation Improves Walking Capacity in Chronic Spinal Cord Injury Patients
Abstract →
BACKGROUND Evidence from preclinical studies demonstrates that the effects of vagus nerve stimulation (VNS) are specific to paired events. Thus, when VNS was paired with upper extremity movements in a recent clinical trial, improvements were observed in the arms and hands of individuals with chronic incomplete spinal cord injury (iSCI), but not in the legs. Walking is often impaired in iSCI patients, resulting in reduced mobility and decreased participation in community activities. Current rehabilitation strategies only produce limited benefits, and patients are still left with significant impairments. OBJECTIVE We reasoned that coupling VNS with lower limb movements may improve walking performance in individuals with iSCI. We tested this approach to demonstrate the event-dependent plasticity capability of VNS in a clinical population, as well as the potential of this treatment to treat several complications of neurological diseases through appropriately targeted synaptic plasticity. (ClinicalTrials.gov: NCT06351111). METHODS First, we developed and implemented an algorithm that uses real-time signals from movement sensors to deliver closed-loop VNS (CLV) during gait exercises. Three individuals who had undergone implantation of a VNS device as part of a prior study completed up to 24 sessions of CLV paired with rehabilitation tasks, which consisted of lower limb exercises and different variations of active walking. Outcome measures included the 6-minute walk test (6MWT) for timed distance, the 10-meter walk test (10MWT) for gait speed, the lower extremity motor score (LEMS), and spatiotemporal gait parameters. Assessments were performed at baseline, after 12 CLV sessions, after 24 CLV sessions, and at a 1-month follow-up. RESULTS No serious adverse events were observed. All three participants demonstrated substantial improvements in lower limb function and walking capacity. CLV led to a clinically meaningful increase in the participants’ motor score (+5.7 points). CLV also significantly increased the timed distance (+123.3 meters), the fast gait speed (+0.22 meters/second), and the stride length (+30.3 centimeters). CONCLUSION These initial findings indicate the capability of VNS to improve functions after neurological injuries when plasticity is appropriately directed. It emphasizes the need for larger studies in these contexts.
Sakib Reza
Poster
115
Quantized Bessel Phase and Physically Embedded Gaussian Tapering for Near-Field Wireless Power Transfer Using Low-Cost Metasurface Lenses
Abstract →
This paper presents the experimental demonstration of a low-cost metasurface lens for near-field wireless power transfer (WPT) at 24 GHz. The lens is horn-fed and engineered to embed quantized spherical-phase control and Gaussian amplitude tapering directly into its physical structure. Both functionalities are realized through geometric variation in cross- shaped unit cells distributed across a 10λ × 10λ aperture. This dual-control strategy enhances near-field focusing by mitigating quantization artifacts and edge diffraction effects. Fabricated using standard PCB processes, the metasurface achieves efficient energy convergence without auxiliary feed shaping. Measurement results confirm a tightly confined beam sustained over a 7.6λ depth and a 5.8 dB enhancement in transmission coefficient |S21| relative to free-space propagation. These findings validate the effectiveness of geometry-driven amplitude tapering in improving spatial energy localization for compact WPT systems.
Prasoon Vishwakarma
Poster
119
Post-Sunset Equatorial Ionization Anomaly (EIA) variations during a weak geomagnetic activity on 23-24 February 2023
Abstract →
This study explores the evolution of the post-sunset Equatorial Ionization Anomaly (EIA) following a weak geomagnetic activity on 23–24 February 2023 over the American sector (65°W), using observations and Global Ionosphere Thermosphere Model (GITM) simulations. Merging of EIA crests is observed on 24 February, with a faster equatorward motion of the northern crest. GITM simulation well captures the observed EIA merging and its interhemispheric asymmetries (IHAs). GITM simulation indicates that reduced upward ion drift partly contributes to the merging and an equatorward propagating traveling atmospheric disturbance (TAD) accelerates the equatorward movement of the NH EIA peak. The IHAs in the TAD propagation result from IHAs in the Joule heating deposition. Stronger Joule heating is deposited in the NH that drives the NH TAD, while no evident Joule heating is deposited in the SH at the same time.
Milad Almasian
Poster
121
Axially swept dithered light-sheet microscope to reveal cardiac morphology
Abstract →
Understanding cardiac microstructure and vascular networks in three-dimensions (3D) is critical for assessing cardiovascular development, disease progression, and therapeutic interventions. Light-sheet microscopy combined with tissue clearing has emerged for high-resolution volumetric imaging of intact organs but facing limitations in compacted and trabeculated myocardium due to trade-offs in compromised light-sheet thickness, constrained frame rates, and restricted working distance of medium-immersion lenses. We present dithered light-sheet (DiLS), which extends the confocal region by at least 40% while preserving optical sectioning quality. By integrating DiLS with the axially swept approach, we introduce axially swept dithered light-sheet (AS-DiLS) microscopy, which increases imaging speed without compromising axial resolution and enabling uniform illumination up to a 12.5-millimeter scanning range. AS-DiLS delivers near-isotropic resolution of ~2.5 μm for investigating intricate ventricular trabeculae, atrial pectinate muscles, vasculature, and extracellular matrix. Collectively, our methodology provides a scalable framework for in-depth assessment of cardiovascular morphology and topology in animal models from embryonic stages to adulthood.
Zhihao Ma
Poster
123
Ultra-Low Temperature Dopant Activation Enabled by Plasma Enhanced Annealing
Abstract →
The formation of highly doped, ultra-shallow junctions is critical for the continued scaling of advanced semiconductor devices. However, achieving high dopant activation while preserving lattice integrity remains a fundamental challenge. Conventional annealing techniques such as furnace annealing and rapid thermal annealing (RTA) typically require processing temperatures above 800 ℃ to activate dopants and repair implantation damage. Although effective, these high-temperature treatments inevitably induce excessive dopant diffusion, junction broadening, and thermal degradation of fragile device materials, restricting their suitability for next-generation technologies. Plasma Enhanced Annealing (PEA) has emerged as a compelling alternative, offering an ultra-low thermal budget through localized and controllable energy delivery from plasma ions. By tailoring ion energy, flux, dose, and species, PEA enables efficient dopant activation at substantially lower substrate temperatures (300–500 ℃). This precise, near-surface energy transfer preserves shallow junction profiles, mitigates implantation-induced defects, and maintains device integrity while achieving high activation efficiency. In this study, implanted silicon wafers were annealed using PEA, and their post-process properties were systematically evaluated using four-point probe measurements and Raman spectroscopy. Relative to conventional thermal annealing (TA) and RTA under identical conditions, PEA demonstrated noticeably superior electrical activation which is demonstrated by significantly reduced sheet resistance together with enhanced crystallinity. 500 ℃ PEA of shallow boron implants reached essentially the same sheet resistance as for RTA at 800 ℃. Moreover, a comparative analysis of different inert ion species revealed distinct influences on dopant activation efficiency and lattice recovery. These results establish PEA as a robust and versatile technique for forming highly activated, thermally stable, ultra-shallow junctions, positioning it as a strong candidate for integration into next-generation semiconductor device manufacturing.
Donald MacPhail
Poster
125
Weight Identity: Both a Risk Factor and a Protective Factor?
Abstract →
Weight identity can be conceptualized as the degree to which one’s weight is an important part of their sense of self (Curll & Brown, 2020; Campbell et al., 2022). Hudson and colleagues (2025) recently published a systematic review in which they evaluated whether identifying as a member of a higher-weight group increased or buffered the negative effects of weight stigma on well-being. This review concluded that higher-weight social identity exacerbated the effect of weight stigma on outcomes such as anticipated rejection and dietary control challenges. But the researchers also found that weight identity reduced the effect of weight stigma on self-esteem among those with low internalized weight bias, showcasing weight identity’s dual role as a moderator of weight stigma. Weight identity has not been readily evaluated across weight statuses, as it has been predominantly assessed among higher-weight samples. Yet, people of any weight status can view their weight as an important part of their identity. The present study aimed to determine whether weight identity moderated the effects of external-based weight stigma (i.e., weight discrimination) and internal-based weight stigma (i.e., fear of fat), while controlling for BMI and gender. A sample of 628 undergraduate students were recruited (Age: M = 21.13 years, SD = 4.58; BMI: M = 24.39, SD = 5.46). Weight identity was found to moderate the effect of weight discrimination on psychological well-being (p = .013) and body dissatisfaction (p = .039), such that the effects of weight discrimination were weakest when weight identity was high. Conversely, weight identity moderated the effect of fear of fat on disordered eating behaviors (p < . 001) and self-esteem (p = .005), but these effects were strongest when weight identity was high. Weight identity did not moderate the effect weight discrimination had on self-esteem (p = .547) nor disordered eating (p = .467), and did not moderate the effect of fear of fat on body dissatisfaction (p = .467) nor psychological well-being (p = .688). These results largely echo the findings of Hudson and colleagues (2025), showcasing that weight identity can operate as either a protective or risk factor, potentially being outcome or context dependent. However, these results also indicate that something as specific as whether the source of negative weight attitudes is external or internal may be the distinction for whether a higher level of weight identity may be beneficial or harmful. Failure to facilitate further understanding of weight identity in this context would severely limit the potency of interventions developed for individuals of all weights and sizes.
Kundan Mishra
Poster
127
Portable Real-time Electrochemical Sensing System for the detection of Mycotoxin
Abstract →
Foodborne toxins such as Aflatoxin B1 and Zearalenone pose significant public health risks worldwide due to their high toxicity and prevalence in agricultural products. Traditional culture-based detection methods, while reliable, are often limited by low sensitivity, long assay times, and the need for laboratory infrastructure. To address these challenges, we present a portable, label-free non-faradaic electrochemical sensing platform for the rapid and accurate detection of Aflatoxin B1 and Zearalenone in corn flour. This system delivers results within approximately 5 minutes and employs toxin-specific antibodies to achieve low limits of detection—0.005 ng/mL for Aflatoxin B1 and 0.05 ng/mL for Zearalenone. The sensor demonstrates wide dynamic detection ranges (0.01–9.151 ng/mL for Aflatoxin B1 and 0.1–25.6 ng/mL for Zearalenone) with strong analytical performance and reproducibility, maintaining inter- and intra-assay coefficient of variation (%CV) below 20%. Validation against benchtop systems confirms its practical utility. This compact and user-friendly platform offers a promising solution for on-site mycotoxin screening, enhancing food safety and enabling field-deployable contaminant monitoring.
Athresh Karanam
Poster
129
AI-in-the-loop: A Probabilistic Perspective
Abstract →
The recent success of machine learning frameworks in applications such as image classification, speech recognition, and natural language processing has largely been driven by the availability of large amounts of data and computational power. However, this approach is infeasible in noisy and data and resource-scarce domains such as healthcare. To this end, we propose a research direction that aims to leverage the rich domain knowledge that experts possess to build more effective, expressive, and explainable models. In particular, we propose an AI-in-the-loop framework that acknowledges the utility of domain experts as beyond mere labellers. We showcase the effectiveness of this framework across different models, data modalities, and types of domain knowledge leveraged.
Ariel Tolfree
Poster
131
Single-Print Multimaterial 3D Printing of Sandwich Structures Enabled by Catalyst-Free Dynamic Covalent Chemistry
Abstract →
Sandwich composite structures are an essential class of lightweight building materials used in architectural, automotive, and aerospace applications. Conventional fabrication relies on stepwise synthesis and assembly, which restricts the complexity of achievable geometries and often leads to adhesion challenges. While additive manufacturing enables complex, computer-designed architectures and the fabrication of sandwich structures, it still imposes certain practical constraints. These constraints often require trade-offs between mechanical performance and the distinctive multifunctional benefits of sandwich structures—such as high strength-to-weight ratios, thermal insulation, and impact resistance. In this work, we demonstrate the single-print fabrication of a foam-core sandwich structure using vat photopolymerization coupled with catalyst-free dynamic covalent chemistry. This approach eliminates the need for a separate adhesion step while taking advantage of 3D printing’s capacity to produce complex geometries and incorporate multiple materials.
Zahra Sepasi
Poster
133
Thermal Stability of Graphdiyne
Abstract →
Graphdiyne (GDY), a two-dimensional carbon allotrope composed of alternating sp and sp² hybridized carbon atoms, has attracted significant attention due to its semiconducting nature, high π-conjugation, and structural tunability. In contrast to graphene’s uniform sp² bonding, GDY incorporate diacetylenic linkages between aromatic rings, which offers unique opportunities for electronic, catalytic, and energy-related applications. However, the intrinsic thermal and structural stability of these acetylenic linkages remains underexplored. The DFT and MD simulation reported so far show strong stability of GDY structure up to 1000 K. On the other hand, the increase in acetylenic linkages (triple bonds) should directly reduce the cohesive energy, indicating a less stable network. GDY’s larger pore size, stemming from extended conjugated triple bonds, leads to weaker π–π stacking interactions, further reducing overall thermodynamic stability comparing to graphene and graphdyne. Here we report an irreversible exothermic transformation of graphdiyne at ambient pressure and low temperatures, ~150oC – 300oC. This transformation is caused by instability of triple-bonded acetylenic groups. The thermodynamic properties, structural changes, and spectroscopy data of triple-bond reaction are discussed. A scalable, metal-catalyzed, graphdiyne powder was synthesized from a commercially available hexakis[(trimethylsilyl)ethynyl]benzene (HEB-TMS) following Li’s et al. deprotection-free protocol. Using Cu(I) Iodide in N,N-dimethylformamide under ambient air at 60 °C, polymerization proceeds with high yield via in situ oxidative coupling. The presence of oxygen is found to be essential for maintaining the high efficiency of the coupling process. Post-synthesis, the material is acid-washed to remove copper oxide nanoparticles and characterized structurally and spectroscopically. Raman spectroscopy confirms the successful transformation from monomer to polymer through the emergence of characteristic D (~1380 cm⁻¹), G (~1570 cm⁻¹), and diyne (~ 2170 and 2282 cm⁻¹) bands. Complementary FT-IR spectroscopy reveals the appearance of strong absorption peaks in the 1950 and 2170cm⁻¹, consistent with the stretching vibrations of conjugated -C≡C- triple bonds in diyne linkages. This confirms the retention and periodic integration of sp-hybridized carbon structures within the GDY network, distinguishing it from disordered carbon or graphite-like materials. Thermal stability and phase transition behavior were studied using simultaneous differential scanning calorimetry (DSC) and thermogravimetric analysis (TGA). An irreversible exothermic transformation is observed in the temperature range of 150–300 °C under nitrogen atmosphere, corresponding to the collapse of triple bonds and reorganization into more thermodynamically stable sp²-rich carbon frameworks. This transformation is supported by a distinct exothermic peak in the DSC heat flow and corroborated by the loss of diyne vibrational modes of Raman and FT-IR spectra in the gradually annealed in vacuum GDY particles. These findings indicate that although graphdiyne exhibits moderate thermal stability, it undergoes a spontaneous and energetically favorable transition to denser carbon phases. The measured enthalpic change during this transition offers a first quantitative insight into the thermodynamics of sp-to-sp² rearrangement in 2D carbon frameworks. This understanding is critical for guiding the processing conditions and device design involving graphdiyne and related nanocarbon materials.
Ishara Ekanayake
Poster
135
Mandrel-free fabrication of giant spring-index and stroke muscles for diverse applications
Abstract →
Methods for making high-spring-index polymer fiber or yarn muscles have required expensive fabrication by wrapping around a mandrel, which limits their practical applications. We demonstrate an inexpensive mandrel-free method for making polymer muscles that can have a spring index of>50 and a contractile tensile stroke exceeding 97%. This method enables the spring index to be varied along a muscle’s length by varying the plying twist, resulting in muscles that transition between homochiral and heterochiral when either heated or cooled. We demonstrate use of these polymer muscles for robots and environmentally driven comfort-adjusting jackets. This mandrel-free method was used to make high-spring-index carbon nanotube yarns for mechanical energy harvesters,self-powered strain sensors, and solvent-driven and electrochemically driven artificial muscles.
Matthew Beltran
Poster
137
Quantifying The Impact of Ball-Milling Parameters on the Ionic Conductivity of Solid-State Electrolyte via Orthogonal Design of Experiments
Abstract →
The performance of solid-state electrolytes (SSEs) is vital for the successful implementation of all-solid-state Li-ion batteries. (ASSLBs). SSE materials are commonly synthesized through a dry ball-milling procedure, eliminating the need for conventional solvents in both synthesis and full-cell configurations. There are several factors in the synthesis procedure that can influence the performance metrics of any particular SSE material; among these metrics, the ionic conductivity (σ_i) is arguably the most important metric for assessing the efficacy of a solid Li+-ion conductor. However, a thorough statistical and regression modeling analysis of the synthesis process is lacking, as it provides an understanding of how specific synthesis parameters impact σ_i. In this work, we devise an orthogonal design of experiments on Li2ZrCl6 (LZC), a notable, cost-effective halide-based SSE, to explore the impact of six synthesis parameters on the resultant ionic conductivity. The orthogonal method reduces the required number of experiments to evaluate the importance of each parameter from several thousand to only 18. The results reveal that the ball-to-precursor mass ratio, the ball-mill step time, and the milling speed are the most significant factors influencing ionic conductivity. The impact of these parameters were modeled and confirmed via multivariate linear regression fitting. Lastly, a Gaussian process regression model was used to optimize the synthesis conditions for this system and accurately predict the output ionic conductivity. The findings reported here may be applied to other SSE systems to enable a more consistent production of high-quality SSEs for the next generation of ASSLBs.
Youhwan Jo
Poster
139
Control of diffusive interfacial heat transport via amorphous interlayers
Abstract →
The 3D-stacked chip design is becoming popular as a solution to overcome the performance limitation of conventional electronics, however, a robust solution is necessary to manage and extract heat across the multiple materials that are intricately integrated. To achieve this, understanding interfacial heat transport between materials is critical, as interfaces act as bottlenecks to heat flow.  In this study, the diffusive nature of interfacial phonon transport and the control of thermal boundary resistance by employing amorphous interlayers are covered based on molecular dynamics simulations of Si-diamond interface models.
Stephanie Yamauchi
Poster
141
INVESTIGATING CHRONIC IL-1-INDUCED CASTRATION-RESISTANT BEHAVIORS IN CASTRATION-SENSITIVE PROSTATE CANCER CELLS
Abstract →
Prostate cancer (PCa), the second-leading cause of cancer-related death in the United States, heavily relies on the transcription factor androgen receptor (AR) to fuel growth and proliferation. Thus, androgen deprivation therapy (ADT) is used to counter PCa growth. However, PCa patients can develop resistance to ADT, termed castration-resistance, due to a multiplicity of reasons, including mutated or overactive AR and adaption to AR loss.
It is established that chronic inflammation can cause cancer initiation and progression. We have found that chronic exposure to the inflammatory cytokine, interleukin-1 (IL-1), can cause castration-sensitive PCa cells to assume more castration-resistant-like behaviors, including decreased sensitivity to exogenous acute IL-1 signaling, AR-independence (i.e., castration insensitive), and repression of the AR-regulated PCa diagnostic marker, PSA. Thus, we hypothesized that chronic IL-1 exposure drives PCa castration resistance.
To test our hypothesis, we compared stable, basal changes gene expression in a castration-sensitive PCa cell line that was chronically treated with IL-1 to that of a castration-resistant PCa cell line. We compared previously obtained RNA sequencing data from castration-sensitive LNCaP cells, novel LNCaP chronic IL-1 sublines generated in our lab (LNas1, LNbs1), and a castration-resistant cell line, C4-2. We wanted to determine whether chronic IL-1 would: (1) drive the LNCaP genome to more closely resemble that of C4-2 cells, and (2) alter AR regulation in LNCaP cells such that it mimics C4-2 cells. Interestingly, while overall bearing little genomic resemblance, a subset of AR-regulated genes behaved similarly between chronic IL-1 treated LNCaP cells (LNas1, LNbs1) and C4-2 cells. Thus, alterations in AR-regulated signaling are expected to underly chronic IL-1-induced castration resistance.
Yuan Tan
Poster
145
Dual-Anion Sodium Halide-based Solid Electrolytes With High Ionic Conductivity and High-Voltage Stability
Abstract →
To address the high cost and safety concerns of the lithium-ion batteries in our phones and cars, our research focuses on a promising alternative: sodium solid-state batteries (SSBs), which use abundant, low-cost sodium and replace the flammable liquid inside with a safe, solid material. The primary challenge is creating a solid-state electrolyte (SE)—essentially a highway for energy ions—that is highly efficient. Many solid materials act like congested roads, slowing down ion traffic and hindering battery performance. Our breakthrough was developing a new electrolyte using a strategy called anion mixing, where we precisely substituted some chlorine atoms with oxygen atoms in the material’s structure. This simple swap transformed the material into a “glassy,” disordered state, creating a superhighway that dramatically increased ionic conductivity, or the speed at which sodium ions can travel. To prove this concept, we built a full battery cell that demonstrated remarkable durability, retaining 73% of its capacity after 1000 charge-discharge cycles. This work highlights a practical and effective method to design the next generation of safer, more affordable, and high-performance sodium batteries.
Changkai Zhou
Poster
147
Bayesian Trajectory Inference for Sparse Spatial Omics Data
Abstract →
The spatiotemporal organization of gene expression is critical for understanding dynamic biological processes, but inferring directional trajectories from sparse, zero-inflated spatially resolved transcriptomics (SRT) data remains a significant challenge. Here, we present a hierarchical Bayesian framework for trajectory inference designed to robustly handle sparse spatial omics data. Our model uses a Zero-Inflated Negative Binomial likelihood and employs a Reversible Jump Markov Chain Monte Carlo (RJMCMC) algorithm to infer an unknown number of change points along potential paths, defining trajectory segments directly in physical space. On simulated data with varying patterns, the method accurately identifies change points, as demonstrated by high Adjusted Rand Index (ARI) and low Hausdorff distance (HD) metrics. We applied the framework to a mouse brain Visium dataset following traumatic brain injury (TBI) and identified migratory trajectories of activated microglia. The inferred paths for key marker genes converge on the injury penumbra, consistent with known microglial response dynamics. This flexible and interpretable framework provides a powerful tool for discovering spatial trajectories in complex tissues, offering new insights into dynamic biological systems.
Marlon Jerez
Poster
149
Hybrid Lithography Process for Scalable Resistive Memory Integration
Abstract →
Conventional memory technologies such as DRAM and Flash face challenges related to power consumption, capacitor scaling, and other scaling limitations. Resistive Random-Access Memory (RRAM) offers a simple Metal-Insulator-Metal (MIM) structure, CMOS compatibility, and resistive switching, making it a strong candidate for dense integration and neuromorphic computing. In this work, we demonstrate sub-micron 32-bit RRAM arrays with 500 nm features and ON/OFF ratio exceeding 1×106. To enable this demonstration, we developed a hybrid lithography strategy combining Electron Beam Lithography (EBL) and photolithography, supported by mix-and-match alignment marks. Particularly, automated alignment procedures performed every 1.3 mm limited write-field shifts to < 30 nm, ensuring high accuracy while improving throughput. This novel strategy provided the scalability required to fabricate dense device arrays. Electrical characterization confirmed functional RRAM devices down to 500 nm. The ON-state current density increased from 3.1×104 A/cm2 (5 µm) to 2.5×105 A/cm2 (500 nm), while the OFF-state current decreased from 0.17 A/cm2 to 0.03 A/cm2. As a result, the ON/OFF ratio improved significantly reaching 7.6×106 at 1 µm device size. These results establish a scalable methodology for fabricating dense RRAM arrays, advancing their integration in next-generation memory and neuromorphic systems.
Jessica Gomez
Poster
151
Chronic IL-1-Exposed Bladder Cancer Cells Maintain Sensitivity to IL-1 Receptor Antagonist
Abstract →
Bladder cancer (BlCa) is the fourth most common cancer in men and has a 71% 5-year survival rate if confined to the bladder; however, this survival rate is reduced to 39% or 8% for regional or distant metastasis. Thus, it is crucial to understand the underlying mechanisms that lead to the progression of BlCa metastasis. The tumor microenvironment (TME) is replete with inflammatory factors secreted by infiltrating immune cells recruited to the TME to kill the cancer cells. Acute inflammation is anti-tumorigenic, yet, if left unresolved, chronic inflammation is a known hallmark for cancer initiation and progression. Our lab has discovered that when chronically exposed, cancer cells can develop insensitivity to the anti-tumorigenic effects of the inflammatory cytokine, interleukin-1 (IL-1), including avoiding IL-1-induced cell death. Consequently, IL-1-targeted therapies, such as the IL-1 receptor antagonist (IL-1Ra), anakinra, would not be effective against IL-1-insensitive cancers, enabling the pro-tumorigenic effects of IL-1 such as tumor angiogenesis, metastasis and drug resistance. Deviating from our findings for other cancer cells lines, we discovered that the BlCa cell line, 5637, maintains IL-1 sensitivity following chronic IL-1 exposure. Thus, 5637 cells maintain sensitivity to IL-1RA. Specifically, we find that IL-1RA prevents 5637 cell-cell dissociation, a phenotype associated with metastasis, and blocks 5637 paracrine activation of angiogenic endothelial cells. However, in line with what we have observed for other cancer cell lines, chronic IL-1-exposed 5637 cells evolve stable changes in the expression of critical tumor-promoting genes. For example, chronic IL-1 exposure causes 5637 cells to evolve high basal levels of SERPINE1. SERPINE1 is frequently overexpressed in BlCa and is associated with poor prognosis, increased tumor aggressiveness, and higher stages of the disease. Taken together, our chronic IL-1 BlCa cell line model is a valuable tool to characterize tumor inflammation and to identify biomarkers that could predict patient response to IL-1RA and other anti-inflammatory therapies.
Priya Christensen
Poster
153
How to Tune your Membrane: Turnover of Lysine modified lipids in Streptococcus agalactiae
Abstract →
Group B Streptococcus (GBS) is a Gram-positive bacterium that colonizes the gastrointestinal and lower genital tracts of humans. Approximately 20-30% of women are colonized with GBS in the United States. GBS may be transmitted from mother to infant via direct infection or interaction with infected fluids during delivery, which can result in severe infections in the newborn, including sepsis and meningitis. Understanding how GBS invades and protects itself from human immune defenses is critical for future anti-GBS therapies. One way GBS protects itself is through the synthesis of positively charged lipids. Lipids form the cell membrane, which is the fundamental barrier that defines the inside versus outside of the GBS cell and is a critical site of host-pathogen interactions. Bacteria typically have a negatively charged cell membrane, which serves as a target for positively charged peptides that are secreted by the immune system. These peptides create holes in the bacterial cell membrane, causing bacterial cell lysis and death. GBS can respond to stress by increasing the synthesis of positively charged lipids through the addition of positively charged lysine to certain negatively charged membrane lipids. The resulting lipids carry a more positive charge. Little is known about how GBS restores its membrane to its “normal” state after it escapes host defenses. GBS encodes a gene known as the alpha-beta hydrolase (ahyD) that we hypothesize is responsible for the turnover of positively charged lysine-lipids back to their original forms. We generated a GBS mutant lacking ahyD and assessed changes to the cell membrane using lipidomic analysis of cultures grown with lysine concentrations that are present at different infection sites. As expected, ahyD deletion altered the ratio of lysine-modified to unmodified lipids, supporting a role for this gene in GBS membrane “resetting.” We additionally determined that ahyD deletion increases the resistance of GBS to a cationic antimicrobial. Overall, our work has identified AhyD as a key player in membrane lipid dynamics in GBS.
Aprana Upuluri
Poster
155
The role of per and polyfluoroalkyl substances in membrane lipid remodeling and modulating antibiotic susceptibility
Abstract →
Per- and polyfluoroalkyl substances (PFAS) are a large class of synthetic fluorinated compounds widely used in industrial and consumer products, characterized by their chemical stability due to strong carbon–fluorine bonds and resistance to environmental degradation, making them persistent contaminants in water, soil, and biota. One of the best studied PFAS compounds is perfluorooctanoic acid (PFOA), an eight-carbon perfluorinated carboxylic acid. PFOA possesses surfactant-like properties that allow it to insert into or associate with lipid bilayers, leading to altered membrane permeability, disruption of proton gradients, and oxidative stress. In eukaryotes, PFOA disrupts lipid metabolism, while in bacteria its effects remain largely unexplored, though recent studies show gut bacteria can rapidly bioaccumulate PFAS at high intracellular levels. The cellular effects of PFOA and related PFAS, particularly their role in remodeling bacterial membrane lipids, remain poorly understood. To address this critical knowledge gap, this study employs genetic and lipidomic analyses to elucidate the cellular response exhibited by pathogenic bacteria in response to PFAS exposure. In our work, different bacterial species and yeast were cultured in growth media supplemented with PFOA and corresponding control conditions and then their lipids were analyzed using liquid chromatography-tandem mass spectrometry (LC-MS/MS). Strikingly, some bacterial species and yeast synthesize novel membrane lipids modified by PFOA. These results are significant because they establish a novel mechanism for the bioaccumulation of PFOA and, potentially, for bioremediation of PFOA in biological systems such as the human gastrointestinal tract. We additionally determined that PFOA exposure altered the antibiotic susceptibility of multidrug-resistant bacteria, demonstrating that synthesis of the novel PFOA-containing lipids by bacteria likely alters their cellular functions. Ongoing studies include the identification of the bacterial genes and biosynthetic pathways involved in the synthesis of novel PFOA-modified lipids.
Seyedeh Leila Mousavi
Poster
157
Reward Cues Trigger Neural Signals in the Entorhinal Cortex
Abstract →
Relapse in cocaine use disorder is strongly driven by environmental cues that reactivate drug-associated memories. The lateral entorhinal cortex (LEC) is important for associative memory and sends projections to the nucleus accumbens (NAc), suggesting a role in cue–reward encoding that may drive relapse. To test this, we used fiber photometry to record calcium activity in the LEC during the presentation of natural and drug reward predictive cues. In our first experiment, adult C57Bl/6J mice were tested for LEC activation during food-predictive cues. The LEC of mice was infused with a non-cell-specific virus expressing GCaMP (pAAV.CAG.GCaMP6f). Animals were later tested across self-administration, where cues coincided with sucrose pellet delivery, and then were tested on extinction and cue-induced reinstatement. We found that these animals readily acquired, extinguished, and reinstated responding for sucrose pellets. Fiber photometry showed significantly higher calcium-dependent activity (465 nm) compared with the isosbestic control (415 nm), with event-related increases around food-predictive cue presentations. We also tested a separate cohort of mice for activation of LEC neurons that project to NAc by cocaine-predictive cues. These mice received bilateral injections of a retrograde GCaMP virus (AAVrg-syn-GCaMP7f) into the NAc and optic fiber implantation in the LEC. Mice readily acquired intravenous cocaine self-administration, reduced responding during extinction, and reinstated responding when cocaine cues were presented. Fiber photometry recordings did not reveal consistent calcium signals in this cohort, highlighting the need for improved targeting of projection-specific populations. Together, these results show that the LEC is engaged by reward-predictive cues during natural reward (and potentially cocaine) seeking and provide a foundation for future work to determine whether LEC–NAc projections are also recruited during cue-induced cocaine relapse.
Yogeswar Reddy
Poster
159
TinyML for ECG Biometrics on Resource Constrained Devices
Abstract →
With the rapid advancements in Artificial Intelligence (AI) and deep fake technology, ensuring data security has become a paramount concern. Traditional authentication methods, such as face ID, iris, and fingerprints, are susceptible to compromise, highlighting the need for more robust biometric systems. Leveraging the unique characteristics of Electrocardiogram (ECG) signals, which vary significantly from individual to individual, shows the potential for enhancing authentication mechanisms. This paper introduces a novel methodology for the authentication of individuals using ECG signals in resource constrained devices like microcontroller units (MCUs) through the implementation of Tiny Machine Learning (TinyML). In this regard, 290 raw ECG signals of different subjects from the Physikalisch-Technische Bundesanstalt (PTB) database were filtered and segmented into 37,752 individual heartbeats. Subsequently, Spectrogram features were extracted to train a Dense Neural Network (DNN) machine learning model within the Edge Impulse platform, achieving an impressive accuracy of 97.6%. The trained model was then converted to a TinyML supported format and deployed onto Espressif ESP-EYE (ESP32 240MHz) devices for performance evaluation. Furthermore, this work evaluated the accuracy performance of spectral features in DNN and statistical, time domain features in the Support Vector Machine (SVM) model to complete result comparison
Upeksha C. Dissanayake
Poster
161
Computational Characterization of DNA Repair Enzymes: Insights into Cancer-Associated MutY Mutations and Mycobacterial Primase-Polymerase C
Abstract →
DNA repair enzymes are essential for maintaining genome stability, and their dysfunction is strongly linked to cancer and other diseases. This work presents computational investigations on two DNA repair enzymes: the Adenine-DNA glycosylase enzyme (MutY) and the mycobacterial Primase-Polymerase C (Prim-PolC).
MutY initiates the repair of mutagenic 8-oxoguanine:adenine (OG:A) mispairs through the base excision repair pathway. Mutations in the enzyme, such as R241Q and N238S, are associated with colorectal cancer. These variants alter a hydrogen bond bridge connecting the active site to the iron-sulfur cluster, disrupting enzyme function while maintaining DNA binding. Our computational studies provide insight into how these structural changes affect the dynamics of the structure and the connection among the residues involved in the H-bond bridge.
In parallel, we examined Prim-PolC, a member of the Prim-Pol family that combines primase and DNA polymerase activities to maintain genome stability. These enzymes play a vital role in genome maintenance pathways, including DNA replication, repair, and damage tolerance mechanisms. A unique structural feature of Prim-PolC is its C-terminal loop 3, which facilitates recognition of short DNA gaps during excision repair. Additionally, the R179 residue functions as a gating side chain, stabilizing the active site in a closed conformation favorable for efficient nucleotide incorporation.
Together, these studies reveal how mutations or unique structural features influence the molecular basis of DNA repair. The findings from molecular dynamics (MD) simulations and quantum mechanics/molecular mechanics (QM/MM) calculations will be further discussed and compared with experimental outcomes, providing mechanistic insights into genome maintenance pathways.
Hsin-Jung Tien
Poster
163
Exploring ALDH1A3-expressing cells in the murine oviduct
Abstract →
The fallopian tube in humans and the analogous oviduct in mice are specialized structures that are critical for fertilization and early embryonic development. The oviduct experiences damage upon exposure to the follicular fluid released from the ruptured ovarian follicle during ovulation. Immature cells, referred to as stem/progenitor cells, are the foundational cells in mature tissues that are essential for tissue repair and homeostasis. The identity of such immature cells in the oviduct remains largely unknown. Defining immature cells is not only important for understanding normal oviduct biology and reproductive abnormalities, but also aggressive ovarian cancer, which is increasingly recognized to originate in the oviduct. Aldehyde Dehydrogenase 1A3 (ALDH1A3) is a marker of progenitors in various tissues, but its presence in the oviduct has not been explored. In this study, we analyzed the expression, distribution, and capacity of ALDH1A3-expressing cells in the mouse oviduct through phenotypic and functional analyses. Through immunostaining, we observed the presence of ALDH1A3+ cells predominantly in the fimbriated region of the oviduct that is proximal to the ovary, and highly expressed in the parenchymal epithelial compartment compared to the surrounding stroma. Using flow cytometry based on the ALDEFLUOR assay, which measures ALDH activity, we detected ALDH+ cells in both oviductal epithelial and stromal subsets. Using Fluorescence-Activated Cell Sorting (FACS)-purified cells, we found that ALDH+ cells from both epithelial and stromal cell fractions exhibited a relatively higher clonogenic capacity in vitro than ALDH- cells indicative of increased progenitor activity. Further research into its functional potential and requirement in vivo, and underlying molecular mechanisms, will provide new insight into oviductal progenitors that can be tapped into for advancing therapeutic strategies in reproductive abnormalities and cancer.
PHD STUDENT SESSION 2 PARTICIPANTS
Mahdi Mosadegh
Poster
102
Revolutionary Debinding Process: Bringing 3D Printed Ceramics to Same-Day Dentistry
Abstract →
3D printing offers dentistry unprecedented design freedom, creating complex bridges, hollow structures, and patient-specific geometries impossible to achieve with traditional milling. However, a critical bottleneck has prevented 3D printed ceramics from reaching chairside dentistry: the debinding process. After printing, ceramic parts contain 40-60% polymer binders that must be carefully removed through heating, a process that traditionally takes 20-100 hours and consumes enormous energy. We have developed a revolutionary ultrafast thermal debinding (UFTD) process that transforms this bottleneck into a competitive advantage. Using vacuum pyrolysis combined with rapid heating via porous graphite felts, our method removes binder materials in under 30 minutes, up to 200 times faster than conventional processes. The technique also reduces energy consumption by over 3,500-fold while producing zirconia restorations with mechanical properties equivalent to traditionally processed ceramics. This breakthrough finally enables same-day 3D printed dental restorations, combining the speed dentists need with the geometric complexity that makes 3D printing unique. Dentists can now print intricate, personalized ceramic restorations, from multi-unit bridges to custom implant components, and deliver them to patients in a single appointment. This innovation doesn’t just improve existing dental workflows; it opens entirely new possibilities for personalized, complex ceramic restorations that were previously impossible or impractical to manufacture.
Rouzbeh Molaei Imenabadi
Poster
104
2023 Continuous Monitoring Of The Bladder Volume Using Non-Invasive Wearable Ultrasonic Sensor (NWUS)
Abstract →
Urinary catheters provide continuous, precise urine output (UO) measurement, crucial for critically ill patients or those with acute kidney injury in surgeries and ICUs, where renal function must be closely monitored. The catheter is connected to a closed collection system with volume markings, allowing healthcare providers to measure urine output over time, typically in milliliters per hour (ml/hr). This monitoring helps detect early changes in renal function. However, catheterization poses risks, as catheters can promote pathogen colonization, leading to catheter-associated urinary tract infections (CAUTIs). Given these risks, non-invasive approaches should be prioritized as the standard of care, particularly for severely ill patients during and after surgeries. Wearable ultrasound systems offer a promising alternative to catheters for continuous bladder monitoring, gaining attention for their non-invasive, real-time assessment capabilities. Current wearable ultrasound technologies face challenges such as high-power consumption, limited computational power, bulky designs, and low-resolution imaging. While these devices may alert users when the bladder is full, they lack the accuracy required for continuous bladder volume measurement, which is essential for diagnostics and treatment. Solutions like phased arrays improve precision but involve bulky electronics or wired connections. Conformable bladder patches provide in vivo monitoring but lack practical power circuitry and rely on A-mode imaging, limiting measurement precision. Existing solutions do not meet all the necessary conditions. Current commercial portable ultrasound devices, such as Verathon BladderScan, Butterfly IQ+, Philips Lumify, and GE Healthcare Vscan, offer robust diagnostic capabilities but fall short in terms of wearability due to their bulky designs. Wearable devices like DFree and SENS-U are limited to single tasks and use A-mode imaging with a few A-lines, providing rough bladder volume estimates that lack the precision needed for critical care. Research prototypes often exceed power budgets, lack wireless connectivity, are not optimized for wearability, and do not offer the accuracy required for precise bladder volume measurements. While recent studies have achieved low power consumption, raw data access, and wireless connectivity, their small aperture and shallow penetration depth make them unsuitable for effective bladder volume monitoring. Previous research introduced a low-power wearable ultrasound patch system with a compact form factor for physiological monitoring. Although the device is flexible and compact, it does not address the dual challenge of high-quality imaging and miniaturized electronics. Therefore, creating a wearable ultrasound system that efficiently collects and transmits data while operating on reduced power with high ultrasound penetration for accurate bladder volume assessment remains a significant challenge. Our main goal is to develop a wearable ultrasound sensor that can be worn via a belt, providing continuous, real-time bladder volume monitoring to reduce the need for catheterization. We hypothesize that this wearable ultrasound sensor will reduce the necessity for invasive catheterization, effectively mitigating catheter-associated urinary tract infections (CAUTIs). The compact, non-invasive design, coupled with continuous wireless communication capabilities, will allow the system to be practical for both clinical and at-home use. Demonstrating the viability of this solution will represent a significant advancement in non-invasive bladder monitoring and could establish a new standard of care for patients requiring frequent bladder volume assessment. This research introduces a wearable ultrasound sensor with machine learning innovation to reduce catheterization, as detailed in the attached document. The system is designed for real-time bladder volume monitoring, using a low-power ultrasound platform combined with FPGA-base.
Arefeh Mirsharifian
Poster
106
Removal of Per- and Polyfluoroalkyl Substances from Contaminated Groundwater using Cationic Metal−Organic Frameworks
Abstract →
Per- and polyfluoroalkyl substances (PFAS) are highly persistent and toxic environmental contaminants frequently detected in water sources, posing serious threats to both human health and ecosystems. Conventional treatment methods, such as activated carbon and ion exchange resins, often show limited effectiveness across the diverse range of PFAS compounds. Metal–organic frameworks (MOFs) have emerged as promising adsorbents for PFAS removal due to their crystallinity, porosity, and high surface area. Cationic MOFs have strong electrostatic interactions with anionic species. In this study, we evaluated the performance of selected cationic MOFs for the adsorption of PFAS mixtures from aqueous systems. Our results demonstrated removal efficiencies of up to 99.9%, underscoring the strong potential of these materials for effective PFAS remediation. These findings highlight the promise of cationic MOFs as viable and efficient adsorbents to help mitigate the growing environmental challenge posed by PFAS contamination in water.
Charles Cahill
Poster
108
Elucidating the Role of the Yeast Asf1 Disordered C-terminal Domain in Binding Histones H3-H4
Abstract →
Intrinsically disordered regions (IDRs) are regions within proteins that do not have inherent secondary structures but can adopt transient structures dependent on context, such as when binding another protein. The resulting flexibility may enable disordered proteins to engage in tunable and multivalent interactions with numerous binding partners. As such, these proteins are often leveraged to fulfill important roles in cellular processes, from cell signaling pathways to transcriptional and translational regulation. One such protein is the histone chaperone Asf1, which has a disordered C-terminal tail that may play a role in binding the H3-H4 dimer. In Saccharomyces cerevisiae Asf1 (yAsf1), the tail is also highly acidic, in addition to being disordered. It is theorized that this acidity may aid the tail in binding basic histones, and, more broadly, acting as a DNA mimic to engage the DNA-binding interfaces of histones. We used a modified hydrogen/deuterium exchange coupled with mass spectrometry (HDX-MS) workflow to study changes in deuterium uptake of Asf1 alone (apo) and the Asf1-H3-H4 complex to elucidate the role of the tail in binding to H3-H4. We further employed a truncated mutant of Asf1 (yAsf1∆CTD) to discern the changes in deuterium uptake of H3-H4 due to the presence of the Asf1 tail. Our work establishes a modified HDX workflow for the analysis of disordered domains and reveals the conformational and allosteric changes that occur in histone binding due to the Asf1 tail.
Oluwaseun Adeyemi
Poster
110
Stable Carbon Pools Associated with Distinct Mineral Phases in Soils: A Cross-Ecosystem Study Using Targeted Dissolution Extractions
Abstract →
Understanding the fate of soil organic matter (SOM) is critical for projecting carbon cycling and climate feedback. We assessed carbon pools associated with minerals across five geochemically distinct soils: agricultural, clay-rich, peat, permafrost, and sediment, using mineral dissolution extractions, including water, hydrogen chloride (HCl), sodium hydroxylamine, sodium dithionite, and sodium pyrophosphate. Our study found that sodium pyrophosphate mobilized the highest carbon concentrations across all soils, with maxima near 25 mg/g in permafrost and clay-rich soils, suggesting abundant mineral-associated organic matter (MAOM). Sodium pyrophosphate chelated and dispersed Fe and Al minerals, thereby resulting in the release of carbon that was complexed with these mineral phases. These mobilizations coincided with the release of Fe, Al, and Si, underscoring the dominant roles of ligand exchange, cation bridging, and mineral occlusion in SOM stabilization. Dithionite and 0.5 M HCl each released substantial amounts of carbon from the same soils (up to ~15 mg/g). Dithionite primarily targets and dissolves crystalline phases, while 0.5 M HCl targets short-range ordered (SRO) phases, as well as Fe²⁺ bound to particle surfaces. The extracted carbon likely includes organic matter occluded within or sorbed to these reactive Fe and Al phases. The C: N ratios of the solubilized SOM ranged from 15.57 to 30.45, reflecting differences in chemical quality and microbial accessibility. C: N ratios of extracted SOM ranged from 15.57 to 30.45, reflecting chemical quality and microbial accessibility. HCl-extracted SOM had the highest C:N ratio (30.45), suggesting N-poor, mineral-bound carbon vulnerable to immobilization. In contrast, dithionite extractions (C: N ~15.6) indicated more microbially accessible SOM, particularly in Fe-rich clay and permafrost soils. Mineralogical analyses (XRD, SEM) confirmed reactive mineral frameworks enabling SOM stabilization. These findings underscore the abundance of MAOM pools in permafrost and clay-rich, with consequences for long-term carbon stability. Our work presents a geochemically grounded framework for quantifying MAOM pools in soil. This framework has potential to examinee how these MAOM pools response to changing environments.
Sina
Khalesidoost
Poster
112
Comparison of NdFeB, Ferrite, and Hybrid Designs for a Magnetic Continuously Variable Transmission
Abstract →
Magnetic continuously variable transmissions (MCVTs) enable variable gear ratios, which can connect fixed-speed machines with variable-speed motion. While rare-earth permanent magnets (PMs) like NdFeB enhance torque density, their high cost and limited availability raise concerns. Ferrite PMs offer a cost-effective but much weaker alternative, while hybrid designs combine both materials to balance performance and cost. This paper investigates four MCVT designs: NdFeB-only, ferrite-only, and two hybrid configurations. Results indicate that NdFeB-only designs achieve the highest torque density but at a significantly higher material cost, while ferrite-only designs are less expensive but have lower torque density. Both hybrid topologies can achieve intermediate VTDs and material costs between the NdFeB-only and ferrite-only designs. The hybrid topology with the ferrite PMs in series with the NdFeB PMs yields higher VTDs and higher material costs than the hybrid topology with the ferrite PMs in parallel with the NdFeB PMs. To account for end effects, the torque was evaluated in 3D and compared with 2D results. Hybrid designs tend to have less torque reduction from 2D to 3D, primarily due to differences in Rotor 1 and Rotor 2 thickness. Thicker modulators and spokes lead to greater torque reduction across all designs due to reducing the reluctance for escaping flux. Torque ripple was calculated for each design, showing relatively low values less than 3.5% for all cases. An efficiency analysis was also performed, revealing that copper loss is the dominant component of total loss at the evaluated speeds. Thus, ferrite designs generally have lower efficiency due to higher copper losses.
Hadi Ismail
Poster
114
Interleaved Dual 3-phase Machine Inter-Turn Short-Circuit Fault Mitigation Topology Using Current Source Inverter Drive
Abstract →
Inter-turn short-circuit (ITSC) faults are among the most critical failures in electric machines, in particular surface-mounted permanent magnet synchronous machines (SPMSMs), driven by their low inductance. These faults create highly excessive circulating currents within the machine, producing localized hot spots that accelerate insulation breakdown in adjacent windings. Left unmitigated, ITSC faults can rapidly escalate into line-to-line, line-to-ground, or even three-phase faults, posing severe safety risks, including fire. Existing mitigation techniques—such as phase redundancy, terminal shorting, negative d-axis current injection, and electrical/mechanical shunts—can limit ITSC effects but often only at high fault currents and with trade-offs in complexity in design and control, output power, reliability, and electromagnetic performance. In contrast, the proposed method reduces ITSC fault currents immediately and significantly (~15% of rated current) while preserving near-normal machine operation. The proposed method relies on the motor special design which converts any turn-to-turn fault into ph-ph a phase-to-phase short circuit, and the fault current must pass through the motor terminals. Dual three-phase current source inverters, each having the DC link inductance split between the positive and negative sides, have been used to greatly reduce this fault current. Its effectiveness has been validated in both steady-state and dynamic conditions, with a systematic study of design parameter impacts. Comparative simulations with conventional methods further demonstrate the robustness and advantages of this approach.
Yuwei Zong
Poster
116
Trustworthy Uncertainty Quantification via Distributionally-Robust Dual Adaptive Conformal Prediction
Abstract →
Uncertainty quantification (UQ) plays a vital role in ensuring reliable and consistent decision-making, especially in emerging domains such as interactive, data-driven modeling and simulation for digital twins, where challenges such as distribution shifts, dynamic adjustments, and deep uncertainty limit the effectiveness of traditional Bayesian methods that rely on prior distribution assumptions. Conformal prediction (CP) offers a distribution-free framework for UQ with guaranteed marginal coverage. While basic methods like split CP and adaptive CP address some practical concerns, they often suffer from unstable prediction intervals and degraded coverage under distribution shift or dependence among samples. To overcome these limitations, we propose a Dual Adaptive Conformal Prediction method that introduces a dynamic data partitioning mechanism to adaptively adjust conformity scores and interval widths based on observed data characteristics and optimize information allocation for improved predictive uncertainty estimation. This dual adaptation improves the adaptability of uncertainty estimation models, increasing their sensitivity to data variations, and improving the stability of prediction results. To evaluate the effectiveness of the proposed method, experiments are conducted on both exchangeable and non-exchangeable datasets across low- and high-dimensional settings. Experimental results demonstrate superior expected coverage and greater stability in prediction intervals compared to traditional CP and Bayesian methods, especially in enhancing the flexibility of CP methods for non-exchangeable data. The proposed approach significantly strengthens conformal prediction’s applicability to real-world, non-ideal conditions—offering a promising direction for future UQ research.
Azin Hojjati
Poster
120
A Compact Near-Field Horn-Fed 3-Bit Phase Gradient Metasurface Lens with Low F/D for Low-Cost Efficient Focusing in ISM-Band
Abstract →
Metasurface-based lenses offer strong potential for enhancing wireless power transfer (WPT) by concentrating electromagnetic energy at targeted focal regions. Conventional designs apply ideal phase compensation assuming planar wavefronts, typically requiring far-field excitation or bulky beam-shaping elements, which limit compactness. In this work, we demonstrate a compact near-field WPT architecture by directly illuminating a passive metasurface lens with a standard horn antenna at a short standoff of 5λ at 24 GHz, achieving a high normalized focal length and an ultra-low F/D ratio of 0.5. The lens comprises a 20×20 array of cross-shaped unit cells on a single-layer, double-sided FR-4 substrate, and employs a 3-bit (8-level) quantization scheme to realize the desired phase profile. Each unit cell occupies λ/2 × λ/2 × λ/8, enabling efficient subwavelength sampling. Full-wave simulations and experimental results confirm strong focal confinement, with FWHM-based analysis validating the lens’s suitability for compact and efficient wireless power transfer.
Shubham Chatterjee
Poster
122
Studying Proteins for Applications from Industrial Wastewater Treatment to Cancer Therapy through the Lens of Computational Chemistry
Abstract →
Proteins, in the form of enzymes, are nature’s catalysts, enabling chemical reactions that would otherwise be unfeasible. This poster presents studies on enzymes with applications from industrial wastewater treatment to cancer therapy, employing existing computational chemistry methods while also developing new ones. Specifically, this poster highlights work on the enzyme Horseradish Peroxidase (HRP), which oxidizes industrial pollutants, where predictions on its catalytic cycle are in strong agreement with experimental observations. This poster further shows how external electric fields may influence catalysis in proteins. Next, this poster highlights studies on Poly(ADP-ribose) Polymerase 1 and 2 (PARP1 and PARP2), which are essential therapeutic targets in multiple cancer treatments. Predictions are presented on whether FDA-approved drugs would bind to a cancer-associated variant of PARP1, along with insights into how predictions on a cancer-related PARP2 variant may aid the design of improved inhibitors. Finally, this poster discusses efforts to calculate/predict enzymatic reaction barriers and reaction energies using Free Energy Perturbation (FEP) method within TINKER molecular dynamics package.
Xinyuan Zhang
Poster
124
Instantaneous volumetric light-sheet imaging of beating heart
Abstract →
Cardiac arrhythmia is a prevalent and life-threatening condition often linked to drug-induced alterations in cardiac electrophysiology. Identifying arrhythmogenic compounds early remains a significant challenge, as arrhythmias can be sporadic and asymptomatic until severe events occur. Zebrafish (Danio rerio) have emerged as a powerful in vivo model for drug screening, owing to their optical transparency and genetic similarity to humans. However, current imaging modalities face challenges in capturing the full complexity of the beating heart at cellular resolution, especially during transient arrhythmic events. By integrating compressed sensing (CS) techniques with rapid volumetric scanning in light-sheet microscopy (LSM), we overcome the traditional trade-off between imaging speed and axial sampling rate, enabling instantaneous volumetric imaging of the zebrafish heart. Specifically, by leveraging a digital micromirror device (DMD) to encode multiple axial planes into single camera frames and employing a plug-and-play alternating direction method of multipliers (PnP-ADMM) algorithm, we reconstruct 4D heart (3D space + time) from sub-Nyquist samples. Achieving imaging rates of 100 volumes per second (vps), this approach enables direct observation of cardiac contractile activity at single-cell resolution. This capability allows us to unveil the underlying mechanisms of arrhythmogenesis and facilitates early drug screening.
Marvin Alvarez
Poster
126
Dual-Task Cost of Sit-to-Walk in Older Adults: Preliminary Results
Abstract →
Introduction: Sit-to-walk is a fundamental activity of daily living for people of all ages. It is not merely a sequential combination of sit-to-stand and subsequent gait initiation, but rather a cohesive, fluid transition between the two movements. Recent studies have examined various biomechanical characteristics of sit-to-walk, such as the total duration, phase durations, and the trajectories of the center of mass [1,2]. Notably, sit-to-walk hesitation, defined as a reduction in forward center-of-mass velocity during sit-to-walk, has emerged as a significant quantitative indicator of the fluidity of this movement [3]. Particularly, sit-to-walk hesitation may indicate age-related mobility decline. However, to date, studies have predominantly examined hesitation and other sit-to-walk variables under single-task conditions. This study aims to compare sit-to-walk hesitation between single-task and dual-task in healthy older adults and to contrast these findings with those from healthy young adults. Methods: We conducted a preliminary analysis for an ongoing study. We recruited 28 healthy young adults (11 females; mean age = 22.4 ± 5.2 years; mean body mass index = 23.9 ± 2.7 kg/m2) and one healthy older adult (male; 75 years; body mass index = 26.8 kg/m2). All participants were cognitively intact, as assessed by the Montreal Cognitive Assessment, and had no history of neurological disorder. To perform sit-to-walk, participants were asked to walk out from an armless chair in a single fluid motion under two conditions: without any concurrent task (single-task) and while performing a concurrent mathematical processing task (dual-task). Each condition was repeated three times. During sit-to-walk, we collected data from 74 reflective markers using a 16-camera motion capture system (Vicon, UK). Data analysis was performed using Visual3D (HAS Motion, Canada). We created a 15-segment biomechanical model and calculated trajectories of the center of mass. We measured sit-to-walk hesitation and the total duration of the sit-to-walk transition. Statistical analysis was performed to compare sit-to-walk hesitation and duration between single-task and dual-task conditions within health young adults using paired t-tests. Differences in these variables between young and older adults under both task conditions were assessed using unpaired t-tests. Results & Discussion: In healthy young adults, sit-to-walk hesitation significantly increased from the single-task condition (30 ± 20%) to the dual-task condition (40 ± 20%) (p < 0.05). Similarly, sit-to-walk duration also significantly increased from single-task (1.8 ± 0.3 seconds) to dual-task (2.0 ± 0.4 seconds) in healthy young adults (p < 0.05). In the single healthy older adult, sit-to-walk hesitation increased from single-task (43 ± 5%) to dual-task (58 ± 8%), and sit-to-walk duration also increased from single-task (1.8 ± 0.1 seconds) to dual-task (2.0 ± 0.1 seconds). However, these changes were not statistically analyzed due to the small sample size. One of the most interesting findings was the difference in sit-to-walk hesitation between the single older adult and the group of young adults. Notably, the older adult’s hesitation during the single-task condition was greater than that observed in the young adults during the dual-task condition. This suggests that even in standard conditions, older adults may experience challenges that are exacerbated under dual-task demands typically observed in younger individuals. The increase in hesitation and duration from single-task to dual-task conditions in young adults indicates that dual-tasking imposes additional cognitive and physical demands, affecting completing sit-to-walk. This aligns with previous research documenting dual-task interference in motor tasks. For older adults, while the trend suggests a similar increase in hesitation and duration under dual-task conditions, the findings should be interpreted cautiously due to the sample size.
Angeloh Stout
Poster
128
Detecting Emotions from Gait Biomechanics: A Machine Learning Approach​
Abstract →
Emerging research suggests that emotions influence the way people walk, positioning gait as a potential modality for emotion recognition. Compared to traditional approaches such as facial expressions and speech, gait-based detection may offer advantages because it does not rely on visible faces or spoken words, which may be unavailable or socially constrained. This study examined the feasibility of recognizing emotional states using 3D gait biomechanics and machine learning. Fifteen healthy young adults participated in gait trials while recalling autobiographical memories designed to elicit five target emotions: anger, sadness, joy, fear, and neutral. Gait biomechanics were captured using a 3D optoelectronic motion capture system, from which 155 biomechanical variables were extracted. To evaluate emotion recognition, five machine learning algorithms—K-Nearest Neighbors, Logistic Regression, Random Forest, Multi-layer Perceptron, and eXtreme Gradient Boosting (XGBoost)—were implemented with Leave-One-Participant-Out cross-validation. Class imbalance was addressed using the Synthetic Minority Over-sampling Technique. Results showed that emotional states could be classified at a level exceeding chance (59% vs. 25%), with XGBoost achieving the highest accuracy (59%) when using the top 20 biomechanical features ranked by a decision tree entropy index. These findings suggest that gait analysis, paired with machine learning, represents a promising alternative pathway for emotion recognition. Beyond theoretical implications, this work may lay the groundwork for novel tools aimed at tracking emotional fluctuations, with potential applications in mental health monitoring and early detection of conditions such as bipolar disorder.
Meghraj Magadi Shivalingaiah
Poster
130
Machine-learning Assisted Discovery of Absorbing Molecules for Tissue Clearing
Abstract →
Optical clearing enables deep imaging in biological tissues, but performance depends critically on dye absorption spectra. In practice, dyes with strong, narrow absorption bands improve transparency and contrast by selectively shaping residual optical attenuation in cleared tissues, as seen in our prior work. Acquiring full molar extinction coefficient (MEC) spectra for each candidate dye is slow and limits screening throughput. We present a data-driven framework that predicts a molecule’s complete absorption spectrum from structure to accelerate dye selection for tissue clearing. Starting from each molecule’s SMILES identifier, we generate cheminformatic fingerprints and train a supervised regressor to map fingerprints to MEC values across wavelengths; the trained model then predicts full spectra for unseen molecules. Our dataset consists of spectra for a few dyes measured in-house via spectrophotometry, enabling held-out evaluation at the molecule level. We assess recovery of peak positions and band shapes relevant to clearing performance. This approach reduces dependence on serial dilution and instrument time, integrates naturally with tissue-clearing workflows, and has broader utility in chemical sensing, environmental monitoring, and optoelectronic devices such as organic photovoltaics and photodetectors.
Rohit Kajla
Poster
132
Limits of transparency in biological tissues via Refractive Index matching
Abstract →
Biological tissues strongly scatter light due to refractive index variations at cellular and sub-cellular scales. This turbidity restricts optical imaging and therapeutic depth, limiting the effectiveness of light-based biomedical techniques. Tissue clearing strategies, such as immersion in refractive index–matching agents, can reduce scattering and improve transparency. However, the ultimate physical limits of how transparent a tissue can become remain largely unexplored. In this work, we investigate tissue as a model system for understanding these transparency boundaries. Using optical clearing agents (OCA), we examine how the optical properties of tissues are modified. By extending the study across multiple species, we aim to identify universal and species-specific behaviours that shape the optical transparency of biological tissue. Our results provide new insight into the interplay of scattering, absorption, and intrinsic tissue structure in defining transparency limits. This framework contributes to the physics of light–tissue interactions and has implications for optimizing biomedical imaging systems and clearing protocols.
Fariha Taskin
Poster
134
Bayesian Order-based Structure Learning with Zero-Inflated Poisson Bayesian Networks for Multiple DAGs
Abstract →
Causal structure learning in complex, heterogeneous systems with multivariate zero-inflated count data—where variables interact through directed acyclic graphs (DAGs)—remains challenging. Classical approaches struggle with identifiability under Markov equivalence, computational burden during graph search, and the unrealistic assumption of a single homogeneous DAG. We introduce a unified, order-based ZIPBN framework that jointly infers multiple DAGs across datasets for multivariate zero-inflated count data and establish its identifiability from cross-sectional observations by utilizing Markov equivalence properties that extend to broader exponential-family distributions. Our approach places a joint posterior over DAGs and their topological orders, using an order-based Metropolis–Hastings scheme that proposes and updates a common order. Conditional on a sampled order, we update dataset-specific structures and parameters, selecting relevant variables and edges per dataset. Comparing sampled graph sizes and (log) joint posterior scores over graphs and orders facilitates reliable recovery of the true causal order and corresponding multi-dataset DAGs, reducing computational complexity and accommodating heterogeneity. Simulation studies show accurate recovery of orders and edges, robustness to zero inflation and dataset heterogeneity, and superior performance compared to Bayesian network baselines.
Drew Miles
Poster
136
Design of Automated Custom Orthoses System
Abstract →
Introduction: 3D printing has revolutionized orthoses manufacturing in assistive technologies, enabling development of custom fitted orthoses in under 24 hours. Physician-engineer collaboration is essential to addressing the mechanical complexities of automated orthoses design. Single-step manufacturing of these devices is necessary for simplistic and translational technologies. Purpose: This work proposes a novel automated program, application programing interface (API), to design a custom single finger orthosis using only basic biometric data (body weight, height, and age) as input to enable fully virtual rehabilitation. Study design: Feasibility study, cross-sectional design, survey study. Methods: Customization without body scans streamlines programming and enables gross estimate of hand orthosis size. Computer-Aided Design (CAD) models are automatically generated using the Automated Custom Finger Orthoses (ACFO) system that was created using several macros with SolidWorks and Visual Basic Editor. A military database was used for the biometric data for the program to establish a relationship between finger size and biometric data. Survey data was taken from 10 human healthy participants (6 male and 4 female , aged between 18 to 60) using Likert scales to check fitness of 3D printed soft index finger orthosis part. Results: CAD models were successfully created using the API using military database, the models were 3D printed and fitted on subjects. The fitness of the device varies significantly under the current parameters. Eight of the ten test devices fit their test subjects with overall average offsets of 7.00 mm, 6.14 mm, 8.89 mm for DIP (L1), PIP (L2) and MCP joints (L3 offset) respectively and the subjects were fully donned. The biasing of current data towards male subjects results in slightly better fit accuracy in males than in females. Discussion: Additional and future revisions of the ACFO system can implement larger datasets including gender for calibrating the code and estimating specific finger lengths. The ACFO system provides an inexpensive method of manufacturing and deploying hand orthoses. Conclusions: Anthropometric studies can aid in discerning fundamental ratios for orthoses design, allowing for custom fitted designs and manufacturing with ease and efficiency.
Gabriella Putri
Poster
138
SharpEarthGAN: A GAN-based Method for Multi-image Super-resolution (MISR) Reconstruction of Earth Observation (EO) Imagery
Abstract →
Resolution is a critical factor influencing the accuracy and effectiveness of object imaging and detection tasks, especially in observing geophysical and geological features of the Earth’s surface. Traditional approaches to enhance resolution are computationally expensive and often impractical at global scale. Recent advances in generative artificial intelligence (AI) provide an alternative by directly learning resolution enhancement from data. We introduce SharpEarthGAN, a conditional generative adversarial network (GAN) framework for multi-image super-resolution of satellite imagery trained on real low- and high-resolution pairs. The generator employs residual blocks with a dual-normalization strategy, incorporating magnitude-based Fast Fourier Transform (FFT) and Meta normalization (MetaNorm), designed to stabilize spectral distribution and adaptively refine features. Information flow is preserved through local skip connections within residual blocks and a global skip connection that bypasses the residual stack and links directly to the post-residual stage, which then performs progressive upsampling. To improve geometric fidelity, Jacobian (edge gradient) and Hessian (curvature) losses are applied on the generator outputs and residuals, encouraging sharper boundaries and improving overall structural fidelity. The discriminator adopts a PatchGAN architecture, with residual blocks stabilized by spectral normalization and extended with a convolutional block attention module (CBAM) enhanced by Jacobian- and Hessian-guided spatial priors. This design improves sensitivity to edge inconsistencies and geometric artifacts in generated images. Experimental results show that SharpEarthGAN achieves sharper textures and boundaries with improved perceptual quality, with gains in SSIM and LPIPS while maintaining competitive PSNR. The framework provides a scalable and transferable solution for Earth Observation (EO) applications, including environmental and geological monitoring and change detection.
Preethi Parupalli
Poster
140
Brown Adipose Tissue FASN Deficiency Rescues Alcohol Induced Hypertriglyceridemia and Liver Damage through Enhanced FGF21 Expression and Thermogenesis
Abstract →
Alcohol overconsumption induced metabolic disorders are increasing worldwide. Specifically, elevated circulating free fatty acid (FFA) and triglyceride (TG) levels have been observed in humans and animal models drinking excessive alcohol leading to alcohol-associated liver disease (ALD). Thermogenesis is a process where the FFAs are utilized by brown adipose tissue (BAT) to generate heat. Understanding of the mechanisms involved in the systemic cross talk between these metabolically active tissues like BAT and liver are still in primitive stages. Recent studies have shown that inhibiting fatty acid synthase (FASN), a key enzyme in de novo lipogenesis (DNL), in adipose tissues increases thermogenesis and uncoupling protein 1 (UCP1) levels. Therefore, we hypothesize that inhibiting BAT FASN can enhance thermogenic action and rescue dyslipidemia induced by alcohol. My study shows that it indeed enhances the thermogenesis and alleviates Hypertriglyceridemia and ALD.
Alex Dao
Poster
142
A novel study on hepatocyte FASN deletion in alcohol induced hepatic steatosis and liver inflammation
Abstract →
Alcohol consumption is typical in many cultures, but its harmful effects are significant, leading to over 200 diseases and 2.6 million deaths globally. Alcoholic-Associated Liver Disease (ALD) develops in stages, beginning with fatty liver and potentially progressing to cirrhosis and liver cancer. Research shows increased lipogenesis is characteristic of ALD, with fatty acid synthase (FASN) being a crucial enzyme. Moreover, deleting FASN in hepatocytes can reverse steatosis and inflammation in obese mice. Interestingly, my preliminary data indicated that FASN deficient mice fed alcohol accumulated more triglycerides but exhibited lower levels of inflammatory gene expression in the liver. These findings suggest a need for further investigation into the underlying mechanisms. The results of this study may provide a new therapeutic strategy for combating ALD.
Lulu Eisenberg
Poster
144
Exploring the relationships between objective and subjective skills, socioeconomic status, and risky decision making
Abstract →
OBJECTIVE: This exploratory study leverages existing data to explore the relationship between objective and subjective graph literacy, numeracy, socioeconomic status (SES) and risky decision making. Understanding graphs and numbers is essential for navigating everyday decisions, from managing finances to interpreting health risks. Socioeconomic disparities may limit opportunities to develop these skills, yet the relationships between SES and both graph literacy and numeracy remain unclear. Here, we examine how these constructs relate to both objective and subjective SES and whether they relate to decision making. METHOD: Using datasets from two adult lifespan (N = 78; N = 119; 25-85) studies, we explored correlations between graph literacy, numeracy, socioeconomic status, and gambling acceptance rate. Objective graph literacy (OGL) was assessed with the Short Graph Literacy scale, subjective graph literacy (SGL) with the Subjective Graph Literacy scale, objective numeracy (ON) with the Berlin Numeracy Test, and subjective numeracy (SN) with the Subjective Numeracy scale. SES was assessed as objective (income and education composite) and subjective (self-perceived rank in country and community composite). Acceptance rate was calculated as the proportion of gambles accepted. RESULTS: We explored the relationships between four individual difference measures (OGL, SGL, ON, SN) with two measures of SES (objective and subjective) across two studies. Sixteen Pearson’s correlations were run (4 measures x 2 SES variables x 2 studies; Bonferroni correction α = .0031). In study 1, only subjective numeracy was significantly related to objective SES, r(76) = .365, p = .001. In study 2, only subjective numeracy was significantly related to objective SES (r(117) = .329, p < .001) and subjective SES (r(117) = .298, p < .001). We also explored the relationships between the four individual difference measures and gambling acceptance rate. Eight Pearson’s correlations were run (4 measures x 2 studies; Bonferroni correction α = .00625). None of the measures showed a significant relationship with acceptance rate. CONCLUSIONS: Although objective skills are often assumed to be shaped by SES, this exploratory work shows they are SES-independent. Rather, it’s people’s perceptions of their numerical skills that are tied to SES. These self-perceptions may impact the decisions people make, potentially reinforcing SES disparities not by objective skill gaps, but by psychological barriers. Graph literacy and numeracy were not related to gambling acceptance rate, indicating that neither numerical/graphical skills nor self-perceptions were related to risky financial decision making.
Dannie Zhabilov
Poster
146
Potential Phytopathogen Antimicrobial Agents Targeting the 2-Methylcitrate Pathway Enzymes AcnD and PrpD in P. aeruginosa
Abstract →
Pseudomonas aeruginosa is a phytopathogen that causes soft rot in a wide variety of plants including white ginger, spinach, and onions1–4. Its rising multidrug antibiotic resistance necessitates novel control strategies5. One possible target is its metabolic pathway which breaks down propionate for energy production utilizing an aconitase-like enzyme, AcnD6,7. However, this pathway has an enzymatic redundancy via prpD, a 2-methylcitrate dehydratase8. Inhibiting this pathway causes propionate buildup, which is toxic to the bacteria, allowing for potential antimicrobial prevention of soft rot7. This study aimed to determine the concentrations needed to produce antimicrobial effects of compounds identified via machine learning and tested based on their Vina docking scores9. Wild type P. aeruginosa, and mutant knockout strains, AcnD, prpD, and a double KO, were evaluated for growth inhibition on agar plates and in a plate reader with sub-lethal propionate. In vitro inhibitory assays of purified prpD were done using spectrophotometry. It has been revealed that Mesaconic Acid has significant inhibition of prpD compared to 2,3-dihydroxy-2-methylbutanedioic acid (2,3-DHMB), which had higher docking scores, though both have the ability to lead to cell death. Not only do the results below indicate that these compounds can inhibit growth of P. aeruginosa, but due to sequence conservation of AcnD with other phytopathogens, like Candidatus Liberibacter, which causes Citrus Greening Disease, these compounds could also provide broad-spectrum applications10. Thus, exploiting propionate metabolism via PrpD and AcnD presents a promising strategy for novel antimicrobial development.
Yihan Liu
Poster
148
REGULATORY NETWORK EFFECT ON E. COLI PHYLOTRANSCRIPTOMIC GROUPS
Abstract →
Urinary tract infections (UTIs) can affect any part of the urinary system, with Escherichia coli (E. coli) phylogenetic group B2 being the most frequent and virulent cause. B2 strains show distinct genomes and transcriptomes compared to groups A and D, with higher DNA gyrase activity and intracellular putrescine levels. Urea also influences these regulatory networks by inhibiting putrescine synthesis, increasing Mg²⁺ transport. Despite some relavant genomic studies, the link between transcriptomic differences and DNA supercoiling remains unclear. This study investigates the key factors driving transcriptomic variation between B2 and AD strains using RT-qPCR and RNA sequencing.
Victor Nguyen
Poster
150
Biomechanical Phenotyping Reveals Unique Mechanobiological Signatures of Early-Onset Colorectal Cancer
Abstract →
Colorectal cancer ranks as the third commonly diagnosed cancer worldwide and the second leading cause of cancer-related mortality in the United States. What is interesting is that, while the rate of average-onset colorectal cancer (AO CRC, diagnosed after 50 years old) has been declining, early-onset colorectal cancer (EO CRC, diagnosed before 50 years old) has been steadily increasing at an alarming rate. As AO and EO CRC are genomically similar, the reason for the rise of EO CRC has been poorly understood. Recent literature has supported that EO CRC may be driven by a combination of external exposures (such as diet and physical activity) and internal exposures (such as the gut microbiome). However, the tumor microenvironment of EO CRC has been relatively unexplored compared to AO CRC. Therefore, in this study, we quantified the biomechanical properties of tumor and matched normal tissue samples AO and EO CRC patients. We found that the tissues from EO CRC patients are stiffer and more fibrotic, primarily through the remodeling of collagen, a key structural component of tissue. These biomechanical alterations cause abnormal cell behavior, elongating cells and activating signaling pathways that drive cancer growth. In particular, we identified the YAP protein, a key regulator of mechanotransduction, as a key mediator of stiffness-dependent cancer growth in EO CRC. To further explore this mechanism, we engineered a model of CRC, where we isolated the effect of stiffness of cancer behavior using a synthetic hydrogel and quantified changes in cancer cell behavior. Through this, we validated that the proliferative effects of increased stiffness on CRC cells are largely mediated by YAP. By integrating biomechanical measurements, microstructural imaging, cellular phenotype analysis, genetic sequencing, and engineered in vitro models, we show that collagen remodeling drives tissue stiffness in EO CRC. In turn, this enhances mechanotransductive signals in cancer cells that lead to increased cancer growth, contributing to the distinct pathophysiology of EO CRC. Understanding how mechanical cues influence CRC cell behavior could inform biomarkers to stratify cancer risk, particularly in younger populations.
Coleman Moss
Poster
152
AI and Lasers: A Smarter Way to Study Wind Farms
Abstract →
Wind farms play an essential role in increasing green energy usage. However, the chaotic nature of the atmosphere and the large size of wind farms, coupled with the cost of traditional measurement techniques such as large meteorological towers instrumented with wind anemometers, make studying wind farms challenging. New developments in the technology of light detection and ranging (LiDAR) make measuring large regions of the atmosphere much easier. Combining the data from LiDARs with massive amounts of data collected from operating wind turbines is still difficult, but new advances in machine learning (ML) are perfectly suited to processing big data in efficient ways. We present the different ways the WindFluX lab at UT Dallas combines LiDAR measurements and ML to study wind farms along with new insights into the interaction between wind turbines and the atmosphere.
Andrew Glick
Poster
154
Tensional Homeostasis in Soft Tissues Arises from a Balance Between Cell Contractility and Extracellular Matrix Densification
Abstract →
To ensure optimal physiological function, cells maintain mechanical homeostasis around a tissue-specific set-point. Such a homeostatic set-point is thought to be established during embryonic development, maintained through cell-matrix interactions and eventually disrupted by maladaptive tissue remodeling. In disease progression, the biomechanical function of tissues can be lost due to excessive production and reorganization of a collagen-rich extracellular matrix (ECM) by resident fibroblasts, leading to fibrosis in vital organs. Therefore, a better understanding of tensional homeostasis is of the essence for improving current antifibrotic treatments. However, it remains unclear how interactions between fibroblasts and a complex network of extracellular collagen give rise to a mechanically homeostatic state compatible with normal tissue function. Much of what is known about mechanical homeostasis of tissues under tension, or “tensional homeostasis”, is based on experiments on tissue equivalents (fibroblast-populated collagen gels). Tissue equivalents cultured in the presence of a physical constraint at their boundary develop tensile forces under static conditions, suggesting the existence of a homeostatic tensile force. Here we present a biomechanical bioreactor fully integrated with a confocal microscope. Our platform carries out dynamic measurements of force evolution, structural remodeling of collagen, and overall tissue cross-sectional area, thereby allowing for the estimation of dynamic stress generation in tissue equivalents. In this study, we used our bioreactor platform to test the hypothesis that fibroblasts achieve tensional homeostasis by maintaining a constant state of stress in different collagen microenvironments. By culturing NIH/3T3 fibroblasts in increasing concentrations of collagen, we observe that cell-generated forces reach a plateau after 48 hours. Consistent with previous reports, the steady-state force value increases with increasing initial collagen density. At the same time, the cross-sectional area decreases with time in a concentration-dependent manner, such that the true collagen density increases over time. By measuring directly both force (f) and cross-sectional area (a), our bioreactor allows us to estimate the true stress (σ = f /a) developed in the tissue over time. Contrarily to our initial hypothesis, we found that stress is not conserved. Instead, tissues with the lowest initial collagen density develop the highest stresses due to higher densification and earlier alignment of collagen fibers. This observation prompted us to ask which physical quantity may be maintained near constant to achieve a homeostatic state in different collagen microenvironments. By employing a dimensional analysis, we found that the product between the total contraction energy W and the true collagen density ρ is maintained. In conclusion, we used a new tissue engineering platform to show that the homeostatic state in tissues may arise from a balance between fibroblast contraction and ECM compaction.
Xiange Jian
Poster
156
Using InSAR to Quantify Groundwater Storage Changes in Response to Atmospheric Rivers
Abstract →
Atmospheric rivers—long, narrow plumes of moisture from the Pacific—can dump huge amounts of rain on Los Angeles, California. Some of that water runs off, but some of it soaks into the ground and refills aquifers—the natural “underground reservoirs” that supply many communities. Within this regional picture, we focus on the San Gabriel Valley—one of Los Angeles’s major groundwater basins. When aquifers fill or are pumped, the land surface can rise or sink by tiny amounts—as if the ground were gently breathing. We use satellite radar—Interferometric Synthetic Aperture Radar (InSAR)—which can measure centimeter-level changes in ground height over large areas, to watch that breathing from 2015 to 2025. We combine the satellite measurements with local well records (groundwater levels) and rainfall from weather radar to answer a simple question: after big storm events, how much water actually ends up stored underground—and where? Our results show that specific atmospheric-river storms cause noticeable, short-term uplift that coincides with rises in groundwater levels, indicating recharge. The magnitude of recharge varies by year and with storm intensity. This work demonstrates that satellites can provide a clear, basin-wide view of when and where storms refill our groundwater, offering a practical tool to support sustainable water management.
FNU Syed Imran Uddin
Poster
158
Additive Manufacturing of a Naturally Closed Smart Gripper for Power-Limited Environments
Abstract →
Soft robotic grippers typically require continuous power to maintain grasping forces, making them unsuitable for environments where energy is scarce or power supplies are unreliable. We present an additively manufactured, naturally closed soft gripper with embedded sensing fabricated in a single cycle, eliminating the need for constant energy input and enhancing suitability for energy-constrained settings. Printing parameters—including temperature, retraction distance, print speed, and layer height—were systematically optimized, yielding high-quality prints with a 10% reduction in fabrication time. Sensor performance enhancement studies achieved a high sensitivity gauge factor of 7 and best-in-class repeatability (coefficient of variation <2%). The gripper is actuated by a high-energy-density Shape Memory Alloy (24 MJ/m³) and demonstrates both proprioception (self-sensing of configuration) and exteroception (contact detection), enabling awareness of its state and interactions. Integrated onto a humanoid robot, it successfully executed pick-and-place tasks with minimal energy demand.
Sushma Shreedhar
Poster
160
Characterizing Novel Binding Interfaces Between Histones H3-H4 and its Chaperone RBBP4
Abstract →
Access to eukaryotic DNA is tightly controlled by nucleosomes, a protein-DNA complex where DNA is wrapped around histone proteins like a lock around a key. To unlock this structure, cells use remodeling machines such as the Nucleosome Remodeling and Deacetylase (NuRD) complex. At the heart of NuRD is Retinoblastoma-Binding Protein 4/7 (RBBP4/7), a histone H3-H4 binder that shapes chromatin architecture and supports DNA repair. Previous studies identified two histone-binding interfaces on RBBP4/7, but we suspect there’s more to the story. To test our hypothesis of a third interaction site, we engineered and purified mutant H3-H4 histones and probed their binding. These insights matter: misregulation of RBBP4/7 is linked to neurodevelopmental disorders and multiple cancers. By mapping its histone contacts, we aim to illuminate how NuRD tunes chromatin access and how things go wrong when regulation falters.
Eshanta Mishra
Poster
162
Monitoring Cascadia’s Slow Earthquakes from Space
Abstract →
The Cascadia subduction zone, a major seismic plate boundary between the Juan de Fuca plate and the North America plate, is well known for posing one of the biggest seismic risks to western North America. The region experiences a range of tectonic activity, from huge earthquakes that occur every few hundred years to more frequent slow movements of tectonic plates called Slow Slip Events (SSEs). Unlike major earthquakes, SSEs happen gradually and are often not felt, but they provide crucial insights into the earthquake cycle. Satellites with Synthetic Aperture Radar (SAR) can measure millimeter-level ground movements from space, making them a powerful tool for studying ground deformation. We use a technique called Interferometric Synthetic Aperture Radar (InSAR), which compares multiple satellite images acquired at different times to track these movements. In Cascadia, however, InSAR data come with limitations such as a low signal to noise ratio (SNR) caused by atmospheric delays and decorrelation noise from the area’s dense vegetation. To address this, we analyzed 10 years of Sentinel-1 SAR images acquired over central Cascadia (43°N–45°N) between 2015 and 2025. Since atmospheric delays and decorrelation noise are considered temporally random, we combined or “stacked” multiple interferometric measurements that span slow slip events. This strengthens the common deformation signal related to the slow earthquakes while reducing other phase components such as atmospheric delays and decorrelation phase noise. Preliminary results show that this method reduces noise significantly and makes the ground deformation signals clearer. Our ultimate goal is to create accurate quantification of slip at depth from InSAR measurements in order to support the assessment of future seismic hazards in the region.
NEXT GEN STARTUP SHOWCASE SESSION 1
Adaeze Ekeleme
Next Gen
202
Nema Auto
Abstract →
Our business is developing a technology-driven, on-demand mobile auto repair service managed through an intuitive app. This solution eliminates the inconvenience of traditional repair shops by bringing transparent diagnostics, scheduling, and repairs directly to the customer. The platform leverages AI to streamline communication, provide upfront quotes, preventative maintenance, and offer a seamless, personalized service experience.
Meerrah Ganeshram
Next Gen
204
Orbital Mentorship Corporation
Abstract →
Orbital Mentorship is a free, inclusive platform that connects mentees with experienced mentors from diverse fields, including NASA and other top organizations. The platform focuses on personalized guidance, helping users develop skills, discover career paths, and apply strategies for growth and self-improvement. It emphasizes equal access, structured mentorship, and hands-on support for meaningful professional development.
Junchen Liu
Next Gen
206
Iron LEO
Abstract →
Iron LEO is developing an automated valet parking and vehicle service system. Our ground-lifting robots safely transport cars from drop-off points to parking spaces, while integrated service modules handle routine tasks such as tire rotation and oil filter replacement. Designed for both civilian and military use, this technology reduces labor costs, optimizes space utilization, and enhances operational efficiency in high-demand environments.
Rithin Pillai
Next Gen
209
MinuteSell
Abstract →
MinuteSell is an AI-powered marketplace designed to make selling faster and easier through intelligent AI search and automated listing support. Our platform features trusted, protected payments with QR-code escrow release, ensuring safety for both buyers and sellers. Unlike competitors, MinuteSell prioritizes speed, trust, and simplicity, creating a more seamless resale experience.
Danish Saeed
Next Gen
211
Cruz
Abstract →
It’s 2025, and somehow we’re still tapping screens while driving.​ We’ve got self-driving cars, voice assistants in our homes, and AI that can write essays — yet when it comes to navigation, we’re still scrolling through menus and fighting with touchscreens just to add a stop or check our ETA. It’s outdated, distracting, and unsafe.
Cruz is what navigation should have become by now. You don’t use it — you talk to it. You say, “Hey Cruz, avoid tolls and find me a coffee on the way,” and it just does it. No tapping, no settings, no fiddling while driving.
The future of driving isn’t about more screens or smarter dashboards — it’s about removing friction entirely. Cruz is that step forward: a natural, conversational copilot that turns your navigation into something that understands you.
More specifically, it’s a voice-based, conversational AI that is fully map-integrated. Cruz can personalize your route, check ETAs, check weather, remember your preferences, and communicate back and forth.
Serenity Underwood
Next Gen
217
Modari
Abstract →
Modari is developing a 3D-printed, modular, customizable shoe for athletes and their respective sports. Through our modular approach, athletes will only have to replace parts of the shoe, maintaining cost-effectiveness and sustainability.
NEXT GEN STARTUP SHOWCASE SESSION 2
Jasmine Curnel
Next Gen
201
Conceited Luxe
Abstract →
Conceited Luxe is a haircare company created for women with curly hair and acne-prone skin. We develop non-comedogenic, skin-safe formulas that protect curls without clogging pores, merging haircare with skincare. Our innovative packaging, including airless pumps, ensures a clean, luxurious, and mess-free experience compared to traditional jars.
Kordel France
Next Gen
203
Scentience
Abstract →
Scentience is building the sense of smell for robotics and embodied artificial intelligence. Scentience designs and manufactures neuromorphic computers, fusion software, and artificial intelligence for olfaction applications in robotics.
Monika Iytha
Next Gen
205
Drivile
Abstract →
Drivers overpay (avg. 20–30% premium variation for identical profiles). Insurers spend heavily on proprietary scoring systems. Our startup provide Driving profile as a service to help customers gain perfect premium for their profiles.
Ekene Obiesie
Next Gen
207
Phase Pal
Abstract →
Phase Pal builds AI-powered, anime-inspired characters that live on your desktop to help with work/studying, boost productivity, or simply keep you company. These characters are able to watch your screen to answer the users’ questions, perform interactive animations, and verbally talk to the user. In only our 1st month, we’ve generated over 8 million views across social media platforms and acquired over 3000 waitlist signups.
Vishwa Pandian
Next Gen
208
Auri
Abstract →
Auri is an emotionally-aware, memory-rich personal AI companion designed to go beyond utility chatbots. It integrates long-term memory, adaptive learning, and personality modeling to build deep, evolving relationships with users. Unlike traditional assistants that only provide information on demand, SocialAI continuously learns from conversations, remembers past interactions, and adapts its behavior and emotional tone over time—acting as a digital twin and supportive partner in daily life.
Avalene T
Next Gen
213
Emunized
Abstract →
Emunized is a mobile app designed to reduce parental vaccine hesitancy. It offers an intuitive, interactive roadmap of a child’s immunization schedule and lets parents track vaccine records over time. Centralized records help parents meet school or daycare requirements and support healthcare providers with more complete patient histories.
Arti Thakkar
Next Gen
214
Tattvaa
Abstract →
Tattvaa offers a clean, luxurious cosmetic line starting with lipsticks-crafted without harmful chemicals, safe for even kids. Rooted in transparency and purity, our products blend high performance with mindful beauty, targeting conscious consumers in the USA and Dubai who demand both safety and sophistication.
Ivan Tong
Next Gen
215
Medceptor
Abstract →
Medceptor is an AI “Field Training Officer” that trains emergency medical first responders through realistic, conversational simulations. It provides vital signs, physical exam findings, and treatment outcomes while also role-playing as the patient or 911 dispatch. In the past 6 months, we’ve built an MVP, onboarded 2 EMS Instructors as official advisors, and are now piloting it with UTD’s multi-award-winning collegiate EMS agency.
Ikeda Trashi
Next Gen
216
BioDelivera
Abstract →
BioDelivera is a platform company solving drug delivery with targeted, multi-payload engineered virus-like particles for oncology, gene therapy, and central nervous system disease, reducing off-target toxicity and improving outcomes.