University Of Pennsylvania

NSF Awards · FY 2026 · 2026-09

This Research Experiences for Undergraduates (REU) site award to University of Pennsylvania, located in Philadelphia, PA, supports the training of 10 students for 10 weeks during the summers of 2027–2029. In this program, participants pursue collaborative research projects, resulting in the development of new catalysts providing alternatives that address pressing chemistry needs and ultimately lower energy costs, diminish waste, and contribute cost and resource efficient systems to meet today’s industrial demands. In addition to participating in front-line research, participants will learn to present their research findings in a professional setting and acquire soft skills for career development as they focus on science communication, networking, and preparation for graduate school in a chemistry related field. By providing research experiences enriched by cohort building, workshops, a poster session, a written report and a focus on mentorship, this site trains undergraduates for careers in STEM and strengthens connections between the University of Pennsylvania and primarily undergraduate institutions (PUIs) from around the nation. The REU students undertake individual projects encompassing a span of technology readiness levels from machine learning for fundamental molecular modeling to development of novel catalysts for various applications and using catalysis for materials synthesis. These projects include screening catalysts, using quantum mechanical modeling (quantum information science, QIS), AI and machine learning (AI/ML) driven optimization, and high throughput experimentation for applications such as electrocatalysis and targeted bond activation. The professional development series includes training on communicating science, ethical research, laboratory safety, and presenting the professional persona through social media. To support the science communication aspect, students meet regularly as a group, guided by senior Penn graduate students, as they learn, refine, and advance their summer research projects and prepare their presentations. Participants also receive opportunities for career exploration and networking through industry talks and access to professional networks with mentorship from Penn faculty and professional staff. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2026 · 2026-07

This proposal seeks to fund US-based students to attend the 2026 Association for Computing Machinery Special Interest Group on Data Communications Conference (ACM SIGCOMM), held in Denver, Colorado on August 17-21, 2026. 2026 ACM SIGCOMM, is the flagship annual conference of the ACM Special Interest Group on Data Communication that attracts high-quality, forward-looking research contributions and provides a vibrant forum for technical and professional exchanges. 2026 ACM SIGCOMM, will expose selected students to cutting-edge developments in the field and enable interactions with world-leading researchers. Students will gain feedback on their ongoing work, broaden their academic perspectives, and build lasting professional connections. This project supports students from US universities to attend the 2026 ACM SIGCOMM conference in person. Students will have the opportunity to present their work and be exposed to state-of-the-art developments in the field. They will also have the opportunity to interact with peers from institutions worldwide, meet with senior researchers, and participate in discussions that are likely to shape the future of the field. This grant will support students who will benefit from attending this conference while Priority will be given to first-time attendees. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2026 · 2026-07

Peripheral nerve injuries (PNI) are a leading cause of chronic disability that affects millions of people in the U.S. Understanding long term effects of PNI is challenging because most diagnostic tools, such as advanced imaging, only provide a snapshot in time. The field of bioelectronics may provide new tools to monitor PNI continuously over long times, if certain challenges can be overcome. This CAREER project will design and fabricate bioelectronic devices that integrate into tissue and continuously monitor the progression of chronic PNI over long times. The outcomes of the project could lead to new personalized treatments of PNI. The research team will partner with the Franklin Institute in Philadelphia to launch "Engineering in Medicine" that will engage high school students in biosensor fabrication. Additionally, the project will establish a new bioelectronics curriculum. Overall, this project will develop novel biotechnologies to better understand nerve injury and will inspire a workforce capable of solving challenges in materials science and healthcare engineering. This CAREER project will support fundamental engineering research to overcome critical barriers of bioelectronics for PNI applications. First, this project will address challenges preventing long-term tissue integration of bioelectronic devices in dynamic nerve environments. An engineered interface will be fabricated to overcome the trade-off between wet adhesion and stability. This design is expected to enable robust attachment to peripheral nerve surfaces over long-term implantation. Second, this project will target the need for continuous biomarker monitoring without sacrificing sensitivity. A biomimetic interface will be developed that supports faithful molecular recognition and active regeneration. Finally, this project will validate the bioelectronic platform in a rat nerve crush model. The new platform will continuously capture multimodal data throughout injury progression and recovery. The sensor outputs will be correlated with nerve function. If successful, this dataset will provide insights into neuro-immune dynamics. The research outcome of this project will motivate data-driven strategies for future therapeutic development. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2026 · 2026-07

This project develops mathematical tools for assessing when artificial intelligence systems can be trusted after their predictions are used to make decisions. Modern AI tools help rank drug candidates, support medical decisions, screen large data sets, and suggest scientific hypotheses. In these settings, predictions are often used selectively and repeatedly: users may follow up only on top-ranked cases, choose confidence levels after seeing outputs, or let automated tools gather evidence over time. Standard uncertainty statements can become overly optimistic in such adaptive pipelines. This project will build a statistical quality-control layer for AI-assisted decisions and discoveries, helping users understand how reliable the resulting decisions are. The work can improve reproducibility and efficiency in biomedical science, drug discovery, and other data-intensive fields where experiments are costly and errors can slow progress. The project will also train graduate and undergraduate students in modern statistics, trustworthy AI, and responsible data science, with efforts to broaden participation. Publicly available software, benchmarks, and teaching materials will support education, reproducible research, and safer use of AI. The technical goal of this project is to develop finite-sample, model-agnostic, and distribution-free inference methods for AI systems used inside adaptive decision and discovery pipelines. The work draws on conformal prediction, predictive inference, selective inference, multiple testing, permutation methods, and anytime-valid testing. The first thrust will develop set-level predictive inference methods for multiple unlabeled instances, including false discovery rate control, family-wise error rate control, global null testing, partial-conjunction testing, and model selection under exchangeability and weighted exchangeability. The second thrust will study predictive inference under adaptive human use, including selective issuance of prediction sets and data-dependent choices of confidence levels, and will develop selection-conditional guarantees, auditing tools, and calibration methods. The third thrust will develop selective inference methods for automated evidence gathering, with a main focus on generated hypotheses, dataset reuse, and optional stopping through e-values and e-processes. The project will also explore extensions to dynamically expanding hypothesis structures and to settings in which automated evidence may be imperfect. Together, these results will expand the mathematical foundations of uncertainty quantification for black-box AI systems and provide open-source tools for reliable scientific discovery. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2026 · 2026-06

As artificial intelligence (AI) increasingly accelerates scientific discovery and engineering design, there is a growing need for models that are not only computationally fast but physically reliable. Many current AI approaches rely purely on massive datasets, predicting physical phenomena without incorporating the underlying laws of nature. This purely data-driven approach can lead to predictions that are unstable or physically impossible. This project develops 'physics-preserving' machine learning models that embed geometric and physical constraints directly into the AI's architecture. By ensuring these models obey fundamental physical laws by design, the research yields simulators that operate thousands of times faster than traditional computational methods without sacrificing accuracy. These advancements directly support Federal strategic interests in artificial intelligence and advanced manufacturing by enabling the creation of real-time, highly accurate digital twins for complex systems in aerospace, materials science, and energy. Additionally, the project supports workforce development by training a new generation of scientists, spanning high school, undergraduate, and graduate levels at the critical intersection of computational mathematics and machine learning.
 This project will create structure-preserving scientific machine learning (SciML) architectures to learn reduced partial differential equation (PDE) models incorporating constrained tensors. Examples include the stress and strain tensors in linear elasticity (symmetric), deviatoric stress (trace-free), and internal stress or linearized Riemann curvature (antisymmetric in pairs, pair exchange symmetry, algebraic Bianchi identities). The PDEs and tensor constraints will be formulated using higher-order differential complexes created by combining de Rham complexes via the Bernstein-Gelfand-Gelfand (BGG) technique. This ensures data-driven surrogates inherit the geometric structure inherent in the tensor objects. Rather than treating physics as data-driven regression, the approach performs learning at the level of the physics to build transformer models thousands of times faster than forward simulators, with geometric structure inherited from the BGG complexes providing guarantees of trustworthy performance. This work will lay the foundation for a unified theory of data-driven physics matching the rigor of finite element methods while preserving the approximation power of modern transformer methods. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NIH Research Projects · FY 2026 · 2026-06

Project Summary During pregnancy and lactation, the female skeleton undergoes significant bone loss and bone microstructure deterioration to provide calcium for fetal/infant growth. While weaning induces substantial bone recovery, reproduction-induced bone loss is only partially recovered after weaning. Our previous studies in rats uncovered several structural adaptation mechanisms through which the balance between the mechanical and metabolic functions of the maternal skeleton is achieved. Moreover, the adapted maternal bone structure from multiple cycles of reproduction and lactation exerts a protective effect against estrogen deficiency-induced bone loss later in life. Furthermore, we discovered enhanced mechano-responsiveness to external loading in the rat maternal bone both during lactation and later in life when subjected to estrogen deficiency by ovariectomy (OVX). These important findings lead to the overall objective of this project to define the mechanisms behind the long-term benefit of lactation, i.e., enhanced mechano-sensitivity in the maternal bone, throughout the lifespan. During lactation, osteocytes (Ocys) are known to actively remodel their surrounding matrix via perilacunar-canalicular remodeling (PLR). The activated PLR during lactation leads to an enlarged lacunar-canalicular system (LCS) and an altered microenvironment of Ocys. Using a multiscale poroelastic model and the LCS data of lactating animals, our results indicated that the PLR-induced alterations in the Ocy pericellular environment would amplify the mechanical and biochemical signal transduction to Ocys, which could in turn enhance the mechanical adaptation of maternal bone to maintain its load-bearing function during lactation. Intriguingly, we also discovered a re-activation of osteocyte PLR in the rat maternal bone (but not in virgine bones), in response to OVX-induced estrogen deficiency later in life, which could partially explain the enhanced bone adaptation to external loading. These exciting findings provide a strong scientific premise for our central hypothesis that Ocy PLR is an essential cellular mechanism through which mechano-sensitivity of maternal bone is enhanced during lactation and in response to estrogen deficiency later in life. In the Aim 1, we will establish the causal role of osteocyte PLR as an important mechanism to enhance bone mechano- sensitivity. In the Aim 2, by employing osteocyte fate mapping in a mouse model, we will for the first time interrogate the mechano-responses between Ocys with and without exposure to prior lactation or lactation- associated hormonal changes within the same bone. The proposed research project will define a novel function of Ocys in regulating mineral release and mechanical integrity, two competing demands on the maternal skeleton, and elucidate cellular mechanisms connecting the two life events (lactation and menopause) and their contributions to women’s bone health, which are well-aligned with the goals of the NOSI (NOT-OD-24-079): Women’s Health Research.

NIH Research Projects · FY 2026 · 2026-06

The dynamic immune landscape of mammalian pregnancy is essential for healthy reproduction. Yet, maternal immunity after birth remains largely undefined. There is a major gap in knowledge in of baseline postpartum immune dynamics and their impacts on disease outcomes during this phase of reproduction. We are poised to address this gap in knowledge given we have recently discovered that that postpartum mothers are more resistant to pathogenic disease outcomes of a lethal viral infection. Based on literature and our extensive preliminary data, we hypothesize that immune changes are enacted after birth to promote enhanced resilience to viral infection in the postpartum period that is dependent upon lactation-associated physiology. This proposal describes targeted objectives to define immune mechanisms that underlie enhanced maternal resilience to severe viral disease during the postpartum period. In Aim 1 we will focus our investigations on uncovering the breadth, kinetics, and immune determinants of protection in postpartum during maternal resilience. In Aim 2 we will focus our studies on the role of maternal caregiving in triggering maternal resilience, decoupling maternal caregiving behavior from lactation-associated physiology. Altogether, outcomes from this proposal will embody new fundamental insights into immune dynamics during postpartum period. Specifically, this proposal will define mechanisms underlying enhanced maternal response to viral challenge after birth. Our findings will advance our knowledge of maternal physiology after birth and underscore mechanisms of biological resiliency during reproduction and provide foundational knowledge that could ultimately be leveraged to improve women’s healthcare.

NIH Research Projects · FY 2026 · 2026-06

Immune checkpoint inhibitor (ICI) has become an important treatment modality for triple- negative breast cancer (TNBC). However, TNBC tumors have relatively low mutation burden and neoantigen counts but are enriched with myeloid derived suppressor cells (MDSC) and M2-like tumor associated macrophages, both shaping an immune suppressive tumor microenvironment. Leveraging on its ability to release neoantigen, chemotherapy is combined with ICI in the clinic to enhance the efficacy of ICI therapy, however the response rate to ICI+chemotherapy is modest. Therefore, developing an effective strategy to enhance anti-tumor efficacy of ICI will address this unmet clinical need. Analysis of The Cancer Genome Atlas (TCGA) suggest that TNBC's glutamine metabolic signature is associated with low anti-tumor immune signature and unfavorable prognosis of TNBC patients. Our preliminary studies have tested two FDA-approved investigational new drugs (IND) and shown that targeting glutamine metabolism overcame the resistance to ICI in TNBC models and glutamine antagonism enhanced the efficacy of ICI+chemotherapy to inhibit tumor growth. In this application, we will validate the strategy of targeting glutamine metabolic signature in the setting of ICI therapy and assess the ability of [18F](2S,4R)4-Fluoroglutamine (4F-Gln) and [18F]FDG PET to detect glutamine antagonism and immune activation, respectively and to predict survival (Aim- 1). We will elucidate cancer versus immune cell contribution to 4F-Gln and FDG PET signal, particularly, we aim to determine the extent to which overall FDG PET uptake at baseline and post-treatment is driven by immune cells versus cancer cells in the tumor, and whether increased FDG uptake post-therapy is associated with immune activation by comparing FDG uptake in immune cells with the expression of immune activation markers including CD69 (Aim-2). To facilitate the clinical translation of precision imaging markers, we will validate [18F]Fluciclovine for detecting glutamine antagonism and implement a dual tracer (Fluciclovine/FDG) protocol in single PET session (Aim-3). Given that TNBC is highly heterogeneous in glutamine metabolic machinery as well as responses to ICI, the highly translational treatment approach and precision imaging markers validated in this application can lead to a window of opportunity trial upon successful completion of this project. .

NIH Research Projects · FY 2026 · 2026-06

PROJECT SUMMARY Ischemic stroke is a leading cause of disability and death in the United States. Despite improved preventive therapies, the incidence of ischemic stroke is still unacceptably high. Progress in preventing ischemic stroke events will require a better understanding of their novel risk factors, biological pathways, and causal mediators. Protein levels can serve as potent, specific, and modifiable biomarkers for ischemic stroke risk, guide preventive therapy and elucidate causal mediators and pathways. Large-scale proteomic technology has become available to scan 6,432 distinct plasma proteins (7,596 aptamers) simultaneously using modified aptamers as binding reagents. We will conduct the proteomic investigation in a case-cohort population of 2,000 participants without history of prior strokes from the REGARDS cohort. The overarching goal of this proposal is to improve incident ischemic stroke risk prediction and thus identify high-risk individuals. We will also combine proteomics with genomics to identify proteins that are causal mediators of ischemic stroke and potential new therapeutic targets for stroke prevention. Using artificial intelligence (AI)-based algorithms, we will evaluate the 3-dimensional structure of proteins looking for binding pockets for small molecule ligands (drugs) of potential causal proteins. These findings will be followed by virtual docking experiments that screen libraries of small molecules for their binding to proteins of interest. We will validate key findings from REGARDS in the Multiethnic Study of Atherosclerosis (MESA study). If we are successful, we will: 1) Create protein-based risk scores for incident ischemic stroke with better discrimination and calibration than current models. 2) Identify protein biomarkers, biological pathways, and nominate new therapeutic targets for ischemic stroke. 3) Identify druggable proteins that contain binding sites for small ligands. Docking data can then be used for in vitro screening of small molecule libraries and potential drug therapies. 4) Generalize and reinforce key discoveries in REGARDS by their validation in the MESA study.

NIH Research Projects · FY 2026 · 2026-06

PROJECT SUMMARY Pancreatic ductal adenocarcinoma (PDAC) has the worst survival of any major tumor type. At diagnosis, most tumors have invaded or metastasized, limiting treatment to systemic therapies with a modest impact on survival. Consequently, there is a pressing need to eradicate tumors before they reach more advanced stages. While early detection for PDAC remains elusive, cancer interception – a strategy that eliminates premalignant lesions before progression to invasive cancer – is an attractive complementary approach. Distinct from either prevention or treatment, cancer interception involves intervening when neoplasms are at an incipient stage. Colonoscopy – an example of “mechanical interception” – represents a successful application of this strategy, in which periodic removal of premalignant colon adenomas has markedly reduced colorectal cancer incidence and mortality. Similar approaches remove early lesions in the cervix, breast, and skin, but such procedures are only practical in easy-to-access tissues with readily identifiable and anatomically defined precursors, making them unsuitable for other solid tumors such as lung, liver, or pancreas cancer. The goal of this proposal is to advance translatable strategies with the potential to interrupt PDAC natural history in high-risk patients. Most PDACs arise from well-recognized precursors known as pancreatic intraepithelial neoplasias (PanINs) and intraductal papillary mucinous neoplasms (IPMNs). As these lesions progress to frank malignancy, they fashion a protective tumor microenvironment (TME) consisting of a dense “desmoplastic” stroma, paucity of blood vessels, and immunosuppressive myeloid cells. These changes are recapitulated in the KPC model of PDAC, which incorporates oncogenic mutations in KRAS and TP53. KPC mice develop PanIN lesions with predictable kinetics (6-9 weeks of age), making this model an ideal platform for exploring PDAC interception strategies. Our preliminary studies with KRAS inhibitors provide strong support for the feasibility of interception approaches. Here, we hypothesize that critical interactions between epithelial and stromal cells constitute a “pre-TME network” that is essential for PanIN lesions to progress to cancer. We further hypothesize that pharmacological disruption of the network will cause its collapse, resulting in effective PDAC interception. Because interception can only be studied in the setting of pre-malignant progression, our existing KPC infrastructure (“Penn Mouse Hospital”), together with experienced bench and clinical research teams, uniquely position us to test these hypotheses in animal models and human patients through three Specific Aims: Aim 1. Determine interception strategies for PanIN lesions and the impact of interception on the pre-TME Aim 2. Enhance the durability of interception Aim 3. Conduct a pre-surgical clinical trial to evaluate PDAC interception in high-risk patients

NIH Research Projects · FY 2026 · 2026-06

3. Abstract describing research plan for R33 The R33 phase will investigate the mechanistic effects of exogenous ketone supplementation (KS; 25g) on brain nicotinamide adenine dinucleotide (NAD+) concentrations and neurocognitive function during acute alcohol exposure in healthy adult drinkers. Building on the R21 findings demonstrating that 25g KS significantly reduced breath and blood alcohol concentrations and that 7T downfield proton magnetic resonance spectroscopy (¹H- MRS) scans reliably quantify the cerebral NAD+ peak at 9.3 ppm, the R33 will test whether ketone-induced modulation of NAD+-dependent metabolism alters central neurobiological and behavioral responses to alcohol. Forty healthy men and women (ages 21-50), who report at least one recent heavy drinking episode but do not meet criteria for alcohol use disorder, will participate in a randomized, counterbalanced, two-period crossover study. Participants will complete two experimental sessions separated by a washout period. In one session, participants will receive 25g KS; in the other, a taste-matched placebo. Thirty minutes following beverage ingestion, participants will undergo multimodal 7 Tesla MRI scanning, including structural imaging, baseline NAD+ quantification using downfield ¹H-MRS, and functional MRI during a Stop Signal Task (SST) to assess inhibitory control. Alcohol will then be administered outside the scanner to achieve a target breath alcohol concentration (BrAC) of 0.08%. Participants will return immediately to the scanner for repeated NAD+ ¹H-MRS acquisition, SST functional MRI, and exploratory spectroscopy to quantify brain alcohol and additional metabolites (e.g., β-hydroxybutyrate, glutamate). Venous blood samples will be collected to quantify alcohol, NAD+, and ketone concentrations. The primary hypothesis is that KS will increase brain NAD+ concentrations relative to placebo and attenuate alcohol-induced reductions in NAD+ levels. The primary outcome will be the interaction between treatment (KS vs placebo) and alcohol (pre- vs post-alcohol) on NAD+ concentration at the 9.3 ppm peak. Secondary hypotheses are that KS will attenuate alcohol-induced impairments in inhibitory control (indexed by stop-signal reaction time) and modulate blood oxygenation level-dependent (BOLD) activation within fronto-temporal inhibitory control networks. Exploratory analyses will examine associations between blood and brain NAD+ levels, alcohol pharmacokinetics, and cognitive performance, and will assess potential moderation by sex and genetic variation in alcohol-metabolizing enzymes. This study is innovative in combining peripheral pharmacokinetic assessment, high-field NAD+ spectroscopy, and task-based functional neuroimaging within a controlled alcohol challenge paradigm. By directly examining NAD+-dependent mechanisms in vivo, the R33 phase will provide mechanistic insight into how ketone supplementation may alter alcohol metabolism and sensitivity at both peripheral and central levels. Findings will advance understanding of cellular and molecular mechanisms of alcohol tolerance and may identify NAD+- dependent metabolic modulation as a novel target for prevention or intervention strategies for alcohol misuse and alcohol use disorder. 4. Aims page revision The Specific Aims have been modified in a minor way from the original peer-reviewed and approved application. In the originally proposed R33 phase, participants were to receive either 10g or 25g of ketone supplement (KS), with final dose selection contingent upon R21 findings. Based on the R21 results, which demonstrated that 25g KS significantly reduced breath and blood alcohol concentrations (p<0.05) whereas 10g did not produce significant pharmacokinetic effects, the R33 phase will utilize the 25 g dose exclusively. This modification reflects data-driven optimization consistent with the original study design and milestone framework. The overall scientific premise, hypotheses, and mechanistic focus on NAD+-dependent modulation of alcohol sensitivity remain unchanged. No other aims, hypotheses, or outcome measures have been altered.

NIH Research Projects · FY 2026 · 2026-06

PROJECT SUMMARY For individuals with chronic illnesses such as diabetes, heart failure, cancer, or obesity, early intervention prevents symptom escalation, acute care use, and mortality. The growing availability of longitudinal electronic health record (EHR), patient-reported outcome (PRO), and mobile health (mHealth) data, including passively collected accelerometer, smartphone, and sensor data, offers new opportunities for proactive intervention. However, the rapid expansion of routinely collected mHealth data has outpaced the research community's ability to interpret it effectively. In particular, the high dimensionality and irregular collection of longitudinal EHR, PRO, and mHealth data introduces key challenges for predictive modelling due to planned sparsity or unplanned missingness. Current methods fall short in three key areas: 1. Informative missingness: Data gaps often carry predictive signal, but are typically treated as nuisance, obscuring meaningful patterns in their timing and duration. 2. Loss of intra-day detail: Fine-grained mHealth data are often reduced to pre-specified daily or weekly summaries, discarding rich intra-day information with potential predictive value. 3. Population heterogeneity: Models trained on populations often perform poorly for underrepresented groups and fail to generalize to individuals, especially when only limited data are available per person. To address these gaps, we propose developing a robust methodological framework for predictive modelling using irregularly collected EHR, PRO, and mHealth data that improves upon imputation-based standard of care methods. In Aim 1, we will develop univariate and multivariate longitudinal models that account for delays between predictors and outcomes and incorporate detailed intra-day mHealth patterns using distributional learning with low-dimensional, near-lossless embeddings. In Aim 2, we address population heterogeneity by personalizing prediction through an embedding-based approach using landmark multidimensional scaling (MDS) and transfer learning, with reweighting of MDS landmarks to improve performance for underrepresented subgroups. In Aim 3, we validate these methods across diverse mHealth and EHR datasets, including NIH All of Us, UK Biobank, and other disease-agnostic and -specific retrospective and prospective datasets, using mixed-methods studies among clinicians to operationalize model outputs for clinical decision support. Though broadly applicable to multi-modal longitudinal data of all types and a range of disease settings, we focus on four chronic conditions with high clinical impact: cancer, congestive heart failure, diabetes, and obesity. The success of this project and its open-source tools will help close a critical methodological gap and enable effective use of multi-modal longitudinal data to improve clinical decision-making for chronic disease management.

NSF Awards · FY 2026 · 2026-06

Artificial intelligence (AI) systems increasingly assist with decisions in areas such as, medicine, science, and education, but they still make simple reasoning mistakes. Thus, they can be unreliable in high-stakes settings. This project develops a science of AI reliability by identifying why these systems fail and by designing principled methods to make them more dependable, both on their own and when working with humans. The project's novelties are a unified framework that connects model-level robustness to human-AI collaboration, grounded in controlled mathematical environments that isolate real-world failure modes while remaining tractable for rigorous analysis. The approach bridges theory with targeted experimentation to produce insights that transfer to full-scale systems. The project's broader significance and importance are providing scientific foundations for safer deployment of AI in critical applications such as medical diagnosis and producing open-source evaluation tools for the research community. In addition to its technical objectives, this project extends its impact through expanding the Learning Theory Alliance mentorship program, promoting research with minimal computational resources, and developing new graduate and undergraduate courses. At the model level, the research characterizes why transformers, the backbone of modern AI systems, learn brittle shortcuts rather than robust algorithms and develops principled interventions in training data, training algorithms, and inference to make reasoning reliable by design. These include representative training sets that teach robust algorithmic behavior, sample-efficient methods for multi-step reasoning, and formal safeguards against adversarial manipulation. At the interaction level, the research develops a theory of human-AI collaboration under complementary information and imperfect alignment. The resulting protocols are auditable, grounded in verifiable conditions such as calibration, and enable humans and AI to combine information and achieve near-optimal outcomes even when the system is not perfectly aligned. Reliable collaboration requires reliable internal reasoning, and failures in collaboration reveal new requirements for model-level reliability. Across both directions, controlled test environments produce precise predictions validated on real benchmarks. This work strengthens the scientific basis for deploying AI safely where the cost of failure is high. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NIH Research Projects · FY 2026 · 2026-06

PROJECT SUMMARY / ABSTRACT Chronic rhinosinusitis (CRS) is a common and debilitating upper airway disease characterized by persistent sinonasal inflammation and infection. Affecting >10% of the U.S. population and accounting for ~20% of all adult antibiotic prescriptions, CRS imposes a substantial health and economic burden. Despite aggressive medical and surgical interventions, many patients fail to achieve lasting relief, highlighting the need for novel, mechanism- based therapies. Recent discoveries have identified bitter taste receptors (TAS2Rs), a family of G-protein- coupled receptors, as important modulators of airway host defense and inflammation. In the nasal epithelium, TAS2Rs are expressed in two critical cell types—ciliated cells and tuft cells. It has been proposed that ciliated cells utilize TAS2Rs, such as TAS2R38 and TAS2R14, to detect microbial metabolites and trigger nitric oxide (NO)-mediated antimicrobial responses. Tuft cells are a major regulator for interleukin-25 (IL-25), acetylcholine (ACh), and defensins, and express TAS2R receptors like TAS2R10 that may regulate type 2 inflammation relevant to CRS with nasal polyps (CRSwNP). Progress in the field has been hampered by key limitations: (1) reliance on in vitro systems, (2) lack of functional orthology between human and mouse TAS2Rs, and (3) functional redundancy within the large Tas2r gene family in mice. To overcome these barriers, we have developed novel Tas2r-deficient mice. We will build on this unique platform by generating humanized mouse models expressing selected human TAS2Rs in specific cell types for CRS-related studies. Aim 1 will determine how cilia-localized TAS2Rs, particularly TAS2R38 and TAS2R14, regulate NO production that controls ciliary function and antimicrobial defense. Using air-liquid interface (ALI) cultures and ex vivo explants from wild-type, Tas2r mutants, and humanized mice, we will test responses to specific TAS2R agonists (e.g., PTC and quinine) in mucociliary transport and bacterial killing assays. These mice will also be tested in vivo using an eosinophilic CRS-like inflammation challenge (ovalbumin and fungal protease), live Pseudomonas aeruginosa infection, and stimulation with the bacterial quorum-sensing molecule (and TAS2R agonist) 3-oxo-C12HSL. We will assess inflammatory responses, epithelial damage, and cytokine profiles in nasal tissues and lavage fluid. Aim 2 will investigate how tuft cell-localized TAS2Rs, particularly TAS2R10, influence IL-25 and ACh secretion to contribute to airway inflammation. We will assess how tuft cells affect calcium signaling, cytokine secretion, and defensin production using ALI cultures and nasal explants. Further analysis of tuft cells will parallel the in vivo challenges described above in Aim 1 to determine how tuft cell TAS2Rs contribute to immune activation and epithelial remodeling. Both aims will also incorporate single-cell RNA-seq and immunophenotyping to reveal the cellular and molecular changes associated with TAS2R-dependent responses. This project will dissect the cell-type- specific roles of TAS2Rs in airway defense and inflammation, provide conclusive mechanistic insight into their functions, and establish the basis for developing TAS2R-targeted therapies for CRS and related airway diseases.

NIH Research Projects · FY 2026 · 2026-06

Proposal Summary/Abstract T cell-mediated inflammatory and autoimmune skin diseases affect millions worldwide, with healthcare costs exceeding $100 billion annually in the United States alone. Current treatments targeting cytokine pathways often fail to achieve lasting remission due to the persistence of tissue-resident memory T cells (TRM) in the skin. The molecular mechanisms enabling human TRM to persist in the challenging skin environment remain unknown, preventing development of more effective therapeutic strategies. Through an unbiased genome-wide CRISPR/Cas9 knock-out screen in a human skin xenograft model, we discovered that protein translation efficiency is essential for TRM survival. This proposal investigates this newly discovered molecular pathway controlling T cell tissue residence through regulation of protein translation efficiency. In Aim 1, we will determine when and where protein translation efficiency in T cells becomes critical to establish tissue residency using single-cell spatial transcriptomics, epigenetic profiling, and intravital imaging. By tracking T cells in real-time within human skin tissue, we will reveal the precise checkpoints where efficient protein translation enables tissue adaptation. In Aim 2, we will define how protein translation efficiency controls the TRM- specific protein synthesis program through stable isotope labeling and polysome profiling. This will uncover how TRM adapt their protein synthesis machinery to survive in the nutrient-limited skin environment. In Aim 3, we will evaluate therapeutic strategies targeting translation termination to selectively eliminate TRM, comparing systemic small molecule degraders with targeted lipid nanoparticle delivery of gene editing components. These complementary approaches will determine the most effective method for therapeutic targeting of TRM while sparing beneficial circulating T cells. Our innovative approach combines: 1) the first genome-wide functional screen in human TRM, 2) cutting-edge spatial and molecular technologies to study translation control in human tissue, and 3) immediate therapeutic applications through clinical-stage compounds. Conceptually, this work departs from the status quo and has the potential to transform our understanding of human TRM biology by revealing how protein translation efficiency enables tissue adaptation while providing mechanistic justification for therapeutic targeting of translation termination in T cell-mediated skin diseases. Success could lead to paradigm-shifting treatments that selectively eliminate TRM, offering lasting remission for millions affected by chronic inflammatory skin conditions that are mediated by T cells.

NIH Research Projects · FY 2026 · 2026-06

ABSTRACT / PROJECT SUMMARY Here we continue our development of the first safe & effective DNA-loaded lipid nanoparticles (DNA-LNPs) for the treatment of common chronic diseases such as atherosclerosis, emphysema, etc. DNA-LNPs are built on the billion-patient success of the mRNA-LNP vaccines for COVID, which showed LNPs’ unprecedented ability to drive protein expression. However, mRNA as a cargo has 2 major limitations that prevent it from being used for common chronic diseases: i) mRNA has a very short half-life of hours; ii) mRNA does not naturally encode cell- type-specific expression, which causes off-target expression. We solved both of these deficiencies in mRNA by loading LNPs with DNA, in a paper in revision at Nature Biotechnology. DNA has long been the goal of LNP delivery, but for 20 years DNA-LNPs proved “too toxic”, as we and others showed that IV injection of standard DNA-LNPs kill 100% of healthy mice within 2 days. We discovered this toxicity is caused by massive inflammation induced when LNPs’ delivery of DNA activates the cytosolic DNA sensor system cGAS-STING. We solved this by loading DNA-LNPs with our bodies’ natural inhibitor of STING, the nitrated lipid NOA, which is produced after viral infection to resolve STING-induced inflammation. NOA-DNA-LNPs completely abrogate STING activation and brought DNA-LNPs’ mortality from 100% to 0%. Further, NOA-DNA-LNPs produce high levels of protein expression for ~6 months in vivo and even accommodate large plasmids (>10kb). DNA-LNPs are poised to treat common chronic diseases in ways no other vector can: express any protein (not limited in size like AAV’s <4.4kb limit) or knockdown any protein (expressing shRNA); achieve cell-type-specificity via cell-type-specific promoter sequences (which do not fit in AAV) and LNPs’ other methods of cell-targeting; express for 6 months per dose; and non-immunogenic compared to AAVs. While our DNA-LNPs have great potential, we need to increase their expression by 2 orders of magnitude to reach levels achieved by mRNA-LNPs and AAV. Increasing expression will open up many additional diseases for treatment, and lower the required dose and thus side effects. Our central hypothesis is 3 hurdles inhibit DNA-LNPs’ expression: i) residual activation of cytosolic DNA-sensors (Aim 1); ii) degradation of DNA by intracellular DNases (Aim 2); & iii) inefficient nuclear transport of DNA (Aim 3). In each Aim, we will engineer “chemical approaches” (loading molecules into DNA-LNPs) & “genetic approaches” (deliver new sequences). The Deliverable by the end of this R01 is a new generation of DNA-LNPs that drives protein expression equivalent to mRNA-LNPs and AAV. This would enable DNA-LNPs to become the next pillar of genetic medicine, alongside mRNA-LNPs, AAV, siRNA, and CRISPR.

NSF Awards · FY 2026 · 2026-06

Some of the most powerful ideas in STEM are expressed in symbols, equations, and proofs. Unfortunately, these formal representations (i.e., formalisms) can be hard to understand, putting distance between millions of learners, practitioners, and scientists and the concepts that could support their work. The fact that scientific texts are becoming increasingly digital provides new opportunities to help readers understand formalisms. Digital texts might let readers look at individual terms in equations, access definitions that are aligned with their own knowledge, and see examples that make their meanings more concrete. However, creating these expanded ways of engaging with formalisms can be difficult for authors, and it is not clear which kinds of support would be most helpful to readers. This project looks to address these questions by developing a design space of ways to better-explain formalisms, tools for authors to create and readers to use these explanations, and studies of how these explanations help authors and readers talk and learn about science. By doing the work with both scientists and students, the project will lower barriers to those struggling with equations and proofs, leading to better STEM education and a stronger STEM workforce. This project's contribution is the design of, evaluation of, and tools for creating "explorable formalisms". These are formal representations, including symbols, equations, and proof, that are made more intelligible through interaction. The research will first contribute interaction techniques for making formalisms' meaning clear through interaction, for instance with on-demand diagrams, walkthroughs, simulations, and logical traces. Second, the research team will provide empirical evidence of the usefulness of the techniques grounded in studies of learning in the lab and classroom. Third, the researchers will develop new tools to facilitate the creation of explorable formalisms. These include a new markup language for rapidly and cleanly associating textual and LaTeX formalisms with granular descriptions, computations, and presentation hints. They also include algorithms supporting equation diagram layout and logical trace discovery. To bring explorable formalisms to scientific texts broadly, this project disseminates the tools, pilots their integration into AI-based systems for answering scientific questions, and incorporates them into learning materials for STEM classrooms. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NIH Research Projects · FY 2026 · 2026-06

Project Summary Comparative effectiveness research is a critical component of medical research that relies on randomized experiments and observational study evidence to identify interventions that improve healthcare outcomes. In many cases, the analysis of the data obtained from such studies is complicated by truncation by death, time-varying confounders, and zero-inflated outcomes. A common approach to address these difficulties is to use models that rely on unrealistic, untestable assumptions. In many cases, the quality of the evidence is low, since it is hard to assess which assumptions are necessary to make inferences. To that end, we develop ablation methods for causal inference. The result is an ablation framework that allows analysts to understand which assumption or combination of assumptions lead to informative conclusions. For instance, analysts will be able to identify the minimum set of assumptions that are needed for either point identifi- cation or the sign of the effect. We develop three different causal ablation frameworks for three specific applications that are common in comparative effectiveness research: truncation by death, time-varying confounders, and zero-inflated outcomes. Under each aim, for each application, we develop (1) appropriate assumption sets that meet a compatibility criterion, (2) a grid search tai- lored to the assumption sets, (3) a corresponding set of point and partial identification estimators, and (4) a summary step that reports results by order of assumption credibility.

NIH Research Projects · FY 2026 · 2026-06

PROJECT SUMMARY This career development award will be supporting Dr. Samuel Shin’s development as an independent scientist focused on iron associated mitochondrial injury in the setting of TBI. Dr. Shin has been a an extremely productive graduate student and medical trainee conducting research in the field of TBI, but in the last 3 years as a junior faculty he has been focused on developing a new pathway that will allow him to uniquely contribute to the scientific community. This current plan will help him build his specialty as a mitochondrial and iron associated oxidative stress expert in TBI. He will gain technical expertise in flow cytometry, transgenic techniques using adeno-associated virus to study development of secondary brain injury, and immunohistochemistry to understand iron metabolism changes after TBI. This will be further supplemented by formal didactics in mitochondrial biology, redox physiology, biostatistics, ethics, and grant writing. This career development effort will be overseen by a mentoring team by faculty from University of Pennsylvania and Children’s Hospital of Philadelphia composed of Dr. Todd Kilbaugh (mentor), a renowned critical care physician who is an expert in brain injury and mitochondrial physiology and Dr. Frances Jensen (co-mentor), an expert in neural plasticity and neurodegeneration. Additionally, he will have four advisors: Dr. Ramon Diaz-Arrastia, a thought leader in TBI therapeutics and biomarkers, Dr. Douglas Wallace, a renowned leader in the field of mitochondrial physiology and genetics, Dr. Rajiv Ratan, an expert in the field of ferroptosis and iron metabolism at Weill Cornell School of Medicine, and Dr. Sharon Xie who has extensive experience in training fellows and junior faculty in biostatistics. The PI (Shin), will also be supported by exceptional resources and faculty at the University of Pennsylvania and Children’s Hospital of Philadelphia for TBI research. The objective of this project is to demonstrate that hemoglobin byproducts and iron leads to mitochondrial dysfunction and subsequent secondary injury. The central hypothesis is that TBI associated hemorrhage leads to accumulation of hemin, which induces accumulation of mitochondrial iron leading to exacerbation of TBI- induced mitochondrial oxidative stress, cell death, and behavioral impairment. To investigate this novel pathophysiological mechanism, the PI proposes to use multiple genetic manipulation techniques to mitigate mitochondrial iron overload and oxidative stress. Specifically, mitochondrial ferritin will be expressed using AAV in mice prior to TBI. Also, mitochondrially directed antioxidants will be expressed, such as mitochondrial glutathione peroxidase-4, periredoxin-3, and mitochondrial catalase to assess for the role of each antioxidant function in TBI.

NIH Research Projects · FY 2026 · 2026-06

PROJECT SUMMARY This project aims to develop theoretical models, validated through experiments, to explain the progression of Metabolic dysfunction-Associated Steatotic Liver Disease (MASLD). MASLD is a rapidly growing global epidemic, affecting over a third of the population and becoming a leading cause of liver fibrosis, end-stage liver disease, and hepatocellular carcinoma. It ranges from simple steatosis (fat-laden liver cells) to advanced liver damage. Though only 20-30% of patients with steatosis progress to severe disease, the high prevalence and risks associated with MASLD make it crucial to identify those likely to progress—something that current methods cannot predict. A key factor in disease progression is elevated cholesterol. Our preliminary data from a rat model show that a high-fat/high-cholesterol diet leads to cholesterol crystal formation in the liver, which increases stiffness and leads to fibrosis. These crystals have been observed in human livers as well. Unlike simple lipid droplets, stiff cholesterol crystals likely impact tissue mechanics, cell organization, and nuclear structure, driving disease progression, though this has not been studied. Building on our previous work in tissue and nuclear mechanics, we propose to integrate and expand models to study the effects of cholesterol crystals on the liver. Through ongoing experiments with primary human hepatocytes and rat models, we aim to achieve the following objectives: 1. Predict and validate the impact of cholesterol crystals on liver tissue mechanics, focusing on stiffness, collagen network plasticity, and the effects of crystal size and shape. 2. Examine the impact of cholesterol crystals on hepatocyte cytoskeletal organization, using models to predict how these inclusions alter cytoskeleton-nucleus interactions in primary human hepatocytes. 3. Determine how cholesterol crystals affect chromatin structure and gene expression, using advanced imaging and sequencing techniques to predict their influence on genome architecture and key gene regulation in primary human hepatocytes. This research could significantly advance understanding, diagnosis and treatment of MASLD.

NIH Research Projects · FY 2026 · 2026-06

Tendons carry tensile loads via their dense extracellular matrix (ECM), which transmits mechanical strain to resident cells. Applying physiological loading promotes an anabolic response, which is the basis of physical therapy to treat tendon disease. Conversely, unloading generates profound catabolic effects, including rapid reduction of chromatin accessibility in mechanosensitive loci and catabolic enzyme production, impairing mechanical properties. Despite the importance of applied loading in maintaining tensional homeostasis, how mechanical cues are transduced to the nucleus within the native ECM is a significant gap in knowledge. Cell monolayer experiments have shown that tendon cells can interpret applied mechanical stimuli via integrin- mediated cell-ECM attachments called focal adhesions, which activate signaling cascades to transmit these inputs through the cytoskeleton to elicit a cell response. Perturbing cytoskeletal tension has profound effects on tissue homeostasis, and one key mechanotransduction protein that regulates cytoskeletal tension is the intracellular protein kinase focal adhesion kinase (FAK). In our ex vivo tendon loading experiments, we showed that applied loads result in elongation of the nucleus, and inhibition of FAK phosphorylation (FAK-I) attenuated this change, indicating its importance in this signaling cascade. Further, FAK-I suppresses the gene expression and cell contraction response to de-tensioning ex vivo. We also generated mice with tenogenic (ScxCre) knockdown of FAK and discovered markedly smaller tendons with altered collagen fibrillogenesis and mechanics. These studies highlight the importance of cytoskeletal tension in tendon function and the role of FAK activity in regulating the response to mechanical cues. We will perturb this mechanotransductive signaling cascade in tendon cells in situ using innovative and targeted means, altering the expression (genetic knockdown) or activity (pharmacological inhibition) of FAK. While other pathways contribute, FAK mechanotransduction, supporting is a key bottleneck in cell positioned downstream of integrins and upstream of the cytoskeleton, underscoring and the need for focused study. We expect that reduced FAK expression and activity will impair the ability of tendon cells to sense applied mechanical stimuli, which will attenuate the downstream cell response. Our overall hypothesis is that FAK-dependent tendon cell mechanotransduction regulates tendon tissue formation and the response to altered mechanical loading. Our two Aims are 1: Determine how altered loading and FAK- mediated mechanotransduction in neonatal tendons regulate growth and 2: Determine the extent to which altered loading and FAK-dependent mechanotransduction regulate adult tensional homeostasis in situ. This innovative and rigorous proposal will define the FAK-mediated mechanisms by which tendon cells respond to mechanical cues, providing insights into cell mechanosignaling due to loading to inform both the pathogenesis and treatment of overuse tendinopathy.

NIH Research Projects · FY 2026 · 2026-06

Project Summary Recurrent breast cancer is typically incurable and arises from the reservoir of residual tumor cells (RTCs) that can persist for decades in patients in a dormant state after primary tumor treatment. Although tumor dormancy and recurrence are responsible for the vast majority of breast cancer deaths, their mechanisms are poorly understood. In this regard, JAK-STAT signaling is activated in many human cancers and exerts a variety of pro-tumorigenic effects. Strikingly, using validated genetically engineered mouse (GEM) models that faithfully recapitulate breast cancer dormancy and recurrence in patients, we have made the surprising discovery that JAK1 deletion or pharmacological inhibition accelerates breast cancer recurrence by inducing RTCs to exit the dormant state and resume proliferation. Based on these findings, we hypothesize that JAK1/2 signaling is required to suppresses breast cancer recurrence and that it does so by enforcing a dormant state in RTCs. If true, this would raise the concerning possibility that administration of JAK1/2 inhibitors to patients harboring dormant RTCs could inadvertently trigger reentry of tumor cells into the cell cycle, which could ultimately result in lethal recurrence. Indeed, we have recently found that breast cancer patients exposed to a JAK inhibitor at some point after their initial diagnosis have a significantly increased risk of recurrence. Since there are >4 million breast cancer survivors in the U.S., and millions of patients treated with JAK inhibitors for a wide array of chronic inflammatory conditions, the possibility that JAK inhibition might increase recurrence risk when administered to early-stage cancer patients with dormant MRD is critical to evaluate. The specific aims of this proposal are to: (1) Determine the contribution of host immunity to dormancy exit and tumor recurrence caused by JAK1/2 inhibition using genetic and pharmacological approaches in GEM models, and (2) Identify the mechanisms by which JAK1/2 inhibition affects dormancy and tumor recurrence by examining the role of STAT proteins, JAK kinase activity, and cell cycle regulators using in vitro and in vivo GEM models. Because we hypothesize that JAK inhibitors may be harmful when given to breast cancer survivors, we must use model systems to test this hypothesis. The mouse has been chosen as a model for our studies both because of the similarity of its mammary gland to humans and because it permits us to accurately model biological variables in patients critical to dormancy and recurrence, such as the immune system and tumor microenvironment. Elucidating the pathways that enforce tumor dormancy and contribute to recurrence is a critical priority for cancer patients. By expanding our understanding of the mechanisms by which dormant RTCs survive and recur, this project has the potential to inform therapeutic approaches in patients who harbor dormant MRD following treatment of their primary cancer. Ultimately, the abilities to therapeutically target escape pathways used by dormant RTCs, while avoiding iatrogenic activation of RTCs that would otherwise remain dormant, has the potential to prevent recurrence, thereby providing new treatment options for millions of cancer survivors.

NSF Awards · FY 2026 · 2026-06

The Philadelphia Harmonic Analysis and Differential Equations (PHADE 2026) conference will be held from July 18th to July 22nd, 2026 at the University of Pennsylvania in Philadelphia. The conference is an officially-designated satellite conference of the 2026 International Congress of Mathematicians, held at the Philadelphia Convention Center beginning on July 23rd. The PHADE 2026 program includes 50-minute talks by 38 international leaders in the field of harmonic analysis and partial differential equations, with particular emphasis on highlighting a number of groundbreaking and transformative advances which have occurred in the field within the past three to five years. The conference will also provide all participants with the opportunity to present their research through poster sessions running throughout the week. A substantial focus on the professional development of students and early career mathematicians gives this event the potential to have long-lasting and far-reaching impacts on the development of the United States mathematical research workforce. Students and other early-career researchers will be given opportunities to disseminate their work through poster sessions and a conference webpage for the distribution of mini-posters. All early career participants will have the opportunity to network and learn from peers and senior leaders during the conference. NSF funding provides valuable participant support. The invited talks will highlight advances in five key areas: subelliptic PDE and generalized pseudodifferential operators, decoupling theory, oscillatory integrals, Geometric Measure Theory (GMT) and related elliptic PDE theory, and Fourier restriction, Kakeya, and Furstenburg problems. Leaders in these areas will be present to share their work with the harmonic analysis community, and by bringing these areas together at the same time, the conference will foster opportunities for making crucial connections. When viewed in its proximity to the ICM, a conference of this scale and significance has not occurred within the US harmonic analysis community in many decades. The broad participation of early-career researchers was a key factor in assembling the tentative schedule of invited talks, roughly half of which will feature talks by mathematicians less than ten years out of the Ph.D. Beyond these invited speakers, holding the event in close proximity to the ICM itself substantially multiplies these opportunities. More information may be found at the conference webpage, <https://sites.google.com/sas.upenn.edu/phade-2026/home>. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NIH Research Projects · FY 2026 · 2026-05

The adrenal cortex is a vital endocrine organ that produces steroid hormones essential for the body's homeostasis. Its organization into three concentric zones enables spatially regulated production of distinct adrenal hormones, a process known as functional zonation. Structural or functional defects in the adult cortex cause primary adrenal insufficiency (PAI), a life-threatening condition affecting millions with no permanent cure, while developmental disruptions in adrenal formation or zonation can lead to congenital adrenal disorders and tumors. In rodents, zonation is established perinatally via centripetal migration and stepwise “transdifferentiation” of subcapsular progenitors within the definitive zone (DZ), regulated by opposing actions of WNT-activating signals from outer capsule (Cap) cells and adrenocorticotropic hormone (ACTH). However, species-specific differences in adrenal organogenesis and steroidogenesis limit the translational value of rodent models for human PAI. To bridge this gap, we developed the first human adrenal organoid model from induced pluripotent stem cells (iPSCs) that recapitulates adrenal development both in vitro and in vivo. This proposal aims to exploit this platform to define and assess the therapeutic relevance of cellular and endocrine/paracrine signaling mechanisms regulating adrenocortical development, with the goal of transforming treatment of life-threatening adrenal diseases. In humans, adrenal cortical zonation begins prenatally, forming three zones: the DZ, the glucocorticoid (cortisol)-producing transitional zone (TZ), and the androgen-producing fetal zone (FZ), which correspond to the functional zones of the adult adrenal cortex. Notably, our organoids can be reliably directed to produce CD10/MME⁺ DZ-like cell (DZLC) progenitors that exhibit striking similarity to in vivo human DZ cells. DZLCs can be further differentiated into TZ-like cells (TZLCs) through combined stimulation with RSPO3 (a Wnt- ligand dependent potentiator of Wnt signaling) and ACTH, and subsequently into FZ-like cells (FZLCs) with ACTH alone. Exogenous RSPO3 is dispensable when DZLCs are co-encapsulated with iPSC-derived RSPO3- secreting Cap-like cells (CapLCs), which mimic the native capsule and enable ACTH-driven transdifferentiation of DZLCs into TZLCs. This contrasts with rodents, where ACTH promotes and WNT suppresses TZ fate. Thus, our in vitro directed transdifferentiation data support the central hypothesis that prenatal human adrenocortical homeostasis is orchestrated by DZ progenitors within a capsular niche that, through self-renewal and transdifferentiation, give rise to both TZ and FZ under coordinated control of WNT and ACTH signaling. Supporting this, our preliminary data show that encapsulation of DZLCs with CapLCs restores functional zonation after transplantation into a hemi-adrenalectomized immunodeficient mouse model, resulting in a long-lived adrenal cortex producing both cortisol and androgens in an ACTH responsive manner. Leveraging our adrenal organoid platform, we will determine whether these progenitor-derived populations can integrate into the host endocrine axis to fully restore adrenal function in vivo.

NIH Research Projects · FY 2026 · 2026-05

Project Summary Multimodal data have been generated and collected as part of HIV care delivery and research including, but not limited to, structured data and unstructured data in electronic health records (EHRs), claims data, pharmacy records data, imaging data, omics data and other molecular biomarker data. Such rich data offer great opportunities for harnessing the transformative power of artificial intelligence (AI) and machine learning (ML) to enhance personalized clinical decision support and address unmet needs in HIV prevention and care. Multimodal AI that can integrate multiple modalities of data encountered in clinical practice has been shown to yield superior performance over simpler, unimodal models in various disease areas outside of HIV. However, multimodal biomedical data are typically complex and heterogeneous, and are fraught with missing data and other sources of biases. For example, patients with less access to healthcare or lower socio-economic status tend to have more incomplete data in their EHRs. Thus, advancing multimodal AI for HIV applications faces significant technical challenges in the training, validation, and implementation, including, but not limited to, quantifying the dimension of heterogeneity, identifying interconnections, and addressing missing data. Another major barrier in advancing multimodal AI in HIV applications is that multimodal data in HIV are typically not publicly available. Our project seeks to address these and other challenges through three specific aims. In Aim 1, we will develop novel accurate, efficient and unbiased multimodal AI models for HIV care and prevention. In Aim 2, we will adapt and create causal knowledge graphs to enhance interpretability for applications in HIV care and prevention. In Aim 3, we will develop synergistic integration of knowledge graphs and multimodal AI models for more precise model and increased usability in HIV care and prevention. We will train and test the proposed multimodal AI models and knowledge graphs using multimodal data from the Veteran Health Administration, the largest integrated health system in the US, and the Veteran Aging Cohort Study for three important use cases in HIV prevention and care, namely, 1) identification of HIV patients at risk of medication non-adherence and/or loss to care; 2) prediction of complications of HIV patients; and 3) identification of patients at high risk of HIV infection. Our model development will be guided by ethical principles to ensure data privacy, security, and transparency. We will adopt a human-centered approach that seeks valuable inputs from and meaningful engagements with key stakeholders informed by the theory of Participatory Action Research. Once successfully completed, our project is expected to advance the state-of- the-art multimodal AI and knowledge graphs that can be applied/adapted to other use cases in HIV and transform HIV care and prevention.