Harvard University

NIH Research Projects · FY 2026 · 2026-02

PROJECT SUMMARY The proposed project aims to elucidate the neuron-specific mechanisms by which SNORD116 loss contributes to Prader-Willi syndrome, a complex neurodevelopmental disorder. We seek to comprehensively characterize the role of SNORD116 in ribosome biogenesis and translation in neurons using advanced high-throughput sequencing techniques like chimeric eCLIP and ribosome sequencing. Preliminary data indicate that Snord116 interacts with rRNA and a neuron-specific homologue of the methyltransferase Fibrillarin, FBLL1, suggesting that Snord116 may act as a ribosome biogenesis factor to meet the unique translation demands of neurons. By testing the consequences of these interactions, we aim to fill a critical gap in understanding Prader-Willi syndrome pathogenesis. Our objectives are to (1) determine the role of Snord116 in ribosome biogenesis, (2) assess the effect of Snord116 loss on neuronal protein synthesis, and (3) explore co-factors like FBLL1 that may modulate its function. Investigating these early molecular events could open possible avenues for correcting the distinctive growth and neurobehavioral features of Prader-Willi syndrome and offer broader insights into other neurodevelopmental disorders.

NIH Research Projects · FY 2026 · 2026-02

Abstract Recent advancements in immunotherapies have demonstrated the potential of antigen-specific patient-derived T cells to combat various diseases, notably cancer. Concurrently, induced pluripotent stem cell (iPSC)-derived technology offers a customizable, potentially limitless source of human T cells for immunotherapies. However, current efforts to generate functional T cells from iPSCs via in vitro selection fall short in achieving efficient T cell production and often produce aberrant T cell phenotypes compared to those educated in the native thymus. While thymic epithelium derived from iPSCs can support T cell differentiation after in vivo transplantation, there has not been success in identifying suitable mesenchymal cells from iPSCs to support thymic epithelium development ex vivo. Additionally, critical T cell education events involve trafficking from thymic cortical to medullary zones during differentiation but reproducing these zones or trafficking dynamics in vitro has not yet been demonstrated. The patterning of cell assemblies in space and time is crucial to tissue and organ development and is likely key to replicating the cortical and medullary thymic tissue sub-architecture in vitro. While the importance of chemical morphogen gradients has long been appreciated, it is increasingly clear that both the specific ECM molecules to which the cells adhere and the dynamic interaction between cell behavior and the matrix, with its time-varying mechanical properties, are important players in morphogenesis. Thus, tissue organization is impacted by the viscoelastic properties of the matrix, which vary from an elastic solid-like response to a liquid-like viscous response, in addition to ECM stiffness and composition. Our long-term goal is to combine iPSC-derived progenitors at the appropriate developmental stage with biomaterials that mimic thymic niches to scale the production of antigen-specific iPSC-derived human T cells for possible future clinical applications. We hypothesize that combining (1) developmentally matched, isogenic iPSC-derived thymic epithelium and mesenchyme with (2) instructive biomaterials that specify thymic zonal identity and allow for trafficking of differentiating iPSC-lymphoid progenitors between these zones will recreate thymic education in an in vitro platform and address the limitations observed in current iPSC-T cell derivations. We will explore this hypothesis through the following: (Aim 1) Developmentally match iPSC-derived mesenchyme and epithelial progenitors for thymic potential, (Aim 2) Engineer biomaterials and culture systems that mimic cortical and medullary thymic niches, and (Aim 3) Evaluate the impact of thymic mimicking niches on iPSC-derived T cell development and TCR repertoire. Success in this project will provide a robust platform for generating patient-specific T cells with functional competency comparable to thymus-educated T cells, generate insights that will have broad implications for cellular and molecular immunology as well as significantly advance the field of T cell immunotherapy using iPSCs.

NIH Research Projects · FY 2026 · 2026-02

Project Summary The human gut microbiota plays a critical role in health, with one of its major roles being to protect the host from microbial pathogens through a phenomenon known as colonization resistance. Effective colonization resistance is associated with increased bacterial diversity within the gut microbiota, which likely enhances the functional capabilities of the community. One microbial function thought to be important for colonization resistance is the production of antimicrobial small molecules by gut bacteria. However, our current understanding of the specific gut microbial metabolites involved in colonization resistance is limited, with identified compounds being largely restricted to previously known microbial products, including short-chain fatty acids, bile acid derivatives, and antimicrobial peptides. We hypothesize that the gut microbiota produces additional metabolites with antimicrobial activity that that are important for colonization resistance. The overall objective of this application is to discover such compounds through high-throughput screening. Specifically, we will screen a library of gut bacterial culture and supernatant extracts for growth inhibitory activity against a panel of bacterial and fungal pathogens and pathobionts relevant to the human gut microbiota (Aim 1). We will then isolate and structurally characterize antimicrobial metabolites from gut bacterial supernatants or cell pellets with promising activity (Aim 2). The proposed studies are supported by a preliminary screen demonstrating the presence of antibacterial and antifungal activity from multiple extracts in the library. To our knowledge, high-throughput screening has not been used previously to identify antimicrobial metabolites from members of the human microbiota. The proposed research therefore has high potential to identify antimicrobial compounds from these organisms, setting the stage for future efforts to understand the biological activities and roles of these metabolites within the gut microbiota, including their contributions to colonization resistance. Ultimately, our findings will provide a foundation for developing novel antimicrobial therapies and enhancing the resilience of the gut microbita against infections.

NIH Research Projects · FY 2026 · 2026-02

Project Summary/Abstract The hydrochlorination of alkenes is a central reaction in introductory organic chemistry, illustrating fundamental concepts in organic mechanism and reactivity. Despite its pedagogical ubiquity, and the growing importance of alkyl chlorides in synthesis and drug discovery, enantioselective hydrochlorination of alkenes remains an unsolved challenge in organic synthesis. This proposal aims to develop a catalytic, enantioselective hydrochlorination reaction of simple olefins by employing hydrogen-bond-donor (HBD) organocatalysis to provide access to enantioenriched alkyl chlorides (K99). The research plan outlines specific strategies that will enable enantioselective hydrochlorination through identification of a suitable HBD catalyst framework and cooperative metal halide cocatalyst. A combination of mechanistic and computational investigations will provide insight into the critical non-covalent interactions and catalyst features necessary for enantioinduction. Subsequent investigations on the stereospecific displacement of the chiral alkyl chlorides will enable the asymmetric synthesis of diverse chiral motifs. These products are well poised for further synthetic derivatization or direct biological evaluation for small molecule-based drug discovery. By facilitating access to these value- added functional units, underexplored chemical matter will be interrogated in biological contexts, contributing to the discovery of new therapeutic technologies for the betterment of human health. In a second research area (R00), cooperative metal halide catalysis will be leveraged to enable the formal activation of canonically inert quaternary stereocenters for enantioconvergent coupling reactions. Dynamic covalent reactivity will be exploited in concert with chiral transition-metal complexes to promote enantioselective reactions from racemic precursors. Mechanistic and computational studies will provide insight on the origins of enantioselectivity and general design principles in cooperative asymmetric catalysis. The proposed career development project will serve as a platform for the teaching and mentoring of high school, undergraduate, and graduate students. The proposed development plan also outlines professional networking and conference opportunities focused on launching a successful independent research career.

NSF Awards · FY 2026 · 2026-02

Entangled materials—such as polymer networks, textiles, and steel-cable structures—are found across length scales and exhibit remarkable mechanical properties driven by both fiber properties and the complex ways fibers entangle and self-contact. However, their behavior remains difficult to predict and design due to the lack of simple models capturing their intricate geometries and physical interactions. This Designing Materials to Revolutionize and Engineer our Future (DMREF) project will address that gap by developing quantitative metrics of entanglement through experiments and microscopy across scales. These metrics will connect entanglement geometries to physical properties, enabling the creation of simplified digital network representations of complex entanglements. These representations will guide the design of future entangled materials with user-defined properties. The project will provide open-source tools and data to support scalable design and optimization of fabrics, textiles, and knits, particularly at industrial scales. Broader impacts include educational integration of network science across institutions and a public art exhibit that will focus on visualizing networks, aiming to raise awareness of network-science-driven materials engineering. The project will establish a closed-loop framework for describing entangled matter using physical networks, correlating structural features with mechanical performance, and using these insights for targeted design. It consists of three unified thrusts that combine theory, computation, and experimentation. To span multiple length scales, the team will use testbeds made of 3D-printed textile architectures and woven metamaterials. Quantitative mechanical measures of entanglement will be obtained both experimentally and numerically. This data will inform the development of network models in which filaments are converted into skeleton and contact networks with geometric and topological attributes. These models will then be used to optimize entanglement geometries for desired performance using graph neural networks and gradient-based refinements. The result will be new material prototypes with engineered entanglements and mechanical properties, along with a broadly applicable design methodology for entangled filament-based materials. 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 · 2026-01

Project Summary/Abstract Research Project: The ability to learn new behaviors, or movements to achieve a goal, is fundamental to all animals. This ability largely relies on trial-and-error, or reinforcement, learning, in which different behaviors are attempted and reinforced if successful. Although the core concept of reinforcement learning is relatively simple, the neural computations supporting it are remarkably complex. Neural circuits must purposely generate high behavioral variability during early learning to explore candidate solutions, drawing on prior knowledge to efficiently “guess” which behaviors to try. As successful behaviors are learned, circuits must then store the solutions without overwriting instructions for behaviors used in other tasks. In the motor domain, these processes are widely believed to depend on cortico-basal ganglia (BG) circuitry. Yet, the specific mechanisms by which it accomplishes these operations for motor learning remain poorly understood. How do the BG leverage prior knowledge (i.e. previously learned behaviors) to generate and constrain motor variability for the purposes of learning new behaviors? How does the circuit modify synaptic weights for new skills while preserving prior learning? Progress on these questions has been hindered by the lack of methodologies to probe and track neural activity over long timescales of learning while simultaneously tracking evolving motor outputs, as well as a lack of candidate models with the power to link neural data with putative mechanisms. However, recent advances in theory and experimental techniques can now overcome these challenges. These include machine vision-based methods to track 3D behavior, techniques for monitoring neural activity over long learning timescales, and optogenetics to test key mechanisms. Using these tools, the proposed research will view these questions through the lens of two critical nodes of the cortico-BG circuit: dorsolateral striatum (DLS), the input nucleus of the motor BG, and its main input, the motor cortex (MC). This will involve using optogenetics to test hypotheses regarding how MC-DLS interactions generate motor variability (Aim 1) and using models and long-term recording to investigate how DLS stores the neural instructions for task-specific movements without overwriting existing ones (Aim 2). Techniques and insights from these aims will then fuel an investigation for how MC may store prior experience and leverage it to expedite learning on future tasks by constraining the activity of downstream structures (Aim 3). These studies will provide crucial insight into how neural circuits balance flexibility for learning with robustness in execution and illuminate principles driving adaptive behavior. Crucially, this work will also shed light on how BG dysfunction in disorders such as Parkinson’s and Huntington’s Disease result in such profound motor learning deficits. The proposed research will be conducted in the Ölveczky Lab at Harvard, which provides an excellent scientific and training environment, and under the guidance of an expert team of mentors and collaborators to advise experimental and computational aspects and professional skills. Together, this will propel the candidate towards her goal of launching a career as an independent researcher.

NSF Awards · FY 2025 · 2025-12

Following the collapse of a cloud core to form a star, the organic-rich icy grains can become incorporated into planet-forming disks, and the presence of large quantities of organic molecules may radically change the chemical trajectory of a forming planet or moon. It is currently unclear if these interstellar organic inventories survive the different stages of star and planet formation. This project will experimentally determine the UV and electron-induced photodissociation cross section of a range of common interstellar organics when embedded in interstellar ice analogs. The project will provide a first comprehensive view of the survival of interstellar organic molecules and organic functionality during disk formation. The project will train a graduate student and a postdoctoral researcher to be interdisciplinary scholars, crossing boundaries between chemistry and astronomy, theory and experiments. This project will experimentally explore the ice-phase UV and electron photodissociation cross sections and dissociation products of 15 organic molecules. The experiments will be carried out under ultra-high vacuum conditions using a pre-existing experimental set-up that enables the continuous monitoring of the ice composition using sensitive FTIR spectroscopy. The experiments will constrain the branching ratios for dissociation events into destruction vs. preservation of chemical functionality. The organic molecules are selected to explore how organic size and functionality affect their resistance to photo destruction, which will be used to make reasonable extrapolations to the complete interstellar organic inventory, thus enabling predictions of the overall interstellar organic survival up until incorporation into icy planetesimals. The immediate result will be a comprehensive database of dissociation cross sections for interstellar model organics across a range of relevant ice conditions. 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 2025 · 2025-12

Fast Radio Bursts (FRBs) are enigmatic pulses of radio waves that come from the distant Universe. In addition to the mystery of what produces them, FRBs can be used as a powerful tool to probe the missing ordinary matter in the Universe. This is the so-called "missing baryon problem", which has posed a challenge to astronomers for decades. The research team at Harvard University have a large data sample of FRBs thanks to the Deep Synoptic Array, an array of radio telescopes near Big Pine, California. They will use this sample to answer critical questions in the distribution of cosmic baryons. These investigations should mature the field of FRB cosmology and prepare the ground for the next generation of radio telescopes, such as the DSA-2000 to be built in Nevada. The team will create interactive 3D visualizations of the baryon cosmic web, with the goal of engaging the public and the broader scientific community. This proposal will use a large existing sample of localized Fast Radio Bursts (FRBs) to measure and map the Universe’s "missing" baryons. FRBs are millisecond radio pulses whose physical origin remains a key mystery in astronomy. Independent of their origins, the dispersion measure of FRBs directly probes all ionized gas between the observer and the source. To fully realize the long-sought goal of FRB cosmology requires a larger sample and new statistical methods that incorporate knowledge of intervening galaxies in their line-of-sight. The proposing team will use a growing catalog of localized FRBs discovered by the 110-antenna Deep Synoptic Array (DSA-110) to measure the distribution of baryons in the cosmic web. The current sample of roughly 70 localized FRBs is expected to double in the next twelve months, thanks to a significant increase in the number of DSA antennas. By 2027, the number of localized FRBs will exceed several thousand, as CHIME/VLBI, CHORD, and the DSA-2000 come online. The coming onslaught of localized FRBs and all-sky galaxy surveys requires a new framework for inferring cosmological gas parameters from observations. This work will be bolstered by forward-modelled mock FRB surveys in cosmological simulations such as MilleniumTNG. 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 2025 · 2025-12

In the current world, there is an ever-growing need for data transmission: not only does more information need to be transmitted, but also more devices are transmitting data: from electric grid infrastructure to traffic signals, more efficiency can be achieved if all equipment is communicating with each other. Wireless communications are simple and easily deployable; however, the range of frequencies available for existing techniques is currently limited to below 300 GHz, thereby not accommodating future growth as devices using the same frequency may be talking over each other. The terahertz (THz) frequency region (from 300 GHz to beyond 1 THz) is located beyond where typical radio equipment typically operates and such high frequencies are particularly suited for these challenges as it not only offers more bandwidth but a larger usable frequency range. Because the wavelength of THz waves is shorter than that of lower frequency radio waves, highly directive beams that are less sensitive to obstacles can be obtained. The aim of this project is therefore to develop new technologies utilizing the terahertz frequency region to obtain high speed and highly directive data transmission between a transmitter and a receiver, using dynamic obstacle avoidance. In addition, water vapor from the atmosphere absorbs light at certain THz frequencies; the use of a widely tunable source of THz will enable the choice of a specific propagation range. The low divergence of THz beams, combined with this tunable propagation range, can drastically limit the eavesdropping and detection probability therefore creating very private communications links. In addition, these techniques would also enable high-speed wireless connections across challenging terrains and into remote areas. The primary goal of this project is to develop two key devices that allow efficient and reconfigurable THz communication links: an amplitude modulator and a spatial light modulator. An amplitude modulator operating at THz frequency is critical to obtaining any sort of data transmission; however, the use of a spatial light modulator is only necessary if one wants to achieve reconfigurable beams (i.e., dynamic beam steering for tracking the detector, or obstacle avoidance). By combining these two devices with state-of-the-art THz sources and detectors, communication links at high data-rate with an intrinsically limited range will be demonstrated. To achieve high speed modulation of a THz field, the free carrier concentration in a low doped (and highly transparent in the THz) semi-conductor such as high-resistivity silicon, will be electrically modulated. The fabrication of such a modulator will require the use of conductive and THz-transparent electrodes using materials such as graphene. The spatial light modulator will be designed using micro-electro-mechanical system (MEMS), where wavelength scale movable mirrors will allow for a full phase front control of the THz beam, thereby allowing the dynamic shaping of the transmitted THz light to avoid obstacles (e.g., using bottle beams). Finally, the two devices will be combined to demonstrate two communication datalinks: an indoor link with obstacle avoidance (<10m) and an outdoor link with high directivity (>100m). These two demonstrations will show the relevance of THz technologies for the future of wireless communications. 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 2025 · 2025-12

PROJECT SUMMARY/ABSTRACT The brain’s ability to maintain stable internal representations of the external world is critical for consistent behavior. Yet recent studies show that neural representations can gradually shift over time, a phenomenon known as representational drift. While initially observed in associative areas like the hippocampus and parietal cortex, drift also occurs in primary sensory regions such as the piriform cortex, a main site of olfactory identity processing. This challenges our understanding of how stable odor-guided behaviors persist despite ongoing neural change. The olfactory tubercle (OT) receives direct input from the olfactory bulb and dopaminergic input from the ventral tegmental area (VTA), positioning it as a site where odor identity and reward value may be integrated to support stable representations. However, it remains unclear whether OT representations are stable or flexible, how they change with learning, and what role dopamine plays in this process. This project will investigate the temporal dynamics of odor representations in the OT and test whether dopamine contributes to their stability or flexibility. By conducting this work in the Molecular and Cellular Biology department at Harvard University, I will be supported by world-class resources and mentorship to elucidate the mechanisms underlying odor representation dynamics in the following aims. In Aim 1, I will use miniscope calcium imaging in behaving mice to track OT population activity during a two-alternative forced choice (2AFC) odor discrimination task. I will assess whether odor representations remain stable during continual reinforcement or drift when reinforcement is paused. I will then test whether new odors disrupt existing representations, and whether removing odors from training accelerates drift. In Aim 2, I will selectively ablate VTA-derived dopamine inputs to the OT using 6-hydroxydopamine (6-OHDA) and examine how this affects drift. I will assess whether dopamine stabilizes representations during extended training or protects against drift when cue-outcome training is paused. Finally, I will test whether dopamine depletion impairs or enhances the incorporation of new odor information, shifting the balance between flexibility and stability in the OT. Across both aims, I will use statistical and computational tools, including similarity metrics, dimensionality reduction, and decoding models, to quantify drift and uncover its structure. This work will advance our understanding of how the brain preserves meaningful sensory representations and how dopamine shapes the trade-off between stability and adaptability. The activities in this grant will anchor my training in systems and computational neuroscience and prepare me for a successful research career. To further my development as an independent scientist, I will supplement this training by presenting my work at national and international conferences, publishing my findings in peer-reviewed journals, and by training the next generation of scientists through teaching and mentorship.

NIH Research Projects · FY 2025 · 2025-12

Project Summary: Atrial fibrillation (AF) is the most common cardiac arrhythmia, significantly increasing the risk of myocardial infarction, stroke, and heart failure. Despite its widespread impact, the mechanisms underlying AF remain incompletely understood. A key driver of AF is atrial fibrosis, which alters the heart’s structure and disrupts electrical signaling, creating a substrate for arrhythmia initiation and progression. Effective prevention and management of AF, therefore, requires targeted strategies to address fibrosis. Growth differentiation factor 11 (GDF11), a TGF-β superfamily member, has shown anti-fibrotic properties in the heart, distinguishing it as a promising therapeutic target. Although GDF11 shares ~90% sequence homology with its counterpart GDF8, their biochemical functions are distinct. Recent work from our lab has established methods to reliably differentiate these ligands and highlighted their unique roles in cardiac health. While circulating levels of GDF11 and GDF8 measured by mass spectrometry show no correlation with cardiovascular disease, aptamer- based dual detection of GDF11/8 cleaved mature domains has robustly predicted AF in at-risk patients. The activation of GDF11 requires cleavage of its inhibitory prodomain by Tolloid (TLD) proteases, a critical step in enhancing its biological activity. Notably, unpublished single-nuclei data from our lab reveals reduced TLD transcript expression in the left atrial appendages of AF patients. Moreover, exogenous administration of TLD- cleaved GDF11 significantly reduces cardiac fibrosis in preclinical models, underscoring TLD protease activity as a vital mechanism in atrial fibrosis and AF pathogenesis. This proposal aims to investigate the role of TLD proteases in experimental atrial fibrosis through the use of gene-edited mouse models, both with and without surgically induced atrial fibrosis. Additionally, the study will explore the therapeutic potential of exogenous, prodomain-cleaved GDF11. By elucidating the molecular mechanisms linking TLD proteases and atrial fibrosis, this research seeks to advance our understanding of AF pathogenesis and identify novel therapeutic approaches for its treatment.

NSF Awards · FY 2025 · 2025-10

As artificial intelligence (AI) systems become more powerful and self-determining, they are shifting from simple tools to collaborative partners that can work with humans on complex, long-term projects. However, current AI systems may not effectively maintain communication with humans over extended periods. They often fail to track preferences, follow instructions, and/or adapt to project-specific context. The proposed work will investigate how to design AI systems that can establish and maintain a shared understanding with humans when collaborating over long-term projects. The work will advance the national interest by developing foundational principles and software for trustworthy AI systems that can support human creativity and productivity in areas such as game design, long-form document writing, and software development . The project will produce open-source tools and guidelines that will benefit industry developers, researchers, and users of AI systems across the United States and Canada. Ultimately, these AI systems will enable more effective and reliable AI assistance across the timeframes of real-world projects and accelerate the development of future AI applications. This research develops novel interaction techniques and technologies for establishing, maintaining, and managing common ground between humans and AI systems during long-term collaborative projects. Common ground represents the intersection where AI systems remain aligned with user intentions as projects evolve and grow complex. This research introduces and explores the interaction model of an “intent specification” (IS). In this work, IS means a human-readable representation of user goals, preferences, and project understanding that grounds long-term AI decision-making. The project investigates three key research questions through empirical studies and system development: 1) how to help users construct and refine their intent by measuring the effect of design decisions informed by cognitive learning theories; 2) how to align AI system understanding with user intentions through interface designs that reify grounding acts from communication theory; and 3) how to manage the accumulation of user intent and project understanding at scale efficiently and accurately through the development of semantic conflict detection and resolution techniques. Findings will be validated through two application domains, game design and long-form document writing. These domains will test generalizability of approaches with empirical methods. These methods will include usability studies of system usefulness (controlled and in-the-wild) and technical evaluations. The project will produce IntentTracker, an open-source software library for managing user intent in AI systems, along with prototype applications GameJammer and DocJammer. By combining theoretical foundations from research in cognitive science and communication theory with evaluations of interface designs and techniques, this work will create design principles for next-generation AI systems. 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 2025 · 2025-10

Artificial intelligence systems are increasingly trained using datasets containing private information about individuals in critical areas such as government services, healthcare, and education. However, these AI systems have a demonstrated risk of accidentally revealing sensitive personal information about the people whose data was used during training, creating serious privacy and security concerns. This problem threatens public trust in AI technologies and creates barriers to beneficial uses of AI in sensitive domains where privacy protection is essential. Currently, many organizations cannot safely use AI because existing privacy protection methods are either inadequate or too difficult to implement correctly. This project addresses this challenge by developing freely available software tools that prevent these privacy vulnerabilities in future AI systems. These tools will make state-of-the-art privacy protection methods practical and accessible to a broad community of developers and researchers. This work serves the national interest by advancing privacy protection for all citizens, strengthening trust in AI technologies used by government and industry, supporting American competitiveness in privacy-preserving AI development, and enabling secure use of AI in critical national infrastructure while protecting individual rights. This project advances privacy-preserving machine learning by developing and implementing novel techniques for training large models with the strong protections of differential privacy and minimal overhead in computation and model performance. The research activities focus on incorporating tools for differentially private stochastic gradient descent into OpenDP, a community-driven open-source software project with a rigorous vetting process. First, these tools will integrate with Opacus, the open-source differentially private machine learning library developed by Meta, with OpenDP efforts strengthening the privacy guarantees offered by Opacus while making the library more easily usable by the OpenDP community. Second, the investigators will improve the efficiency and utility of differentially private stochastic gradient descent by optimizing the choice of noise distributions and their samplers. Third, the team will implement sophisticated privacy accountants to measure the protections of differentially private stochastic gradient descent as the algorithm runs. Finally, the investigators will develop and implement a framework in OpenDP for differentially private federated learning with precisely specified privacy guarantees. The project will explore applications in genomics and educational technology, demonstrating privacy-preserving data utilization across multiple domains while ensuring the trustworthiness of the software through community-driven development and expert vetting rather than reliance on single commercial entities. 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 2025 · 2025-10

Healthy indoor air quality is paramount to human health, especially considering that on average, Americans spend approximately 90% of their time indoors. Current commercial indoor air quality sensor systems are unable to identify or distinguish specific chemicals in the air, including toxic compounds from natural substances. As a consequence, these systems fail to determine the level of harm and alert about negative effects on health, which can contribute to cancers, cardiovascular and neurodegenerative diseases and other medical conditions. Current air quality sensors also tend to report inaccurate concentrations of chemicals in the air, require frequent recalibration, and are expensive. To address this technology gap and to keep Americans safe, this project seeks to develop and commercialize a sensory nature-inspired technology platform for monitoring indoor air. The product is a low-cost, highly accurate indoor air quality monitoring system that provides precise, real-time information about hazardous chemicals in indoor air. Offering seamless building integration, the technology first alerts customers, such as building managers. Knowing the status of the air inside each room and space gives the customers the ability to quickly identify problems and take immediate action, such as building ventilation adjustment, air cleaning, or, in extreme cases, tenant evacuation. In addition to the immediate public health benefits, the outcomes of this project provide impacts across multiple U.S. sectors that require real-time gas sensing, including disease diagnosis, search-and-rescue, food spoilage monitoring, and hazardous waste identification. The project’s technology platform for monitoring indoor air incorporates bio-inspired technological concepts to overcome the limitations of current indoor air quality sensor systems, providing exceptional sensitivity, specificity, and robustness to variations in factors like temperature and humidity. Specifically, the technology’s key elements include (1) an array of inexpensive off-the-shelf chemiresistive sensors, (2) the modulation of the intake and expulsion of air (akin to biological sniffing), (3) state-of-the-art temporal data processing, and (4) machine learning models to distinguish specific chemicals in the air. As a low-cost, highly accurate, real-time indoor air quality monitoring system, this technology is poised to improve the ability to measure indoor air and prevent negative health effects stemming from hazardous air quality, and usher in a new era of indoor air quality sensing. This investment will de-risk this technology platform, positioning it for potential future commercialization by the private sector. 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 2025 · 2025-09

Project Summary Piperidines are the most common nitrogen-containing, privileged scaffold, present in numerous FDA- approved drugs. However, to derive desired physiochemical properties, biological activities, and target selectivity, chiral substituents are added to the piperidine core in enantio- and diastereoselective manners. Accordingly, medicinal chemists prepare various stereoisomers of the drug candidate to test and compare their properties, but the traditional preparation of chiral piperidines drugs are a target-oriented synthesis which that relies on chiral ligands, catalysts, and auxiliaries for every derivative. Therefore, chiral piperidine synthesis remains a bottleneck in drug optimization campaigns. To address this challenge, I propose that a hydrogen-bond donor (HBD) catalyst can generate a common chiral environment that enables a general, diastereoselective strategy for selectively accessing highly substituted diastereomeric piperidine products bearing an epoxide moiety. Once the common HBD/iminium ion intermediate is created upon the addition of a Lewis acid, two subsequent synthetic pathways emerge: the “oxidation-first” pathway and the “nucleophile-first” pathway. Hence, the proposed diastereoselectivity will be controlled by order of addition of nucleophiles and oxidants; since the positive charge of the common intermediate persists when the oxidant is added first, the HBD catalyst is still bound to the substrate which presents an opportunity to set different chiral centers. If nucleophile is added first, then the positive charge of the iminium ion is quenched, and the stereochemistry of the subsequent epoxidation will be controlled by the d.r. of the substitution at the C2 position. This proposed system will be realized by two aims. First, the prochiral piperidine substrate will be rapidly diversified by adding amine-, alcohol-, and TMS-based nucleophiles to form new C-N, C-O, and C-C bonds at the C2 position. To ensure success of this aim, the structure of HBD catalyst will be rigorously optimized by tuning its pyrrolidine arm, anion-binding bridgehead, and electron-withdrawing substitutions. Once optimal catalysts and conditions are found for each type of nucleophile, mechanistic studies will be performed to understand molecularity, rate-, and selectivity-determining steps. Then, the resulting epoxide moiety in the chiral piperidine product will be further functionalized through nucleophilic ring-opening strategies. In the second aim, a concurrent, reigodivergent reduction of the epoxide moiety of the HBD/iminium ion intermediate will be carried out to achieve net reductive resolution and 1,2- transposition of the 4-OH group of the piperidine substrates. This goal will be achieved by using magnesium catalysts with butyl and bistriflimide ligands that facilitate regioselective hydride attack at the epoxide. Overall, the transformations outlined in this proposal closely mirror the crucial principles of drug design, lowering the barrier to accessing valuable chiral piperidine building blocks.

NIH Research Projects · FY 2025 · 2025-09

ABSTRACT Metabolic health depends on acquiring an appropriate amount of energy and allocating it efficiently across tissues. The human gut microbiome modulates energy acquisition and allocation and exhibits high sensitivity to modifiable lifestyle factors such as diet, exercise, and social contact, resulting in gut microbiome contributions to metabolic health that differ across individuals and within individuals over time. Ecological plasticity of the gut microbiome is a double-edged sword: it may increase disease risk if the microbiome departs from profiles to which the human body has adapted, but it may also confer protection by affording humans some capacity to adapt rapidly to changing ecological conditions. However, health outcomes of lifestyle-associated alterations in the gut microbiome remain untested. This situation complicates the interpretation of human gut microbiome variation, hinders efforts to develop equitable microbiome-targeted therapeutics, and obscures our view of features characterizing a ‘healthy’ human microbiome. In the coming five years, we will perform experiment-based, causally informative research toward three synergistic themes that elucidate when, why, and how lifestyle affects gut microbiome contributions to metabolic health. In Theme 1, we interrogate how ecological factors such as dietary digestibility, caloric intake, and exercise shape gut microbiome contributions to host metabolic health and the conditions favoring inflection points between beneficial and detrimental contributions. In Theme 2, we evaluate how ecological context shapes the metabolic health consequences of host-microbiome interactions and whether it may benefit health to restore putatively beneficial taxa depleted in industrialized or non-industrialized gut microbiomes. In Theme 3, we investigate how social microbial transmission may mediate metabolic disease risk and resilience, testing the idea that non-communicable conditions have a communicable component by virtue of the social exchange of risk-associated microbes. To address Themes 1 and 2, we will exploit natural variation in human microbiomes and controlled diet and exercise interventions in mice to foster conditioned gut microbiomes that can then be transplanted into germ-free animals to study the causal health impacts of a given microbiome profile. To address Theme 3, we will examine the transmissibility of microbes between focal individuals with metabolic disease and their social partners, varying focal and partner social networks to evaluate the risk landscape of social transmission in relation to social position and connectivity. Work toward Themes 1-3 will be enhanced by three overarching principles guiding our work: (1) consideration of both bacterial and non-bacterial constituents of the gut microbiome, (2) rigorous host and microbiome phenotyping enabling the elucidation of underlying mechanisms, and (3) development of machine learning models that identify metagenomic and metabolomic features predicting gut microbiome contributions to metabolic health. Collectively, our proposed research will enhance our understanding of human energy metabolism and help lay the groundwork for future health-directed manipulation of the gut microbiome.

NIH Research Projects · FY 2025 · 2025-09

Project Summary/Abstract Nicotinamide Adenine Dinucleotide (NAD+) serves as a crucial redox cofactor in metabolism and acts as a substrate for poly (ADP-ribose) polymerase and sirtuins, playing roles in DNA repair, metabolism, and stress response and involving in cancer and longevity. Sirtuins consume NAD+ and produce a unique byproduct, O- acetyl-ADP-ribose (OAADPr), which holds significant potential as a signaling molecule in regulating various biological processes. Yet, the biological roles and metabolic regulations of OAADPr remain poorly elucidated. Colorectal cancer (CRC) is the third leading cause of cancer deaths in the United States. The poor outcome of CRC highlights an urgent need to identify mechanisms that regulate CRC metabolism and growth. Studies have shown elevated NAD+ levels and increased expression of NAD+ biosynthesis and salvage enzymes in CRC, suggesting a critical role of NAD+ metabolism in CRC progression. Sirtuins are involved in CRC and conflicting roles have been reported in CRC. Yet, the mechanistic links between sirtuins and cancer remain incompletely understood and the exact mechanistic links between NAD+-related metabolites and cancer remain unclear. To identify novel proteins that interact with NAD+-metabolites, I applied a proteomic thermal stability assay and discovered a previously unknown interaction between medium-chain specific acyl-CoA dehydrogenase (ACADM) and OAADPr. ACADM is a key enzyme in β-oxidation, and aberrant fat metabolism has long been recognized in CRC. I hypothesize that OAADPr may play a tumor-suppressive role in CRC through the inhibition of ACADM activity and fat utilization. I will test this hypothesis in two aims: 1) elucidate the mechanism of how OAADPr affects ACADM activity and 2) examine the impact of OAADPr-mediated inhibition on ACADM both in vitro and in vivo. In Aim 1, I will determine the kinetic and equilibrium parameters of the inhibition of OAADPr on ACADM and its mechanism. In Aim 2, I will modulate OAADPr levels in CRC cells by overexpressing (OE) SIRT3 and/or knockout (KO) MACROD1, two key enzymes that regulate OAADPr metabolism. With that, I will first assess fat oxidation in wild-type, SIRT3 OE, and/or MACROD1 KO CRC cells by measuring the rate of fatty acid oxidation, profiling acyl-carnitines, and conducting C13-palmitate tracing experiments. Then, I will assess if OAADPr modulates CRC proliferation and tumor growth by measuring cell proliferation in vitro, organoid growth ex vivo, and xenograft or spontaneous tumor growth in vivo with altered OAADPr levels. Lastly, I will investigate whether dietary supplementation of NAD+ precursors affects intratumor OAADPr levels and tumor growth in organoids and the genetically engineered mouse model. Our research is conceptually novel and will address an important gap in our understanding of the mechanism by which NAD+ metabolism affects tumor proliferation. The proposed study will also uncover new roles for OAADPr in metabolism and cancer and open a new area in NAD+ and sirtuin biology. The insight gained may open avenues for novel approaches leveraging NAD+ metabolism in cancer therapy.

NIH Research Projects · FY 2025 · 2025-09

Project Summary Transition metals are elements with unique reactivity that have changed the trajectory and success rate of small-molecule drug discovery and manufacturing over the past decades. A quarter of the top 20 most used reactions in medicinal chemistry are catalyzed by noble metals, and the majority of the top 200 selling small- molecule drugs now contain either a biaryl or arylamine linkage (motifs easily accessible by Pd-catalyzed cross- coupling). The dependence of medicine on transition metals is alarming, as most of the important metals are toxic, expensive, environmentally damaging, and, most importantly, increasingly scarce, with several facing effective extinction in the next century. Accordingly, there has been an urgent need to discover strategies to force other more available and benign elements to substitute for transition metals in these essential chemical processes. As an alternative, we ask whether, rather than using a single element, we might use entire organic molecules (or fragments) as oversized surrogates for metals (“pseudometals”). First, we aim to show that organic, metal-free catalysts can be capable of performing processes traditionally considered only possible with transition metals, with special emphasis on oxidative addition and reductive elimination. We will then assemble these elementary steps into catalytic mechanisms to achieve mimicry of cross-couplings, the most important class of metal-catalyzed reactions in drug discovery. Second, we will extend the pseudometal concept from mimicry to unprecedented reactivity, realizing metal-like transformations that are impossible with conventional, periodic-table metals. These new cross-couplings will enable access to desirable chemical space for accelerated drug discovery and more rapid construction of bioactive small-molecule agents. Finally, in a parallel direction, we will investigate the combination of electrochemistry with rationally tailored redox-active organic molecules to generate very weak X–H bonds akin to first-row metal hydrides, with the goal of functionalizing ubiquitous alkenes and alkynes through a biologically-inspired hydrogen-atom transfer process. This research goal will establish the potential of organic pseudometals to cover the whole range of transition-metal reactivity and beyond, providing a roadmap for the design of efficient and enabling catalytic transformations to sustain biomedical research in a transition-metal-free future.

NIH Research Projects · FY 2025 · 2025-08

Project Summary/Abstract Antibiotic-resistant Gram-negative infections pose a major threat to human health. Gram-negative pathogens are intrinsically resistant to most clinically used classes of antibiotics due to the presence of an outer membrane that prevents antibiotic entry. Many Gram-negative pathogens, including Pseudomonas aeruginosa, Acinetobacter baumannii, Klebsiella pneumoniae, and E. coli, are now multi-drug resistant and can only be killed by colistin, an old antibiotic that was previously almost abandoned because it has dose-limiting toxicity. My lab played a large role in discovering the machines that assemble the Gram-negative outer membrane and has developed a comprehensive set of in vitro and in vivo tools to study outer-membrane assembly. Here, we will use our existing tools as well as new approaches to elucidate the mechanisms of action of two peptide antibiotics that target the lipopolysaccharide transport (Lpt) pathway, thanatin and murepavadin. The molecular mechanisms by which these compounds disrupt outer membrane assembly remain poorly understood. We hypothesize that a molecular understanding of how these compounds block lipopolysaccharide transport will enable the development of analogs with more promising properties as drugs. In Aims I and II, we will address the mechanism of action of thanatin using a range of biophysical and biochemical assays, including single molecule TIRF microscopy in living cells, in vitro biochemical assays that monitor different steps in LPS transport, and in vivo crosslinking assays that monitor LPS transport in cells. In Aim III, we will investigate the mechanism of action of murepavadin, and we will characterize the Pseudomonas aeruginosa multiprotein complex that we propose it targets. Using cyclic peptides, we have recently established that this multiprotein complex is a valid therapeutic target in Acinetobacter baumannii to treat Carbapenem-resistant Acinetobacter baumannii (CRAB) infections, so we will also use the Pseudomonas multiprotein complex in a screen to identify other antimicrobial peptides that may share a similar mechanism of action. In this way, the fundamental knowledge we obtain about how thanatin and murepavadin function could enable discovery of other classes of compounds to expand the spectrum of antibiotic-resistant Gram-negative pathogens that are susceptible to lipopolysaccharide transport inhibitors.

NIH Research Projects · FY 2025 · 2025-08

This application requests support for the Biophysics Graduate Program at Harvard University. The mission of the Program is to provide students who have strong undergraduate backgrounds in quantitative sciences (especially physics and mathematics) with broad training in the biophysical, chemical, and molecular concepts and techniques that are required to address outstanding problems in biology and biomedical sciences. This profoundly interdisciplinary program supports the training of students from a broad, multidisciplinary spectrum of academic backgrounds and experiences to develop and use experimental, computational, and theoretical approaches to address important questions at the interface of physical and biological sciences. The cross-campus Biophysics Program unifies 60 Training Faculty from all across Harvard University, including Harvard’s main (Cambridge) campus and the basic science and clinical departments of Harvard Medical School (HMS) and its teaching hospitals, creating a highly interactive and collaborative training environment for impactful and rigorous biophysics research.  The Program offers a flexible curriculum with two required courses and electives drawn from offerings across the Harvard campuses. Courses and structured activities — including a 2nd-year mini-symposium, a preliminary qualifying exam based on an original research proposal on a topic distinct from the dissertation research, a student research seminar series, and an annual off-site Research Retreat with student and faculty research talks and a poster session — provide a strong foundation that emphasizes research design, rigor and reproducibility, and written and oral scientific communication to a variety of audiences. The Program supports trainees to pursue dissertation research in a variety of disciplines relevant to molecular biophysics with strengths in the areas of structural biology, computational biology, quantitative cell biology, single-molecule biophysics, neuroscience, and imaging. The Program also leverages rich institutional resources to further support the trainees’ professional development. The Program empowers the trainees’ career exploration by engaging with a rich alum network, including bimonthly “Chats with Biophysicists” events and an annual dinner gathering of admitted and current students with faculty and local alum during the Recruitment Visit. The Program aims to equip trainees for a wide range of science-related careers. 2009-2024 training grant eligible (TGE) alum outcomes indicate success: 70% are in primarily research positions; all others are in clinical or science-related careers.  In the past 5 years, the time-to-degree averages 5.4 years, and the Program continues to improve how it tracks and supports trainee progress. Over 35 years of funding, the previous grant has supported up to 16 trainees (12 currently). In this proposal we request support for 16 trainees; with rare exceptions, 8 students will each be funded in their 1st and 2nd years of graduate studies.

NIH Research Projects · FY 2025 · 2025-08

Project Summary/Abstract There exists a major gap in our knowledge on what genes and regulatory regions control the differences in skeletal morphology between humans, and our best studied model system, the mouse. This is despite the human skeletons’ importance to walking/running, tool use, and childbirth, and that congenital, trauma-induced, and aging diseases of the skeleton are very common. This gap has not been spanned by the recent efforts of large scale consortia (e.g., ENCODE), which have been generating extensive functional genomics datasets on human and mouse fetal and adult tissues to shed light on organ biology, but which have neglected the skeleton and its cell types. Here, we propose to remedy this situation and take a first step to fill the gap in knowledge through our proposal to conduct targeted studies on the human fetal skeleton, at site-specific anatomical levels, using the functional genomic techniques of single cell multiomics, to detect expressed transcripts/genes and regulatory elements per cell for each tissue, and spatial transcriptomics for each anatomical region to detect gene expression at high definition histologically. Importantly, we will generate transcriptomic and epigenomic maps for all the large joints in the post-cranial body, which will then be intersected with actual human disease-causing genetic variation from Genome-wide Association Studies to find causal variants. These will be tested in a high throughput assay to examine each regulatory variants impacts on gene expression. In Aim 1, single cell multiomics will be performed on human joints for the shoulder, elbow, wrist, interphalangeal joint of the hand, ankle, interphalangeal joints of the foot, and lumbar sacral joint to map the transcriptome plus epigenome of each cell type at each joint. These data will be combined with identical data we generated on the hip and knee, thus covering all large joints of the post-cranium. All of these datasets are compared across three timepoints to build human-specific maps, noting similarities and differences in transcriptomic and epigenomic usage, reflecting differences in the attainment of morphological differences in anatomical regions. In Aim 2, high-definition spatial transcriptomics will be performed on the same anatomical regions as in Aim 1 on human samples to reconstruct tissue (histological) level gene expression at single cell resolution. These data are then compared across time- points to build human-specific maps, noting similarities and differences in genic usage at the cell type, and anatomical-specific levels. In Aim 3, we propose to use a high throughout reporter assay, called the Massively Parallel Reporter Assay, to test the regulatory functions of thousands of human variants (involved in normal or disease biology) residing in musculoskeletal cell type regulatory regions. This approach will generate a compendium of human regulatory elements with effects on such cell type gene regulation, and in doing so will shed light on the complex regulatory architecture underlying normal human biology (e.g., the shape of the birth canal or our bipedal knee) and common human skeletal diseases such hip dysplasia or osteoarthritis.

NIH Research Projects · FY 2025 · 2025-08

Mitochondria (mt) are regarded as critical components for cardiomyocyte homeostasis and are the nexus of many signaling events that maintain normal cardiac function. As a result, mutations in mitochondrial DNA (mtDNA) have been shown to affect cardiac myocyte function and are associated with the development of cardiomyopathy in ~25% of all mtDNA mutation carriers. However, it is difficult to define mtDNA mutations as the cause of a particular pathology because the penetrance of mtDNA disease is not mendelian and is dependent on the amount of heteroplasmy of mtDNA mutations per cell, the cell type expressing mutations, as well as the specific mutation. Some canonical mtDNA mutations, in particular mt-tRNA mutations, have been shown to be displayed at higher heteroplasmic states and are defined as causative sources for cardiomyopathy, such as in mitochondrial myopathy, encephalopathy, lactic acidosis and stroke-like episodes (MELAS) syndrome caused by a mt-tRNALeu mutation (m.3243A>G). Many mtDNA diseases, such as MELAS, present with aging, yet it’s not completely defined why and how different cardiomyocytes or other cardiac cells i.e., endothelial or fibroblasts, are affected by the presence of MELAS mutations. Understanding of the mechanism of cardiomyopathy by mtDNA mutations would be advanced by useful mtDNA gene editing strategies. In this proposed research plan, we present a novel Cas9 based mitochondrial gene editor system (mEditors) that we designed to introduce mtDNA mutations into human IPSC cardiomyocytes (IPSC-CMs), with the ultimate goal to ascribing pathogenic and potentially causal relationships between specific mtDNA mutations and cardiomyopathy. We hypothesize that mEditors can be used to introduce, rescue, and regulate mtDNA heteroplasmy states of the common MELAS mutation (m.3243A>G), allowing hypothesis-testing on the presence and level of mtDNA mutations with functional effects on cardiomyocytes. The following aims are designed to test these hypotheses: Aim 1: To test the hypothesis that mEditorS can be used to introduce mtDNA mutations in human cardiomyocytes. Aim 2: To test the hypothesis that mEditorS can remove mtDNA mutations in human cardiomyocytes. Aim 3: To test the hypothesis that mtDNA heteroplasmy can be regulated by mEditors in conjunction with functional alterations in human cardiomyocytes. This project will advance a novel mtDNA editing tool with the goal for ascribing causality for mtDNA mutations and possible future therapeutic development for mitochondrial diseases.

NSF Awards · FY 2025 · 2025-07

This award funds research to develop a new methodology to distinguish between research findings that can be generalized across populations, places, and time, and those that cannot be generalized. This research addresses a fundamental challenge in empirical research: determining which experimental or observational findings are generalizable across different environments and populations. Existing methods do this by using restrictive assumptions, potentially leading to false generalization. By introducing a new methodology to detect generalizability, this research helps identify features of the environment, population characteristics, and treatment conditions that systematically contribute to generalizable results and those that exhibit context-specific or unpredictable results, and therefore not generalizable. The research results improve the reliability of evidence-based decision recommendations and the quality of decision design. By offering a rigorous approach to identify generalizability, this research makes significant contributions to economics science and beneficially informs decision makers and practitioners. The results of this research aid improved decision making, speed up economic growth, and hence improve living standards. This award funds a research agenda that develops a new methodology to distinguish between research results that are generalizable and those that are not. Methodologically, the research advances statistical meta-analysis by developing estimators and classification tools that distinguish between predictable (generalizable) and unpredictable (environment-specific) treatment effects. Unlike standard approaches, this framework allows researchers to pinpoint critical environmental or demographic factors driving effects heterogeneity. The resulting methodology can be integrated into various fields---including economics, public health, and education---enabling more nuanced insights into whether, when, and why certain policies or interventions are particularly effective. Empirically, the project applies these techniques to varied datasets in economics. 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 2025 · 2025-07

The water held in soil is a critical resource for sustaining ecosystems and agriculture and, through bare-ground evaporation and transpiration through plant leaves, a primary source of surface humidity over land and driver of the land-based hydrological cycle. But despite its central importance for continental climate and hydroclimate the basic science of soil moisture, at least at regional to continental scales, is not well developed. There are several reasons for the lack of basic science understanding, among them the complexity of factors that determine soil moisture, including physical climate factors like precipitation and the surface energy available for evaporation, and biological factors that determine how much water is taken up by plant roots and transpired through leaves. The development of a basic science is also hampered by lack of observations, and by a tendency for research to focus on more applied science issues like the use of soil moisture as an input for subseasonal weather forecasting. The Principal Investigator (PI) of this CAREER award seeks to develop the basic science of soil moisture at regional to continental scales by addressing two questions: first, what are the key controls on the spatial variability of soil moisture in present-day climate? One issue here is why the latitudinal profile of continental soil moisture has a characteristic "W" shape, with a maximum at the equator and minima in the subtropics of the Northern and Southern Hemispheres. Second, how will soil moisture change in a warming world? The question is motivated by the drying trends found in many of the world's semi-arid regions including the US Southwest, which raise concerns about future water resources as well as increases in the severity of heat waves and the risk of wildfire. But evidence of a drier future is inconclusive as climate models show subtantial disagreements in their projections of soil moisture change under warming, as well as important discrepancies with the observed record. Issues to be addressed under this award include why the soil moisture response to warming is muted compared to the precipitation response and what factors determine whether a region gets drier or wetter as a result of warming. The research questions are addressed through the construction of a hierarchy of soil moisture models starting from the simplest configuration and incrementally adding complexity to determine the most parsimonious version that can account for the behaviours of interest. The simplest configuration is the model of Stahl and McColl (2022) in which the moisture budget of a thin layer of soil (meaning moisture added by precipitation and removed by evaporation, transpiration, runoff, and drainage) is boiled down to a formula involving only precipitation and sunlight received at the surface. Despite its simplicity and neglect of important factors such as vegetation physiology the model successfully reproduces the annual cycles of soil moisture seen in a variety of soil moisture regimes over the globe. Complexity is added to the model through more realistic representations of factors including the nonlinear effects of soil saturation fraction on evaporation, runoff, and drainage. In addition to the work with simple models, which are compared to observations and climate model output, the PI conducts experiments with a high-resolution atmospheric model (a version of SAM, the System for Atmospheric Modeling, see AGS-2218827) coupled to a land surface model. The educational component of the CAREER project uses art to teach students about the water cycle and the balance of sources, sinks and storage that accounts for the presence of water on land. The effort builds on a pilot course developed by the PI in collaboration with the Harvard Art Museum in which students are asked to consider the water balance implications of images of water-dominated and water-depleted landscapes such as the Salton Sea and the surrounding drylands that were, until recently, submerged by it. The course, which is intended for a general audience and requires no specific math or physics background, is scaled up under this award to class sizes of 100 or more and is further developed into a class that can be taught indepedently at other universities. The class is developed so that it can be taught online and also tailored to museums near other colleges and universities, most of which have art work that has appropriate depictions of water. 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 2025 · 2025-05

This project will develop novel statistical theories and methods for handling large, heterogeneous datasets. Modern scientific applications often produce heterogeneous data of different types for the same problem. For instance, a single-cell biologist may observe multiple types of sequencing data from diverse instruments, all relevant for understanding the biological pathways of a single complex disease. The challenge lies in effectively combining these different data types to build statistical pipelines that outperform those developed using any one data type. Traditional statistical approaches struggle with this challenge. This project will establish a new statistical paradigm to address the complexities of such heterogeneous data while accounting for datasets with billions of variables. The project outcomes will facilitate principled prediction and inference in applications ranging from single-cell biology to precision health and neuroimaging. The project will involve graduate student participation and the development of new curricula at graduate and undergraduate levels that incorporate the project outcomes. Additionally, the research will engage medical professionals to facilitate the dissemination of the research products in current biomedical practice. This project will develop a modern statistical framework to address data heterogeneity in high dimensions, focusing on three key sub-themes: (i) creating principled and robust prediction strategies for multi-view learning, (ii) developing new inference pipelines and prediction analysis frameworks for meta-learning, and (iii) introducing novel inference methods for low-dimensional functionals under transfer learning. In multi-view learning, this project will quantify optimal strategies for cooperative learning, devise new adversarial learning techniques, and analyze the effects of interpolation learning. In meta-learning, this project will introduce new debiasing strategies to tackle inference questions that arise during fine-tuning following an initial phase of pre-training. In transfer learning, this project will develop general-purpose strategies for ranking source distributions and establish new inference schemes for low-dimensional functionals of scientific relevance. On the technical front, this project will introduce novel comparison inequalities, algorithmic proof methods, and leave-one-out techniques that effectively capture the interplay between high dimensionality and heterogeneity. 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.