Carnegie-Mellon University

NSF Awards · FY 2025 · 2025-01

Solid-state single-photon detectors are a class of light sensors that are designed to detect and process light at single photon resolution. These sensors enable a wide variety of modern technologies like quantum computing and integrated biosensor platforms for biomedical analysis. One novel technology trend is to integrate single-photon sensors with microelectronic chips. This integration provides on-board signal processing in addition to ready interfacing with other systems. While there has been much progress in the development of integrated single-photon sensors, there remain quite a few scientific and technological barriers to their widespread adoption. For example, the tradeoffs between noise, temperature, resolution, and photodetection efficiency remain unfavorable for many applications. The proposed research takes a layered approach toward solving some of these outstanding issues, in addition to uncovering new applications in which integrated single photon sensors can provide added benefit or entirely new capabilities. Furthermore, a broader impact of the proposed research program will be a framework that features early educational experiences in optoelectronic science and engineering for middle and high school students. The framework will also include a robust translational research component and an entrepreneurship practicum for both undergraduate and graduate students. The project will feature the development of single-photon avalanche diodes (SPADs) in which the spatiotemporal profile of avalanche currents may be tailored dynamically. This feature, called avalanche confinement, represents a stark departure from current approaches in the development of SPADs; it will yield improved signal-to-noise ratio at room temperature without compromising photon detection efficiency and resolution. The project will be structured in three independent and synergistic research tasks. The first task will focus on establishing a scientific framework that explains avalanche confinement. This will include the development of calibrated device-physics models that leverage quantum kinetic equations to describe the carrier transport processes at play during avalanche confinement. These device-physics models will be used to augment electronic circuit models that can be used in high-level system designs. The first task will include the development of both silicon and silicon carbide SPADs. Under the second task, silicon SPADs will be integrated in standard, i.e., non-specialized, complementary metal-oxide semiconductor (CMOS) microchip technologies to demonstrate a new class of imagers in which pixels operate under avalanche confinement. This integration will enable single photon imagers that can detect light near the quantum limit and at room temperature, i.e., without cooling. Imagers developed under the second task will be evaluated in passive imaging and in fluorescence lifetime imaging for in-incubator biological cell culture analysis. Lastly, under the third task, applications of avalanche-confined SPADs to true random number generation and physically unclonable functions for hardware security will be explored. Together, work under these three tasks will advance the field by establishing a transformative paradigm for the design of single-photon detection and imaging. 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 · 2025-01

PROJECT SUMMARY/ABSTRACT Everyday listening requires auditory selective attention: the ability to direct attention to a single sound source (e.g., your friend talking to you) amongst competing sources (e.g., other patrons at a restaurant). Spatial information is particularly useful for identifying and directing our attention toward a sound source, and the perceptual benefit we receive from the spatial separation of sound sources is called spatial release from masking (SRM). Two pieces of acoustic information are useful in SRM: (1) the difference in the time of arrival of a sound between the two ears, or the interaural time difference (ITD), and (2) the difference in the intensity level of the sound at the two ears, or the interaural level difference (ILD). Normal hearing (NH) listeners rely on both ITD and ILD information to successfully deploy spatial auditory selective attention. However, bilateral cochlear implant (BiCI) users do not have robust access to ITDs, as their devices do not provide temporal information at a fine enough grain. Since ITDs are considered the dominant spatial cue, BiCI users demonstrate poorer spatial hearing outcomes than their normal hearing peers. Instead, they must rely on ILDs. In this project, we propose to remediate spatial hearing in BiCI users through ILD Magnification. This strategy enhances source lateralization by applying larger-than-natural ILDs. ILD Magnification has been shown to improve performance on spatial hearing tasks in both NH listeners listening to vocoded sound and BiCI users, but the neural mechanisms underlying this benefit remain unclear. Here, we will test whether ILD magnification provides a benefit to sound source segregation or to spatial selection. We will measure known neural signatures of auditory attention using functional near-infrared spectroscopy (fNIRS) and electroencephalography (EEG) while subjects complete a spatial attention task. These experiments will be conducted both in NH listeners (Aim 1) and BiCI users (Aim 2). These two aims will affirm the perceptual benefits of ILD magnification and identify the neural mechanisms underlying these benefits. This project offers essential training opportunities for the proposed fellow in his career as an auditory neuroscientist. This includes gaining experience in fNIRS and EEG data collection, as well as proficiency in signal processing and statistical analysis. Additionally, there will be hands-on experience in recruiting and working with cochlear implant patients and exposure to clinical procedures. The training will also encompass professional development, including teaching, mentoring, and science communication. The proposed work will contribute substantially to our knowledge of the relationship between spatial acoustic information and networks of auditory attention. Simultaneously, this project introduces a new method for improving spatial hearing outcomes in bilateral cochlear implant users, working toward the NIH's goal of the application of knowledge to improve health. Overall, the studies described here will improve our understanding of the neural mechanisms underlying spatial hearing and implement a possible solution for those with cochlear implants.

NSF Awards · FY 2025 · 2025-01

Matroids are combinatorial objects that appear in many important areas of mathematics because they capture the notion of “independence” in diverse mathematical constructs, such as graphs, field extensions, hyperplane arrangements, matchings, and discrete optimizations. Using various techniques developed in recent years, the PI aims to deepen the interaction between combinatorics and algebraic geometry, both in matroid theory and in contexts beyond matroids such as Coxeter combinatorics and algebraic statistics. The PI plans to involve undergraduate and graduate students in the project. The PI will work on several projects in matroid theory are proposed using the new framework of "tautological classes of matroids." Many of these projects concern new properties for numerical invariants of matroids that have implications to some long-standing conjectures in matroid theory. These projects also inspire projects in Coxeter combinatorics and algebraic statistics. Completion of these projects will reveal new structural properties of combinatorial objects such as polymatroids and delta-matroids, and will inform the complexity of maximum-likelihood problems in algebraic statistics. 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 · 2025-01

The computing landscape has been gradually shifting from monolithic to distributed systems, catering to, for example, cloud computing and Internet of Things (IoT) applications. This shift challenges the development of such applications because they no longer may act autonomously but have to interact with other, concurrently running components in an orchestrated way, following intended protocols. A compounding factor challenging the understanding of the behaviors of such applications is their heterogeneity: rather than being developed in one common programming language, an array of languages is used, with some components of applications being actual physical objects. The project's novelties are the development of a framework for verifying compliance of these heterogeneous systems with the intended protocols of interactions between them, and the application of the framework to the verification of IoT systems. The project's impacts are (a) foundational reasoning techniques that cater to the heterogeneity of today's systems, allowing not only guarantees of software written in multiple languages, but also of software that interacts with untrusted objects such as sensors; (b) a novel case study of the use of the techniques in verifying a practical system; and (c) the training and development of graduate and undergraduate students. The verification framework uses linear session types to define semantic logical relations for protocol compliance and resource management and explores two scenarios: trusted environments and untrusted environments. The environments convey, respectively, whether an application developer may or may not assume that the foreign objects with which they interact are well-behaved. Possible applications to validate the framework, for both scenarios, include validation of properties such as parametricity and noninterference, as well as IoT applications. While semantic logical relations enjoy popularity overall, only recently have they been introduced to the session-typed setting by the investigator. The project addresses how semantic logical relations express various desirable properties, such as confluence, deadlock freedom, and resource management, which traditionally have been expressed syntactically using types. The project not only intends to scale session type theory to new application domains but also develops techniques that serve as a stepping stone for the development of more general reasoning frameworks such as fully dependent session type theories. 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 · 2024-12

Activities of daily living require a distributed group of brain areas to perform a series of computations that determine and drive action sequences. Compromising even one of these areas (e.g., due to injury or disease) can cause severe behavioral deficits. Thus, many disorders (e.g., Parkinson’s disease) often result in some degradation in planning and/or execution of sequences. Studying the neural mechanism of sequence generation has been challenging due to the distributed nature of the neural operations that drive them. Prior work on primary motor and dorsal premotor cortex (PMd/M1) has established that these areas are critical for sequencing. Yet during motor preparation, PMd/M1 activity reflects only the first movement within a sequence. Information regarding each subsequent movement arrives ‘just in time’ as the previous movement ends. Thus, PMd/M1 must rely on upstream areas to determine which actions should be performed in what order. To assess what these upstream areas are, and through what mechanism they facilitate determination and generation of sequences, I developed a new cognitive-motor behavioral paradigm for Rhesus macaques. My platform requires subjects to deploy previously learned abstract rules to determine a multi-step sequence on never-before-seen trials, while I perform large-scale electrophysiology from a circuit that spans prefrontal cortex, supplementary motor area, premotor cortex, and the basal ganglia. Analysis of these measurements using neural network modeling and population analyses, will help me elucidate the neural mechanisms that 1) allow the brain to determine new movement sequences by deploying prior knowledge, and 2) drive flexible modification of multi- step motor plans. These findings coupled with my flexible behavior and large-scale electrophysiology platform will launch my independent career, where my initial work will focus on 1) discovering the causal link between neural population activity and movement sequencing, and 2) establishing electrophysiological mechanisms of action for commonly prescribed psychoactive drugs. This project will facilitate my training as an independent scientist through new experience with ultra-large- scale muti-region electrophysiology and training in state-of-the-art statistical and neural network modeling methods. This project will involve collaborations between experimental neurophysiologists (Dr. Mark Churchland), theoretical neuroscientists (Dr. Ashok Litwin-Kumar), and physician-scientists (Dr. Mike Shadlen) at Columbia University. This award will help me achieve my long-term career goal to lead an independent neuroscience laboratory that 1) elucidates causal neural mechanisms underlying deductive reasoning and motor control, and 2) uses those insights to test and develop targeted neuromodulation therapeutics for cognitive-motor disorders. The aims of this proposal will also contribute to a core goal of NINDS: acquiring fundamental knowledge about movement control as well as elucidating currently unknown mechanisms of action of prescription drugs used by hundreds of millions of people for cognitive-motor disorders.

NIH Research Projects · FY 2026 · 2024-12

ABSTRACT Exposure to airborne pollutants and harmful chemicals, along with smoking, can lead to a wide range of respiratory diseases, including chronic obstructive pulmonary disease (COPD), bronchiectasis, and asthma; COPD is the third leading cause of death worldwide, with disease counts still on the rise. These diverse respiratory disorders have a pathological hallmark in common: cilia beating defects (ciliopathies), which lead to impaired removal of foreign particulates via the mucociliary escalator, airway obstruction, and increased mortality. Despite the importance, traditional methods for assessing cilia function are equipment-demanding and time- consuming, due to cilia’s nano-scale size and high beating frequency. To overcome this hurdle, this proposal will combine advanced engineering and computational analysis of apical-out airway organoids (AOAOs) to generate physiologically relevant quantitative metrics for modeling mucociliary dysfunction in the human airway. The AOAO exhibits novel behavior that translates the nano-scale cilia beating into micro-scale cilia-powered organoid locomotion, dramatically improving spatiotemporal resolution and enabling cilia functional analysis using computer vision to deliver unprecedented throughput without the need for specialized equipment. Furthermore, the AOAO enables non-invasive pollutant introduction directly to the physiologic, outward-facing apical epithelial surface. The central hypotheses of this project are that the AOAO locomotion correlates with and predicts cilia function and that its apical-out epithelial polarity will allow close mimicry of in vivo injury response dynamics induced by environmental pollution. To test these hypotheses and, thereby, attain the overall objective, the following specific aims will be pursued. Aim 1 will deliver computational tools for rapid ciliopathy diagnosis using point-tracking algorithms to extract AOAO locomotion metrics to correlate with core aspects of cilia function (density, beating frequency, and coordination). Accuracy and accessibility to later users will be further enhanced by utilizing machine learning algorithms to provide automation and high-level feature extraction. Aim 2 will further assess this experimental and computational pipeline for evaluating mucociliary dysfunction by exposing AOAOs to Diesel Particulate Matter (DPM), a model pollutant and major respiratory health threat with close relevance to real-world pollution exposure. AOAOs will be evaluated through computer vision and single-cell transcriptomic analysis to assess the theragnostic utility of the platform for recapitulating native airway-pollutant interactions. The rationale for the proposed research is that a stem cell-based, high-throughput model of respiratory injury will enable accelerated and personalized therapeutic development and clinical management. Concurrent with the pursuit of this research, this project will facilitate the PI’s mastery over organoid engineering and computational analysis.

NSF Awards · FY 2024 · 2024-12

Modern computing systems, especially those supporting machine learning applications, are increasingly burdened by highly variable and unpredictable workloads and service capabilities. For example, machine learning tasks such as generating movie recommendations or interacting with large language models exhibit significant variation in arrival times, user characteristics, and computational demands. Service variability and uncertainty also become pervasive as computing systems scale, due to differences in hardware generations and the dynamic sharing of resources among many applications. These challenges render traditional resource allocation algorithms ineffective as they rely on knowing specific details about job demand and server capabilities. This project aims to advance the fundamental knowledge and to innovate algorithm design for resource allocation in modern computing systems, addressing the challenges posed by high variability and uncertainty in both demand and service. The resulting algorithms aim to improve the performance and energy efficiency of data centers, thereby reducing their carbon footprint while maintaining flexibility, affordability, and scalability in computing services. Additionally, the project will promote interdisciplinary collaboration and educational initiatives, offering mentorship opportunities for underrepresented groups in STEM and contributing to high school and undergraduate curricula in data science and machine learning. The demand and services have become highly heterogeneous and unpredictable in modern computing systems, especially those serving machine learning applications. Such uncertainty poses great challenges to developing performant resource allocation algorithms. Existing performance analyses and algorithms based on traditional methods of queueing and stochastic systems that rely on the knowledge of arrival and service rates are ineffective in these modern ML-workload-dominated computing systems. Although significant progress has been made from the systems perspective to address these challenges, there lacks a theoretical foundation that can offer a deeper understanding of the fundamental limits of system performance and guide the designs of resource allocation algorithms. This project aims to make fundamental advances to queueing and scheduling algorithms for modern computing systems with machine learning workloads. It focuses on two dimensions of uncertainty in computing systems, namely, demand uncertainty (Thrust 1) and service uncertainty (Thrust 2). The proposed research brings two new techniques to the design of computing servers and resource allocation algorithms: 1) coding-theoretic techniques to design more flexible servers to handle multiple job types and heterogeneous demands without having to overprovision resources, and 2) online-learning-based job scheduling strategies that dynamically estimate service capabilities. The research exploit the cross-pollination of ideas between the queueing theory, information/coding theory and online learning. 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 2024 · 2024-10

Retrieval-Enhanced Machine Learning (REML) refers to a subset of machine learning models that make predictions by utilizing the results of one or more retrieval models from collections of documents. REML has recently attracted considerable attention due to its wide range of applications, including knowledge grounding for question answering and improving generalization in large language models. However, REML has mainly been studied from a machine learning perspective, without focusing on the retrieval aspects. Preliminary explorations have demonstrated the importance of retrieval on downstream REML performance. This observation has motivated this project in order to provide an alternative view to REML and study REML from an information retrieval (IR) perspective. In this perspective, the retrieval component in REML is framed as a search engine capable of supporting multiple, independent predictive models, as opposed to a single predictive model as is the case in the majority of existing work. This project consists of three major research thrusts. First, the project will develop novel architectures and optimization solutions that provide information access to multiple machine learning models conducting a wide variety of tasks. Next, the project will study training and inference efficiency in the context of REML by focusing on the utilization of retrieval results by downstream machine learning models and the feedback they provide. Third, the project will study approaches for responsible REML by examining data control for content providers in REML and fairness and robustness across multiple downstream models. Without loss of generality, the project will primarily focus on a number of real-world language tasks, such as open-domain question answering, fact verification, and open-domain dialogue 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 2024 · 2024-10

This Civic Innovation Challenge (CIVIC) Stage 1 project will support research that intends to develop and provide a roadmap for revitalizing legacy rail lines—which are abandoned, out-of-service, or underutilized railroad tracks. There are thousands of miles of existing but underused legacy lines in urban and rural communities across the country. This existing infrastructure is poised to improve accessibility by better connecting underserved communities to essential services. The primary barrier to using legacy rail lines is the lack of affordable rail-defect detection systems, which are important safety mechanisms. This research project aims to overcome that barrier. The project team — academic, civic, community, and industry partners — will use Philadelphia’s Delaware River waterfront as a demonstration site using a low-cost, artificial intelligence-based technology previously developed by the team. This technology, which is installed onboard trains, detects broken rails and other track damage in near real-time, facilitating a transition of a legacy line to active passenger rail in months rather than years. Research completed in association with project is the culmination of a decade-long process of public engagement and research, identifying rail as the preferred choice for improving accessibility because it better connects with other transit modes, supports high-capacity needs, stimulates local economies, adapts to increasing demand, offers more reliable service, and produces fewer emissions per passenger than cars and buses. By repurposing dormant rail assets, the project aims to enhance connectivity, reduce traffic congestion, promote environmentally friendly transportation, revitalize communities, and improve living standards in both urban and rural settings. To achieve these goals, the research will strive to reintroduce affordable, dynamic, and community-driven legacy rail systems by transitioning new, safe technologies into an industry traditionally adverse to rapid change. The onboard monitoring and detection technology deployed and tested through this project combines two complementary sensing modalities: real-time acceleration data to inform statistical anomaly detection algorithms and real-time automated computer vision for detecting broken rails and classifying track damage. This adaptable system provides actionable data on integrity and ridership, reducing the time and financial barriers compared to traditional rail upgrades, and facilitating quicker regulatory approval and community acceptance. In Stage 1, this research project will: (1) integrate this technology within passenger rail intending to demonstrate that it is adaptable to various types of trains and locations, (2) assess the economic impacts of revitalized legacy rail, (3) gain community feedback on implementation needs, and (4) address the regulatory, financial, and governance implications of deploying advanced technologies for revitalizing legacy rail infrastructure. This project is in response to the Civic Innovation Challenge program’s Track B. Bridging the gap between essential resources and services & community needs and is a collaboration between NSF, the Department of Homeland Security, and the Department of Energy. 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 2024 · 2024-10

U.S. policy is spurring a range of ambitious public and private investments that seek to improve national industrial capacity, economic security, and competitiveness – large-scale capacity investments will create thousands of new jobs, often in occupations without a significant base of current employment in the regions where investments are made. Meeting this new demand for skills will require policymakers, employers and trainers to identify other occupations that partially meet new job requirements, and to quantify what skills may be needed for workers to transition into new opportunities. This project will develop and improve methods to help evaluate the potential readiness of regional workforces to meet the skill demand created by large-scale industry transitions. This project will also help produce capabilities that can provide insight into possible transition opportunities for workers whose employment may be disrupted by technological and economic transformation. Firms, trainers, government, labor groups and other key decision-makers lack consistent, data-driven methods for evaluating workforce feasibility. Rather than a one-time study for a specific project or technology, a flexible and repeatable capability is needed for decision-support across a range of industrial scenarios, to identify for any given investment proposal the conditions under which that proposal may be feasible from a workforce standpoint, and to support the development of a data-driven strategy for meeting workforce needs (such as identifying skill gaps to be closed through training programs). This project will leverage an approach to estimating the similarity of requirements between different occupations, as well as other potential indicators of the feasibility of worker transitions from one occupation into another. These indicators will be tested against longitudinal evidence of realized occupational transitions and used to specify models that quantify the number of workers who may satisfy a minimum level of readiness for demand in any given occupation, and boundary estimates on the rate at which such workers might transition into the in-demand occupation. The outputs of this project will include an empirically validated user-tool that helps estimate potential stock and mobility of workers in every US metropolitan statistical area to meet demand for any occupation within the BLS SOC taxonomy, and the potential gaps between "candidate" occupations and the in-demand occupation. Underlying analytical models of the tool and its performance against historic mobility trends will be published. This tool will be applied to support a regional case analysis of workforce readiness for capacity-building in microelectronics in Florida and energy storage in New York. 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 2024 · 2024-10

Environmental sustainability has become a critical global concern in both technological and economic spheres. Artificial Intelligence (AI), while addressing the environmental consequences of human activities, also contributes to environmental degradation due to its reliance on energy-intensive hardware like GPUs, which are used in diverse applications, ranging from chatbots (e.g., ChatGPT) to AI for Science applications (e.g., AlphaFold3). The project’s novelties are the introduction of a holistic approach combining system architecture, machine learning (ML), and software engineering to boost energy-efficiency, sustainable, and collaborative AI development. The project's broader significance and importance are its offering (i) easy-to-use, low-cost, scalable, and standardized open-source software to promote environmental sustainability in AI development; and (ii) multi-disciplinary education and research training opportunities across various educational levels. This project takes a holistic view of all levels of the AI development pipeline and seeks to harness recent progress in ML Systems, AutoML, and Collaborative Learning (CL) techniques to foster an environmentally sustainable approach to AI and LPM development across natural language processing, computer vision, and AI4Science domains. The project’s goal is to enhance training efficiency, reduce power consumption, improve hardware utilization, and expedite model development explorations. The project introduces a suite of measures, including efficient parallelization strategies, well-tuned hyperparameters with minimal trial rounds, and optimized hardware utilization. By aligning all aspects of the computing process with sustainability, a comprehensive approach to AI computing can be adopted, fostering a more responsible and environmentally mindful practice. The project’s objectives include (i) sustainable hardware utilization through optimized distributed ML strategies; (ii) efficient collaboration via cognitive scheduling and CL; (iii) learning to develop AI models and hyperparameters faster; and (iv) developing software toolkits for sustainable ML at all levels through multi-level optimization. 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 2024 · 2024-10

This project will explore new ways to address complex mathematical problems by integrating advanced machine learning techniques with automated reasoning. By combining artificial intelligence with formal mathematical methods, the research team will advance the knowledge of some long-standing open problems in mathematics and computer science. These disciplines are crucial for the development of technologies that ensure software reliability, security, and efficiency --- key aspects in the digital age. The project not only supports the exploration of theoretical knowledge but also the practical application of these new algorithms to improve the tools that are integral to the technological infrastructure. The project will support two students, a mathematician and a computer scientist, who will closely work together to achieve the proposed goals. In technical terms, the project will use three specific open problems within graph theory and combinatorics as test cases to evaluate the effectiveness of new algorithms. The first objective will involve applying machine learning to develop efficient symmetry-breaking clauses to determine the values of small Ramsey numbers. Secondly, transformer-based methods will be used to generate small Folkman graphs. Lastly, the project will tackle a realizability problem related to point sets in a plane, aiming to understand and create configurations with larger than previously known planar discrepancies. This project will be a collaboration between mathematicians and computer scientists aiming to explore the synergy between machine learning and SAT solvers. The research team will improve methods for addressing these difficult problems, potentially obtaining both theoretical insights and novel computational techniques in mathematical and computational sciences. 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 2024 · 2024-10

In recent years, artificial intelligence (AI) has made major progress. Generative AI and, in particular, large language models (LLMs) have revolutionized how we think of AI. These methods are used broadly and are now also applied to program development. Before LLMs, human contributors, commenters and curators would ensure at least some minimum level of correctness when looking up such code fragments on programming-related internet sites. These guard rails are gone when using LLMs, as the system cannot provide any guarantees and insights to ensure the correctness of the produced examples. This LLM hallucination becomes a huge issue when LLMs are asked to provide examples for algorithms with complex and hard-to-understand behavior and where correctness is not obvious. In the worst-case scenario, generative AI provides incorrect and outdated code snippets from undocumented sources that are strung together into a “reasonably looking” example that may well be wrong in a subtle way, but only experts could spot the problem. Any effort that helps guard AI from acting outside its bounds and that ensures that one can trust in AI has a major scientific and societal impact. This project focuses on enabling trust when using AI as an aid to programmers. This project develops an experimental approach to the above-described problem for a class of important core numerical algorithms. The team of researchers takes their inspiration from a well-understood insight in mathematics and computer science that proving a solution correct is easier than finding a solution. In the context of LLMs, they investigate how to utilize LLMs to guess an implementation of an algorithm, within clear implementation constraints. Then, they develop an extension to the SPIRAL system (www.spiral.net) that implements symbolic execution and theorem-proving capabilities to derive the semantics of the LLM-generated code using SPIRAL’s formal framework and engine. For the problems addressed by this project, the researchers aim to demonstrate how rule-based symbolic AI can bring guarantees to generative AI. While the problem of deriving the formal semantics of a piece of numerical code, in general, may well be unsolvable, for the class of algorithms that the SPIRAL system understands, the problem becomes tractable. The approach is a variant of lifting that leverages the detailed knowledge of numerical algorithms encoded in the rule-based AI components of the SPIRAL system. 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 2024 · 2024-10

Although healthcare Artificial Intelligence (AI) research has made notable progress, its integration into healthcare practice remains limited. Clinicians typically make decisions by reviewing electronic health records that include clinical notes, medical images, and lab results. However, most current healthcare AI tools are incapable to performing multimodal analysis. In addition, existing prediction models are inflexible and lack the interactive capabilities to address the complex and dynamic requirements of clinicians. A systematic and comprehensive evaluation framework is also missing in these healthcare AI tools. In this project, the team aims to develop an end-to-end AI system to overcome these limitations. The proposed framework brings a novel AI approach to multimodal data by training foundation models to simultaneously combine and analyze four data modalities. These modalities include: data over time (i.e., time-series), text from notes, image, and numerical data. The system then provides comprehensive evaluation and tailored feedback. The resulting technologies will facilitate seamless integration of AI systems into real-world clinical settings. This project pursues three closely connected research thrusts: modeling multimodal clinical data, enhancing interactions between clinicians and AI, and enabling comprehensive evaluations of healthcare AI systems. Specifically, the team will develop new process for training foundation models to enable joint consideration of four data modalities including time-series, text, image, and tabular data. Clinicians will be able to interact with these models in natural language, and in return these models will generate contextualized clinical text. We will enhance usability, trustworthiness, and personalization by introducing three key features: (1) clinician intent elicitation, (2) trustworthy content generation grounded in multimodal input data, and (3) clinician-oriented computationally efficient content personalization. To make sure that the models can be effectively deployed in everyday routines of busy medical practitioners, we will develop standardized evaluation metrics in collaboration with clinicians, including a wide array of objective and subjective measures such as predictive accuracy, bias, toxicity, fairness, and robustness, and preferences of the clinicians. The outcomes will include innovative methodological contributions to AI, as well as datasets and benchmarks for evaluate these systems. Additionally, the project will develop key capabilities for enhancing patient safety during the high-risk transfer of care, and will make broader scientific and engineering contributions to computer science, information sciences, and statistics. This project’s findings will be promptly shared with the public and industrial partners to maximize the impact and facilitate the transition to practical applications. 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 2024 · 2024-10

The transportation industry has seen rapid technological change over the past decade, much of which has been enabled by the advancement and integration of artificial intelligence (AI) — from route planning and dispatch to autonomous vehicle technology. This project builds on the team’s in-depth research across roadway transportation modes (taxi, transit, and logistics) to inform a set of responsible AI practices for the sector. Specifically, the research focuses on the development of a framework and set of tools for participatory labor impact assessment — or, comprehensive analysis of how the introduction or expansion of AI technologies will affect workers within the industry from the perspective of those set to be most impacted (namely workers, but also bus and taxi passengers, residents of neighborhoods surrounding logistics clusters, and other community stakeholders). Over the two-year planning grant, the multi-disciplinary team, spanning the fields of human-computer interaction, social informatics, and sociology of work, will convene taxi, transit, and logistics workers and their representatives to assess AI awareness, on-the-ground needs and to specify participatory labor impact assessment tools. This process will result in the following contributions: (1) a comparative analysis of how AI is used and regulated across different transportation modes; (2) a participatory framework for evaluating existing AI systems and identifying opportunities to enhance safety and equity, and finally; (3) a transferable set of design principles for responsible AI development and implementation that considers the needs and concerns of workers. This project will provide valuable insights into the diverse application of AI within transportation, and its significant effects on the industry’s workforce and, in turn, on other stakeholders such as customers and service recipients. It will propose novel methods for assessing labor and labor-driven impacts and a set of responsible AI design principles, offering a foundation for future research and policy 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 2024 · 2024-10

The Workshop on Operational Factors for AI Diffusion and Productivity Growth will convene leading experts from academia, industry, policy and other sectors to discuss the economic impacts of AI technologies on domestic and international workforces. Given the significant investments being made in AI in both the private and public sectors, this workshop will not only provide a timely forum for a discussion on the topic but will also help seed new interdisciplinary and multi-sector collaborations to contribute new perspectives and enhanced dialogue on the diffusion of AI technologies and their economic impact. The workshop will help generate further avenues of study to understand these impacts. While there is currently much discussion about the effect that AI diffusion will have on the economy, the actual extent of its potential real impact is yet unknown. Large upfront costs mean that only large firms have the scale necessary to justify adoption as most firms find the technical feasibility far exceeds economic feasibility. The direction of organizational change will also determine whether gains in AI diffusion accrue in ways that tend to replace or augment workers. The meeting will bring together industry practitioners, federal policy leaders and researchers in AI and economics to discuss the impact of AI on the workplace, productivity, and pay equity. Topics to be explored include the impact of AI adoption on productivity, the potential efficiency gains from this technology’s deployment, the reorganization of production and changing the composition of tasks within the firm, and the generation or creation of new tasks and work products. The workshop will facilitate in-depth discussions and connect experts to find avenues ripe for impact and further study. The workshop will feature several prominent researchers in the field as invited speakers. Results from the workshop will be widely circulated via a workshop report which will include findings and future directions for impact and further study. 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 2024 · 2024-10

Modern software applications in all facets of society, including commercial, scientific, and non-profit enterprises, rely on databases to store information. These organizations often want to use their data in ways they cannot easily express with existing database query languages, especially in the context of artificial intelligence and data science applications. This mismatch means such applications wait longer for answers about their data, inhibiting them from reacting to changes as quickly as possible and impeding their goals. This research addresses this problem and develops foundational techniques that automatically removes such inefficiencies without requiring organizations to perform costly rewrites of their application code. It enables organizations to ask more complex questions about their data and extrapolate new knowledge from it, all while using less computing and energy resources than today’s systems. Many database management systems (DBMSs) extend the query language SQL to support user-defined functions (UDFs) written in procedural programming languages. Despite their software engineering advantages, UDFs are notoriously difficult to optimize within database systems, and DBMSs often resort to executing them iteratively (row-by-row). This project focuses on developing optimization approaches to overcome SQL and UDF boundaries via automatic code transformations that pass critical information between them to enable more effective query planning and compilation. These strategies include methods for programmatically deconstructing UDFs into smaller pieces, manipulating them individually, and reconstructing them into the calling query to optimize performance. This project will address three fundamental research challenges: (1) improving the performance of UDFs without requiring modifications to the application code, (2) optimizing external language UDFs (e.g., Python) that rely on dynamic types and library calls, and (3) generating new optimizations that leverage information about UDFs across the entire lifecycle of a query and multiple invocations. By eliminating performance penalties associated with UDFs, this research will enable organizations to improve the efficiency of applications and support more complex workloads, including leveraging machine learning and data science libraries. 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 2024 · 2024-10

Decision-making in many fields involves managing complex systems made up of smaller, interconnected parts. These systems can be modeled using weakly coupled Markov decision processes (WCMDPs), which are groups of smaller Markov decision processes (MDPs) linked by shared constraints. WCMDPs are applicable in various fields such as job scheduling, resource allocation, electric vehicle charging, and supply chain management. However, despite their widespread application, many fundamental questions on WCMDPs remain unanswered. Efficiently computing near-optimal decision rules, i.e., policies, for WCMDPs is still an open problem. Furthermore, when the problem parameters are unknown, reinforcement learning (RL) approaches are needed, but effective RL algorithms for WCMDPs are currently lacking. A key challenge is that the shared constraints create coupling among the smaller MDPs, which prevents making decisions for each MDP individually and thus leads to hardness results when the number of MDPs is large. This proposal aims to establish a theoretical foundation and innovate algorithm designs for WCMDPs. The proposed research will develop theory and techniques to “decouple” large WCMDPs into their smaller parts and then “reassemble” them properly. This research will draw on a new approach devised in the preliminary work, named the “one-to-many” approach, for tackling decision-making in large, complex stochastic systems. This new approach will be combined with classical techniques from large stochastic systems, including the Lyapunov drift method, Stein’s method, and rate conservation law, as well as recent advances in reinforcement learning. The algorithms and theory developed in the above research will be evaluated in both simulated problems and in the resource management problem in large-scale computing systems, using real-world data traces from Google’s datacenters. The results from this project are expected to enrich the traditional algorithms and theory not only for WCMDPs but also for large-scale MDPs in general. This research will be accompanied by curriculum development, mentoring programs, and initiatives at conferences designed to recruit students from underrepresented backgrounds into research on decision-making in large stochastic 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 2024 · 2024-10

Understanding and predicting earthquakes is a critical endeavor that has profound implications for public safety and disaster preparedness. By developing cutting-edge machine learning models and algorithms, this project seeks to uncover the intricate dynamics of earthquakes, potentially identifying precursory signals that precede major seismic events. The broader impact of this work includes enhancing our ability to forecast earthquakes more accurately, thus mitigating risks and improving resilience in communities prone to seismic activities. Additionally, the project will create open-source tools accessible to researchers and practitioners worldwide, fostering global collaboration and knowledge sharing. This initiative not only aims to advance scientific understanding but also to inspire and educate the next generation of geoscientists through engaging sonification/animation products and educational programs. This research addresses the challenge of understanding earthquake physics and forecasting its collective behaviors by utilizing high-resolution, high-dimensional earthquake catalogs and continuous geophysical measurements. Traditional forecasting models struggle with the complexity of such detailed data; thus, this project proposes novel approaches grounded in marked temporal point processes. The key strategies include developing advanced Hawkes process models with deep neural triggering kernels to gain nuanced insights into earthquake dynamics, creating a novel generative framework to explore complex seismic patterns, and applying these methods to recent earthquake sequences in California, Japan, and Türkiye. The project will produce open-source software tools to support these efforts. The intellectual merit lies in integrating advanced statistical models, machine learning techniques, and high-resolution earthquake catalogs to address longstanding challenges in geoscience. By enhancing the representation of earthquake dynamics with deep neural triggering kernels within Hawkes process models, the project aims to overcome limitations of traditional forecasting methods. The generative framework for marked temporal point processes will enable systematic exploration of intricate seismic patterns. International collaborations and the development of accessible, open-source resources exemplify a commitment to impactful and practical research. Additionally, the project will offer Research Experiences for Undergraduates (REU) at Georgia Tech, promoting interdisciplinary collaboration and broadening participation in geosciences. The collaboration between an early career machine learning PI and a mid-career earthquake seismologist further underscores the project's innovative and interdisciplinary nature. 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 2024 · 2024-10

Silicon is the cornerstone material for building microelectronics, essential in a wide range of devices from smartphones and personal computers to electric vehicles. However, its inherent material limitations also pose challenges for advancing future computational technologies. As the thickness of silicon decreases to sub-3-nm range, its carriers suffer from significant scattering, leading to dramatic performance degradation. This limitation results in a “silicon-impossible” territory. According to the International Roadmap for Devices and Systems (IRDS), further scaling of silicon technology nodes will reach a plateau at physical channel lengths of 12 nm by 2037. In contrast, the atomically thin body thickness of two-dimensional (2D) semiconductors offers superior immunity to aggressive scaling, presenting a distinct advantage for advancing transistor technology. This program aims to experimentally demonstrate wafer-scale 2D transistors in the “silicon-impossible” territory (sub-5-nm channel length) and investigate their fundamental limits through a combination of experimental and theoretical efforts. Key metrics of the extremely scaled 2D transistors will be benchmarked with the IRDS projections for both high performance and low power applications, as well as the state-of-the-art industrial technology nodes. This program will generate critical knowledge and technologies for next generation of energy efficient computing and a roadmap of further optimization of device structures and material designs. Additionally, this program will provide training opportunities for the future workforce in the semiconductor industry, covering K-12, undergraduate and graduate students. The evolution of silicon complementary metal-oxide-semiconductor (CMOS) transistor technology, driven by Moore’s law scaling, has led to impressive advancements in computing, communications, robotics, and healthcare. To keep up with the shrinking lateral dimensions of transistors, their vertical dimensions must also be reduced in order to prevent short channel effects. Nonetheless, as silicon thickness approaches sub-3-nm scales, the presence of dangling bonds leads to substantial scattering of charge carriers and degrades the carrier mobility in silicon. Therefore, there remains a challenging "silicon-impossible" zone, defined by parameters including a channel body thickness of less than 3 nm and a channel length of less than 5 nm. The atomic thickness of 2D semiconductors, particularly 2D transition metal dichalcogenides, makes them highly suitable for ultimate scaling of transistor technology. Unlike silicon, 2D semiconductors benefit from their van der Waals bonding, which eliminates dangling bonds and keeps carrier mobility immune to thickness scaling. This makes them a promising alternative for advancing Moore’s law into the “silicon-impossible” domain, offering excellent subthreshold performance and ultra-low power consumption. The proposed program aims to use a combined experimental and theoretical approach to systematically investigate the fundamental limits of 2D transistors in the “silicon-impossible” territory, with the following activities: (1) developing a high-throughput, scalable technique for fabricating wafer-scale 2D transistor arrays with physical channel lengths below 5 nm; (2) creating a quantum transport simulator to elucidate the key device physics affecting sub-5-nm 2D transistors and to outline a strategy for further optimization of device designs; and (3) establishing an open-access database to systematically catalog the subthreshold and on-state properties of highly scaled 2D transistors. This program will broadly impact multiple disciplines including electrical engineering, physics and materials science and offer unique outreach and educational opportunities to undergraduate and graduate students and K-12 students. 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 2024 · 2024-10

Public and community mental health services, which provide essential support and treatment for individuals with behavioral health conditions, have a profound impact on communities and millions of individuals across the United States. Un- or under-treated, significant mental illnesses are among the biggest sources of years lost to disability and economic burden in the United States. Communities served by public and community mental health services represent some of the most vulnerable in our society and can benefit greatly from the development of data-driven intelligent technologies. This planning grant focuses on strengthening and empowering peer-led organizations in public and community mental health services. As the behavioral health workforce crisis deepens, there is a growing imperative to expand and enhance the role of peer providers in public behavioral health, particularly within underfunded and technologically under-resourced peer-led agencies. These organizations, vital for delivering a diverse range of support to address individual holistic needs, including social, physical, emotional, and environmental by offering individual peer support, self-help groups, education and training, navigation to legal advocacy, face significant challenges due to inadequate technological infrastructure and reliance on low-tech service delivery methods. This planning grant builds on our ongoing partnership with Collaborative Support Programs of New Jersey (CSPNJ), a state-wide, peer-led community-based behavioral health agency, which is known for its innovative work serving people with complex behavioral health, social and economic challenges. We will work together to create community-centered AI solutions that meet the needs of both peer service providers and recipients in resource-constrained peer-led public mental health services. We aim to build a collective understanding of community needs, success criteria, and challenges, and facilitate rapid prototyping of proven AI solutions through participatory methods. Leveraging multidisciplinary expertise in Human-computer Interaction (HCI), Artificial Intelligence (AI), mental health, and social work, and in close collaboration with our community partners, this planning grant will help us gain a robust understanding of (1) how service providers and recipients perceive success in peer-run community mental health services and the challenges they face in achieving such success. Based on these insights, we will (2) collaboratively identify specific AI-driven technologies that could be piloted in Stage 2 to enhance the capabilities of safety-net peer-run organizations. This will empower them to better meet the needs of underserved communities and potentially serve a broader population, thereby strengthening their impact in the community mental health sector. This project is in response to the Civic Innovation Challenge program’s Track B. Bridging the gap between essential resources and services & community needs and is a collaboration between NSF, the Department of Homeland Security, and the Department of Energy. 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 · 2024-09

Models of alcohol use disorder (AUD) risk increasingly conceptualize relationship factors as being critical to the understanding of problem drinking. A substantial literature exists linking lower relationship quality (e.g., partner conflict, relationship dissatisfaction, insecure attachment) to the development of AUD, and interventions focused on improving relationship quality have demonstrated some success. However, we lack an understanding of how and why lower relationship quality shapes AUD risk over time. A better understanding of such mechanisms could have broad implications for understanding AUD risk in close relationships more generally. The proposed project merges insights from the alcohol administration literature and research conducted by social psychologists with expertise in close relationships to offer a novel conceptual model that depicts how lower relationship quality impacts the way that couples experience alcohol, which in turn sets them on a path toward heavier alcohol consumption and the development of AUD symptoms. More specifically, we predict that couples with lower (vs. higher) relationship quality experience heightened alcohol-induced social bonding and emotional rewards, and that this increased sensitivity to alcohol’s social and emotional rewards in couples with lower relationship quality is a key mechanism that drives problematic drinking over time, which then increases risk for AUD. Two-hundred-fifty-two couples (N=504 young adult drinkers; aged 21-30) will drink together over 36-min a moderate dose of alcohol (males: 0.82 g/kg; females: 0.74 g/kg) or a placebo beverage. Alcohol reinforcement (e.g., social bonding, reduced social tension) will be assessed during a free interaction period and a conflict resolution discussion using a broad range of measures across multiple response systems (e.g., self-reports, observational measures). Drinking behavior, alcohol reinforcement, acute alcohol-related problems, and relationship functioning will be assessed in daily life during three subsequent ecological momentary assessment bursts, and AUD symptoms will be assessed at baseline and at 12-month follow-up. We predict 1) an alcohol by relationship quality interaction, such that alcohol’s social and emotional reinforcing effects at the individual and dyad-levels, measured across multiple response systems, will be stronger in lower (vs. higher) RQ couples, 2) alcohol’s social and emotional reinforcing effects in real-world contexts will vary across couples as a function of the blood alcohol curve (BAC) and partner drinking status, such that couples with lower (vs. higher) RQ will experience stronger reinforcing effects on the ascending limb of the BAC when both members are drinking together (vs. when couples are together but only one member is drinking), 3) higher alcohol reinforcement on the ascending limb of the BAC will predict heavier alcohol consumption and more acute alcohol problems (both within-person and across-partners), and 4) lab-based and real-world measures of alcohol reinforcement will prospectively predict heavier drinking and more AUD symptoms at 12-month follow-up, particularly for lower RQ couples. This project integrates two prominent literatures that have not been connected before (i.e., social psychological theory and research in couples with longitudinal alcohol administration work) and promises to have broad conceptual, methodological, and clinical implications..

NSF Awards · FY 2024 · 2024-09

The World Health Organization defines self-care as "the ability of individuals, families, and communities to promote health, prevent disease, maintain health, and to cope with illness and disability with or without the support of a healthcare provider". Individuals recovering from medical treatments tend to perform self-care alone and post-operative patients are often expected to maintain a balance between rest and physical activity. While doctors often want the patients to start moving as soon as possible, moving too much or too intensely can increase recovery time and risk injury. The goal of this project is to guide patients through the recovery process using lightweight sensing and behavior modeling using situational and context-aware voice-based guidance. The team will develop machine learning methods to optimize outcomes and help patients manage their condition. The project will unify multimodal data on activity and behavioral sensing on a watch with an artificial intelligence (AI)-driven intervention agent. This should advance aspects of at-home healthcare for an underserved portion of the population and, more generally, contribute to basic science in adaptive, real-time, in situ multimodal interactions. The team will apply advances in basic science to post-operative care procedures for surgical excision for skin cancers on the head and neck. In this project, the team proposes to solve these challenges using situated and context-aware voice-based guidance. The system will use watch-based multimodal sensing to guide the patient through three care procedures with technical demands of increasing complexity: maintaining appropriate activity levels, pain management, and wound care. The agent will recognize the patient’s actions and behaviors, and pro-actively intervene only as needed. It will augment doctors' understanding of patients' situations with a meaningful and appropriate level of explanation when a problem occurs. Where there is repetition, it will adapt to users’ changing needs as they develop familiarity with the associated procedure. The proposed work will contribute to science through (i) the development of new machine learning models that use sparse, multimodal, in situ training data to identify user actions, motion and body pose; (ii) development of generalizable physical behavior modeling; (iii) advances in knowledge regarding how state-of-the-art sensing and natural language processing combine to better support at-home patient recovery; and (iv) identifying effective strategies to determine the level of support and intervention the user needs during post-operative recovery and adapting system performance accordingly. 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 2024 · 2024-09

Strengthening American Infrastructure (SAI) is an NSF Program seeking to stimulate human-centered fundamental and potentially transformative research that strengthens America’s infrastructure. Effective infrastructure provides a strong foundation for socioeconomic vitality and broad quality of life improvement. Strong, reliable, and effective infrastructure spurs private-sector innovation, grows the economy, creates jobs, makes public-sector service provision more efficient, strengthens communities, promotes equal opportunity, protects the natural environment, enhances national security, and fuels American leadership. To achieve these goals requires expertise from across the science and engineering disciplines. SAI focuses on how knowledge of human reasoning and decision-making, governance, and social and cultural processes enables the building and maintenance of effective infrastructure that improves lives and society and builds on advances in technology and engineering. Facilities operations encompass a wide range of essential services, such as ensuring occupant thermal comfort, energy management, and overseeing space design and management. Public and commercial enterprises, such as healthcare facilities, schools, and community centers, must confront the rising costs of operating their facilities every day. For such enterprises, the most significant and costly operations inefficiencies arise from an inability to accurately predict facilities usage, which can be highly dynamic and uncertain. This project pioneers a new approach to modeling and predicting human-infrastructure interactions, behavior, and decision making within facilities, to address the inefficiencies that often arise from these factors. By developing a Human-centric Intelligent Facilities Integration framework that combines real-time sensing, computational modeling, and automated decision making tools, facilities management will be moved from a conventional, operational focus to a dynamic, human-centered approach. Using this framework, facilities will be able to proactively adapt to and anticipate the needs and behaviors of their users, fostering more inclusive, sustainable, and efficient environments. These models will empower cities to better prioritize investments and target specific outcomes, as facilities are one of the most important elements needed to start and maintain transformation of deindustrialized centers. Despite their integral role, there is limited understanding of how human interactions impact facilities and how this knowledge could enhance efficiency, utilization, and resource distribution. Progress requires creating interdisciplinary opportunities for computational behavioral research to integrate with engineering for measurement, prediction, and automation. This project’s Human-centric Intelligent Facilities Integration framework addresses this challenge by bridging three fields: behavioral and cognitive science, cyber-physical systems, and design. This research develops new cognitive models to represent psychological decision making in the context of facility usage, focusing on how individuals interact with and navigate through physical and social spaces. These models are contextualized by dynamic instances and features not previously studied in highly dynamic facility environments. New privacy-preserving technologies are developed to map spatio-temporal social interaction data into existing facility management digital infrastructure. Finally, a novel hierarchical, learning-based model fusing cognitive, social, and spatial layers is developed for predictive modeling of human adaptations to changes in facilities management. The project also generates valuable datasets that document human-infrastructure interactions in facilities, allowing other researchers to further advance understanding. Additionally, the project provides open-source tools and implementations of various framework components to facilitate replication and collaboration within the scientific community. 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 · 2024-09

PROJECT SUMMARY The formation of biomolecular condensates (BMCs) is an important phenomenon in biology, creating transient compartments that can accelerate biochemical reactions or sequester molecules during times of stress. Aberrations in BMC formation are implicated in a growing number of diseases. Thus, molecular tools that enable researchers to monitor and manipulate BMCs are in great demand. Ideally, these tools could be deployed without engineering the biomolecules that drive BMC formation in order to avoid artifacts introduced by altering the sequence of the RNA or protein components. Our proposal is focused on developing RNA-targeting probes that can be used to (a) fluorescently label RNAs suspected of participating in BMCs and (b) disrupt specific RNAs from entering BMCs. The molecular probes we will use for these experiments are based on peptide nucleic acid (PNA), a synthetic version of DNA in which the natural sugar-phosphodiester backbone is replaced by an extended peptide. Hybridization of the PNA to its RNA target will be used to introduce a fluorescent dye for visualization in microscopy. Moreover, PNA hybridization should prevent that region of the RNA from either binding to a protein or interacting with other RNAs, blocking its incorporation into a BMC. The advantage of this approach is that the RNA target need not be modified in any way, i.e. we will target endogenous RNAs. Additionally, the PNA can be introduced in a reversible manner, meaning it can be removed from the RNA at any time, releasing the RNA to participate in BMC formation when the researcher deems it appropriate. Our model system will be the C9orf72 gene bearing expanded repeats having the sequence G4C2/C4G2. The resulting RNAs, both of which are present due to bidirectional transcription, have been implicated in numerous toxic gain-of-function phenomena that are central to the pathology of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). PNAs that can block the entry of these RNAs into BMCs can help assign roles to specific molecular components of these complex, multicomponent condensates. Moreover, since numerous neurodegenerative diseases feature expanded repeats of other sequences, the tools we develop in this project can be readily adapted to other targets and diseases, greatly enhancing the potential impact of our proposed research. Finally, while antisense approaches targeting the C9orf72 RNA have not yet led to viable therapies, it is possible that the high affinity of PNA and a focus on targeting both of the expanded repeat transcripts will offer better outcomes.