University Of Southern California

NSF Awards · FY 2025 · 2025-01

Understanding how climate has changed in the past (the study of paleoclimatology) can help societies better adapt to and anticipate future climate change. To fulfill this promise, researchers need to have rapid access to the abundance of datasets and analytical methods available to them. Furthermore, they may need assistance in choosing the most appropriate technique(s) for their data, applying them correctly, and interpreting conflicting or ambiguous results. Artificial intelligence (AI), and in particular large language models (LLMs), can help in this regard. Indeed, they have already proven useful as coding assistants. This project develops an artificial intelligent system that uses the power of generative AI while incorporating paleoclimate knowledge. The resulting system, PaleoPAL, is used in the context of three paleoclimate studies to evaluate its effectiveness. In addition, the project engages with publishers to provide guidelines for study incorporating AI assistants in the research. PaleoPAL uses Retrieval Augmented Generation (RAG) to incorporate paleoclimate knowledge (e.g., data, software, methods, workflows, literature) into existing LLMs to create an AI assistant as a Jupyter Notebook, an environment familiar to scientists. This AI assistant helps in the investigation of three paleoclimate problems: placing recent El Niño-Southern Oscillation variations in the context of the last 10,000 years, detecting climate tipping points and their potential precursors, and generating empirically-based, low-cost climate projections. The proposal supports training activities that build capacity in the US workforce, and in particular, teaching a diverse cross-section of the next generation of geoscientists to work with AI assistants, learning from them and challenging them as they would a mentor. This award by the Division of Research, Innovation, Synergies, and Education within the Directorate for Geosciences is jointly supported by the National Discovery Cloud for Climate initiative within the Office of Advanced Cyberinfrastructure within the Directorate for Computer and Information Science and Engineering. 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

ABSTRACT - More than 1 in 9 seniors in the United States is living with Alzheimer’s disease (AD), and the number of people affected is expected to double over the next two decades, rising to 13 million in 2050 (Alzheimer’s Association, 2022). The recent shift to creating a biological definition of Alzheimer’s disease for research purposes- advocated by the NIA and the U.S. Alzheimer’s Association task force, has led to great interest in neuroimaging and other biomarkers of brain health from MRI and PET. While emerging treatments may resist cognitive decline by 25-37%, there are risks for severe side effects. Efforts to identify precise markers of disease progression in the living brain as early as possible are essential. Furthermore, knowing if, how, and when the early deviations can be adjusted through modification of lifestyle factors can be essential in preventing decline. Unfortunately, many efforts to track AD reduce the rich biomarker information available in neuroimaging to a single metric - hippocampal volume or amyloid positivity. The vast richness of brain wide data is often overlooked and the temporal sequence of biomarker changes is often ignored when evaluating each risk or resistive factor. We lack causal models of the biomarker changes, to identify modifiable factors in our lifestyle or genome that could resist or delay the onset of the disease, or guide treatment selection, even in the presence of confounding factors. Here we address these gaps in the AD research field with a multi-arm project that seeks to use sophisticated high-dimensional deep learning, event based modeling to elucidate the spatial rich causal sequence of brain biomarker changes that lead to AD and causal factors that influence them. We have 3 Specific Aims: Aim 1) We will fit novel adaptations of multishell dMRI models to more widely available single-shell clinical dMRI data from multiple independent public datasets to predict CSF and PET-derived Aβ load in symptomatic vs pre-symptomatic individuals. Aim 2) Use our novel AI models to generate individualized whole brain MRIs at future time points based on sex-specific normative models, creating individualized prediction maps like ‘digital twins’ of brain morphology at any age. We expect these to be more sensitive to identifying abnormal aging trajectories than current full brain methods, and more specific to disease pathology and risk factors than “brain age” alone. Aim 3) Use our novel marginal sensitivity model for continuous treatments to identify causal relationships, which are partially robust to confounding, between modifiable lifestyle factors including specific diet scores and physical activity and brain alterations in key dementia related neurodegenerative pathways. We will model causal effects and their dependency on age and sex. We hypothesize that neuroimaging metrics that show associations with earlier pathologic changes, including microstructure, will be more tolerant to confounding in retaining a causal relationship between lifestyle and brain integrity. The resulting set of tools will be more inclusive of the rich biomarker data in AD, offering a means to track its evolution and the causal factors that affect it.

NIH Research Projects · FY 2026 · 2025-01

Project Summary Precise spatiotemporal regulation of lineage-specification genes is critical not only for directing proper development in the early embryo but also for safeguarding cell fate decisions in adult tissues. Histone methyltransferases (HMTs) restrict transcription factor binding by condensing DNA into heterochromatin and thus prevents the formation of developmental disorders. This process can be reversed through the action of pioneer factors, which are capable of binding and decondensing heterochromatin. However, the mechanisms governing where and when a pioneer factor can bind to a given region remains unclear. How different states of heterochromatin may be resistant to pioneer factor binding and activity. Recent work by the Bell lab has identified a unique overlap between the histone 3 lysine 9 (H3K9) dimethyltransferase G9A/GLP and pioneer factors at lineage specification genes bound by the zinc finger protein ZFP462. However, while these overlapping regions are more accessible than those bound solely by G9A/GLP, they are not fully derepressed. In contrast, H3K9 trimethyl (H3K9me3) rich DNA established by SETDB1 remained inaccessible and devoid of pioneer factor binding. Therefore, I propose that unlike TF-resistant heterochromatin established by SETDB1, G9A/GLP creates a poised heterochromatin state which is semi-permeable for pioneer factors and thus facilitates cell fate plasticity during differentiation. To test this, I will compare the heterochromatin states established by G9A/GLP and SETDB1 at unbiased loci using two distinct methods. In aim#1, a tethering reporter strategy pioneered by the Bell lab will recruit G9A/GLP or SETDB1 to the same reporter locus through tethering minimal interaction domains (IDs). Two reporter loci will be utilized for this, one knocked into the endogenous Oct3/4 locus to investigate the established heterochromatin in the context of the OCT4 positive feedback loop and the other in a gene desert on chromosome 7 allowing for modular addition of pioneer factor binding motifs. Repression of the locus will be assessed through expression of the reporter as determined by flow cytometry and ChIP-qPCR probing of the targeted locus for the downstream histone marks and repressive machinery recruited by G9A/GLP and SETDB1. In aim#2, the ability of pioneer factors recruited by endogenous cis- regulatory elements to open heterochromatin during differentiation will be tested by a ZFP462 fusion protein with the C-terminal G9A/GLP-ID swapped with a SETDB1-ID. Through ChIP- and ATAC-sequencing, the establishment of heterochromatin and accessibility of ZFP462 target loci during directed neural stem cell differentiation will be assessed. Additionally, 10X single cell RNA-sequencing during undirected EB differentiation will determine any differences in the capacity of the ZFP462-ID fusion proteins to direct differentiation into the three germ layers. In aim#3, a 6mA footprinting analysis in conjunction with long-read genome sequencing will be performed in the chromosome 7 tethered reporter lines to identify differences in nucleosome arrangement in the different heterochromatin states, both alone and in the presence of pioneer factor binding motifs.

NIH Research Projects · FY 2026 · 2024-12

ABSTRACT We launch the MEGA-OCD Initiative - the largest, most highly powered neuroimaging investigation of obsessive compulsive disorder (OCD), drawing on diverse worldwide data to rigorously address pressing questions in OCD with unprecedented statistical power. OCD is a leading cause of global disability, and a severe and debilitating mental disorder. Symptoms start in childhood for about 50% of cases, and less than half of patients respond to first line interventions, so most OCD patients have a chronic course. Multimodal neuroimaging can elucidate circuitry and mechanisms underlying disease and treatment response, but studies have failed to address 3 major barriers to discovery: 1) inadequate sample sizes with low statistical power, coupled with variation in unimodal imaging methods, yielding a reproducibility crisis; 2) clinical heterogeneity (where brain correlates depend on developmental and disease stage); and 3) poor representation of diverse populations. To address this and NIH’s Call for Diversity, we create a highly-powered international study of unprecedented scale - the first with the global inclusion and data harmonization expertise to identify factors robustly influencing disease course, and genetic drivers of brain alterations. Drawing on 10 years’ experience coordinating international OCD studies, we combine data from two worldwide OCD consortia: ENIGMA-OCD Consortium - where 43 sites in 18 countries actively participate with data from 3,583 OCD cases (668 children/2,915 adults) and 3,505 controls (654 children/2,851 adults), and the Global-OCD Consortium - evaluating 268 deeply-phenotyped unmedicated adults with OCD, 256 healthy controls, and 80 unaffected siblings across 5 continents (North and South America, Africa, Asia, and Europe). Leveraging both samples and the expertise of these consortia, our Aims are: 1) establish biosignatures of OCD by combining structural, diffusion, and resting state functional MRI, and relate these ‘multi-layer’ signatures to age, developmental and disease stage, medication, and comorbidities. We launch worldwide analyses of task-based fMRI (focusing on emotion, inhibitory control, executive function) to yield novel functional insights into OCD; 2) use multimodal machine learning to fuse clinical and imaging metrics to: discriminate OCD from health, identify OCD biotypes corresponding to clinical subtypes, and predict treatment response; and 3) determine how OCD brain signatures relate to gene expression. A global resource for OCD researchers, our novel MEGA-OCD Initiative will identify reproducible multi-modal biosignatures of disease that incorporate diversity in developmental stage, clinical profiles (subtypes, comorbidity, cognition) and culture. Robust brain-based signatures of OCD profiles and biotypes and understanding neurogenetic risk for OCD will advance neurocircuit models of OCD, contribute to new treatment targets, and provide a foundation for precision psychiatry. Leveraging diverse samples (ENIGMA-OCD: 43 sites/18 countries; Global-OCD: 5 sites/5 continents), our findings will have both local and global impact and be relevant to populations around the globe.

NSF Awards · FY 2024 · 2024-12

The International Conference on Continuous Optimization (ICCOPT) is a premier event for optimization experts, crucial for solving complex problems in engineering, economics, business, and other domains. The conference is thus a valuable forum for graduate students in this discipline to present their work, receive feedback from other researchers working on similar or related topics, and to make professional acquaintances within the discipline. This award supports the participation of US-based students in ICCOPT 2025, held at the University of Southern California, July 19-24, 2025. Participation at this conference will enhance the research experiences for students and provide increased opportunities for new collaborations. Students form an integral part of the conference, whose first two days, known as the Summer School, cover contemporary and cutting-edge topics in optimization as presented by leading experts in the field and which are often not found in traditional textbooks. The conference also promotes the interaction between academia and industry, which benefits students through greater employment prospects upon graduation and/or better anchoring of academic research themes to industrial relevance. Funds offered through this award will support the attendance at the 2025 ICCOPT by students enrolled in a graduate or undergraduate program in the United States and whose papers have been accepted for presentation. 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

ABSTRACT Bipolar disorder (BD) is a chronic illness in which nearly half of all patients attempt suicide and no biological measures exist to reliably guide diagnosis or treatment. This project builds off 15 years of successful team science, reviving existing and independently collected neuroimaging and clinical data from around the world to create the largest studies of BD and the brain. In this new initiative from the Enhancing Neuro Imaging Genetics through Meta-Analysis (ENIGMA) Consortium Bipolar Disorder Working Group (ENIGMA-BD), we will apply a recently standardized method for large-scale voxel-wise brain mapping to improve replication and generalizability BD-related brain signatures, its clinical factors, and transdiagnostic comparisons to major depressive disorder (MDD). The ENIGMA-VBM (Voxel-wise Based Morphometry) Pipeline includes built-in voxel-wise processing, quality control, and analysis tools for both pooled mega- and federated meta-analysis. Whereas most prior ENIGMA studies have used region of interest measures in which local brain variations are averaged across parcels of the brain defined by anatomical atlas, ENIGMA-VBM quantifies whole-brain, atlas-free variations at the voxel level. This new direction allows for focal mapping across the entire brain, including frequently omitted structures such as the cerebellum. Aim 1 will compute ENIGMA-VBM measures across a target sample of 3,500 BD and 8,500 healthy controls, and analyze focal morphometric patterns associated with BD diagnostic subtypes, illness severity, polypharmaceutical treatment effects, and other important clinical factors such as age, sex, age of onset, and comorbid substance abuse. Aim 2 will derive VBM features from the ENIGMA-Major Depressive Disorder Working Group (ENIGMA-MDD) and identify overlapping and distinct voxel-wise patterns across the largest samples of BD and MDD ever studied (N=22,500). Interactions between transdiagnostic brain variations and key clinical factors such as age of onset, duration/severity of illness, number of overlapping/distinct mood episodes, and pharmaceutical treatment will be assessed. The project represents a new phase of the ENIGMA Consortium and will create a novel set of atlas-free brain phenotypes to empower the largest transdiagnostic neuroimaging studies of BD, MDD and beyond.

NIH Research Projects · FY 2026 · 2024-12

Project Summary/Abstract A sudden drop in hearing in a previously healthy adult patient is a common patient presentation. Medical terms that are used to describe the various forms of this disorder include sudden sensorineural hearing loss, cochlear hydrops, Meniere’s disease, and autoimmune inner ear disease. However, the underlying pathophysiology is poorly understood. To date, post-mortem temporal bone histopathology and animal models have provided the best data about these conditions, and it appears that a relatively common finding is increased endolymph volume, termed endolymphatic hydrops. The link between endolymphatic hydrops and human hearing loss has not been established. Here we propose to use optical coherence tomography (OCT), an imaging technique based on the use of low-power light, to visualize the fluid compartments of the human inner ear. We have built a device that permits us to see inside the lateral semicircular canal and measure the relative amounts of endolymph and perilymph. We will characterize the normal endolymph-to-perilymph ratio in control subjects undergoing mastoid surgery for middle ear disease who have normal cochlear function. In addition, we will compare the endolymph-to-perilymph ratio in subjects with Meniere’s disease and vestibular schwannoma who are undergoing mastoid surgery to treat their disease. If our hypotheses are correct, the endolymph-to- perilymph ratio will be tightly regulated in normal controls but elevated in subjects with inner ear disease. Thus, this study is designed to validate the use of OCT for human inner ear imaging and to support the use of the endolymph-to-perilymph ratio as an outcome measure for clinical trials of new treatments for Meniere’s disease. This exciting technology may also offer new approaches to assessing outcomes of innovative treatments of hearing loss or vestibular dysfunction, such as hair cell regeneration or gene therapy.

NSF Awards · FY 2024 · 2024-11

Faint dwarf galaxies are sensitive to dark matter and early universe physics. The investigators will develop a model that comprehensively predicts the population of dwarf galaxies surrounding the Milky Way as a function of dark matter properties and early universe dynamics. Comparing these predictions to the data will yield constraints on dark matter particle models and cosmic inflation. The key product of this work is an efficient and accurate model for the Milky Way’s dwarf galaxy population in non-standard cosmologies. This work will train early career researchers in broadly-applicable computational methods and support undergraduates from underrepresented groups in physics through the National Society of Black Physicists and Society for the Advancement of Chicanos/Hispanics and Native Americans in Science programs. Dwarf galaxies occupy the smallest dark matter halos that can form stars. These systems are extremely sensitive to unknown dark matter and early universe physics. To date, the faintest dwarf galaxies have exclusively been detected as satellites of the Milky Way. This situation presents a theoretical challenge: to discover fundamental physics in these data, we must model the specific formation history of our Galaxy’s dark matter halo and satellite population. The investigators will perform new suites of cosmological simulations constrained to match the Milky Way, with initial conditions appropriate for a variety of dark matter and early-universe scenarios, which will be used to calibrate the Galacticus galaxy formation model. The calibrated model, Galacticus-MW, will rapidly and accurately predict the Milky Way satellite population as a function of the beyond-CDM linear matter power spectrum. Comparing these predictions to the data will yield the first near-field measurement of small-scale matter clustering, along with constraints on dark matter properties (including its particle mass, interactions, and production mechanism) and on cosmic inflation. This work will enable cosmological inference using dwarf galaxy populations detected by Stage IV surveys over the next decade. 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 · 2024-11

Scientific data analysis is a large-scale process that involves instruments generating data at one site and networks moving the data to a high-performance computing facility where the analysis happens. Programmable networks are capable of reading the data inside data packets, opening the possibility of performing computations while data is still in transit. However, computing within the network is challenging given the limited compute and memory resources of programmable network devices. The SciWiT project plans to implement a prototype system for performing scientific data analysis using in-network and near-network resources in an optimal way. Many scientific workflows continuously monitor a phenomenon in search for rare events. This process generates enormous amount of data; thus, researchers rely on change detection algorithms to locate the rare event information. SciWiT is a computing model where programmable network resources operate on the raw data streamed through it. While data is still in transit, network identifies the regions of interest from the data stream and to provide feedback to the instrument. SciWiT plans to investigate how scientific workflows can leverage network-based in-transit computing and to develop novel in-network and near-network computing mechanisms to operate on scientific data streams. SciWiT will benefit scientific applications that rely on change detection. Moreover, this project will enhance the viability of making programmable networks an inherent computing element in the scientific data processing pipeline, effectively making the technology widely available to the scientific community. Similar to cloud environments, in-transit computing environments distributed across campuses will onboard scientific computing community to leverage the benefits of high speed programmable networks. Wide-spread adoption of the developed solutions and the downstream research enabled by the findings of this project could result in acceleration of the scientific discovery process through a fractional increase in the resources, thus benefiting the wider public. More details on SciWiT can be found at [https://gitlab.com/sciwit/public/] 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 evolution of the modern human form, with its characteristic long legs and comparatively short arms, is yet to be fully understood. This project investigates this problem by examining the fossil remains of juvenile human ancestors. The study informs how the modern human body shape evolved and contributes to the understanding of modern human adaptability and physical capabilities. This project employs advanced virtual imaging techniques and morphometric methods. By sharing the discoveries through public lectures, educational programs, and collaborations with museums, the project aims to engage and educate the public about modern humans’ ancient ancestors. This study advances scientific knowledge in developmental biology and medicine, providing educational opportunities that inform students and the public at large about human history and origins. The transition from the ancestral body shape of australopithecines to the modern human form remains unclear. To bridge this knowledge gap, this study addresses the challenge of understanding body size, proportions, and growth patterns in early Homo species, by focusing on Homo erectus. Since Homo erectus is the first hominin whose postcranial body shape is similar to the one observed in modern humans, analyzing the ontogeny of this species is fundamental to the evolutionary understanding of the the modern human body shape, body energetics, locomotion, and behaviors like long-distance walking and endurance running. Research goals include comparing growth patterns in axial and appendicular skeletons and validating anatomical reconstructions with high-resolution CT and nano-CT imaging of Nariokotome and the subadult skeleton from Dmanisi. In addition, understanding the relationship between axial and appendicular skeletons through cellular anabolic deposition patterns enhances knowledge of human skeletal growth, shedding light on the developmental processes that shaped our species. This collaborative U.S.-Swiss project is supported by the U.S. National Science Foundation (NSF) and the Swiss National Science Foundation (SNSF), where NSF funds the U.S. investigator and SNSF funds the partners in Switzerland. 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

Many geologic basins around the world are heavily faulted, with the number of faults in the hundreds. Predicting the response of such basins to earthquakes or other subsurface events requires understanding the mechanical response of the fault network to external forces like the flow of underground water. Dynamics of individual faults in the network can be different when compared with the dynamics of the overall network because the faults interact with each other via stress transfer mechanisms. This interactive behavior among faults affects the earthquake distribution and ground deformation pattern for the basin. Most existing geological models struggle to capture these dynamics because they lack the ability to account for such differences among faults within the network. This project builds an AI Model using satellite data, subsurface imagery, and other geological information to better understand fault dynamics. This work will enable assessment of regional earthquake and hazard probabilities in tectonically active regions. The model will further provide insight into sustainable and cleaner energy processes of the future. Joint workshops on AI in computational mechanics and seismology will be held to train, upskill, recruit, and reward a diverse body of undergraduate and graduate students. This project builds a multiphysics fault network model to discover reduced-order governing equations for the evolution of stress in complex fault systems. The study region is the Southern Permian Basin in the Netherlands. It uses a novel Computational Graph Discovery and Completion algorithm with Gaussian Process kernels to discover the reduced-order governing equations describing the evolution of stress and stability in the network. These network governing equations are hypothesized to provide orders of magnitude gain in computational speed relative to the current direct numerical simulation algorithms used in the field and will additionally provide insights into multiphysics effects of fluid injection/extraction on stress transfer mechanisms. This model will create new opportunities in subsurface imaging by assimilating flow, petrophysical, seismic, and geodetic data to discover hidden fault networks capable of hosting earthquakes. This award by the Division of Research, Innovation, Synergies, and Education within the Directorate for Geosciences is jointly supported by the Division of Mathematical Sciences within the Directorate for Mathematical and Physical 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

Advances in machine learning offer a tremendous opportunity to deploy learning-enabled multi-agent systems (LEMAS) in uncertain and high-stakes environments. Examples of a LEMAS could include a team of self-organizing drones in wildfire prevention or robots in semi-automated warehouses. Guaranteeing safety and robustness of LEMAS is challenging as agents form complex interacting networks where agents make individual decisions with local information. A major challenge is the effect of imperfect learning-enabled components (LECs) on the network structure. For example, a poorly calibrated perception system can result in loss of information between agents, which challenges the network's safety and robustness. Additionally, LEMAS are high-dimensional which makes it difficult to foresee all possible failure scenarios, which is a combinatorial and computational bottleneck. Addressing these challenges requires a fundamental rethinking. This project does that with a novel statistical neurosymbolic approach for the design of safe learning-enabled multi-agent systems. These approaches are neurosymbolic as they use symbolic reasoning over a novel multi-agent logic and statistical, as they provide probabilistic end-to-end safety guarantees. The project’s impacts are new theories and algorithms that will be relevant to any safety-critical LEMAS with significant societal impact on civilian and commercial applications. The project will also have significant impact on wildfire prevention by teams of drones, the primary application domain. The broader impact lies also on the educational agenda involving undergraduate and graduate level education at the University of Southern California. Each agent within a modern LEMAS uses state-of-the-art learning-enabled components for perception, trajectory prediction, and control. While learning-enabled single-agent systems are fairly well understood, it is unclear how to design safe LEMAS due to their size, complex network structure, and data dependencies between agents. To formalize safety for LEMAS, the project proposes the formal specification language “multi-agent spatio-temporal logic” (MASTL). MASTL enables to jointly reason about the network structure of LEMASs and the reliability of learning-enabled components. The project proposes a combination of statistical and formal verification techniques for MASTL to obtain provable safety guarantees for LEMAS utilizing surrogate models, reachability analysis, and statistical tests. Alongside, the project proposes reinforcement learning techniques under MASTL specifications that result in self-organizing swarm behavior. The techniques proposed in the project use photorealistic simulators that provide cheap yet realistic dataset. Because deployed LEMAS may behave differently in the real world and be subject to data distribution shifts, the project proposes training algorithms that are robust against data corruption and distribution shifts. These algorithms enjoy formal safety guarantees under quantifiable assumptions on the distribution shift. To detect extreme events and unknown unknowns in the environment, the project proposes adaptive and robust predictive runtime monitors. In the last step, the project provides real-time algorithms for self-organizing drone swarms for wildfire prevention and robots in semi-automated warehouses. 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 award will promote the progress of data science in support of the nation’s smart transportation systems through establishing a theoretical foundation for quantifying the value of information from various data sources in a transportation network. Major transportation network operators, such as the California Department of Transportation, depend on good-quality data to operate and manage their systems. Acquiring and managing data can be expensive. A crucial question arises: What data should a network operator acquire to optimize planning and operations? Priorities in data acquisition plans must align with their value for the network operator’s decision-making. Currently, there's no established theoretical framework for developing these plans. This research establishes a unifying theoretical framework for evaluating and optimizing data acquisition in a transportation network. The project will directly benefit society by facilitating effective utilization of information and leading to more sustainable and efficient transportation systems. Interdisciplinary curriculum development supported by the research findings, including modular course materials that can adapt to varying learning needs, will help better prepare and broaden participation of next-generation professionals in the smart transportation innovation ecosystem. The project introduces novel concepts that quantify the value of data through the lens of robust estimation and decision processes and translates the impact on robustness to sensitivity analysis of optimal planning problems. The project centers on three tasks: Task 1 quantifies how changes in data affect estimates of network performance metrics, which will enable a network operator to identify what data is important and how the importance varies spatially and temporally. Task 2 concentrates on modeling of data acquisition and leads to stochastic optimization models that prescribe the best data acquisition plan in support of the subsequent estimation of performance metrics. Task 3 creates three case studies for the purpose of testing and validating the methods using both real-world and synthetic data. 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

Over time, the daily decisions people make about what to eat generate a significant impact on their health and longevity. Decisions about what to eat are often influenced by the resources people have available and the options they can access. However, people with limited resources often live and spend time in neighborhoods such as “food deserts” and “food swamps” with few healthy options, leading to dietary inequities. While previous research has focused on the role of food options in people’s home neighborhoods, this project will investigate how the places people spend time in throughout their daily routines influence their dietary choices by analyzing vast amounts of human mobility data collected from smartphones. In this project, mobility data will be analyzed using Artificial Intelligence to evaluate policies that equitably increase access to healthy foods and lead to improved food choices by individuals. Working with government partners in public health and urban planning and by eliciting community feedback, interventions with high potential to increase access to healthy foods will be implemented in underserved communities in Los Angeles County. The approach provides a prototype for cross-sector partners to efficiently work together to use novel data science tools to develop and compare evidence-based policies for increasing equity in healthy food access. The SCC-Food Environment Dynamics (SCCFED) project will develop foundational methods in artificial intelligence (AI) to extract insights on the influence of food environments, including causal relationships with diet-related health, from a large body of anonymized human mobility data collected from smartphones of approximately 15 million U.S. adult residents from 2019-present. Working with community partners, these insights will directly guide the design of interventions to food environments that optimally increase equitable access to healthy foods. The research team will work with two partners in policy: the Los Angeles County (LAC) Department of Public Health and the Department of Regional Planning to co-design, estimate potential impacts of, and pilot test food environment interventions in communities in LAC’s unincorporated areas. SCCFED will develop transferrable data science tools and systems science frameworks to support cross-sector partners in research and government to leverage mobility data to evaluate which interventions can create the greatest impact, representing a new paradigm for scalable, evidence-based food environment intervention policy design. New techniques extracting greater meaning from human trajectories through time and space to study causal relationships between mobility and food choices offer broader impacts on a range of future issues in public health, urban planning, and transportation management. 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

Geologic storage of carbon dioxide (CO2) can be used to reduce CO2 in the atmosphere by removing it from the surface carbon cycle. This Industry-University Cooperative Research Center for geologic CO2 Storage Modeling, Analytics, and Risk Reduction Technologies (CO2-SMART) creates a multidisciplinary program that accelerates the safe and cost-effective sequestration of CO2 in the deep subsurface, at scale. This is done via use-inspired, basic research and workforce development in an area needed by industry. CO2-SMART is a partnership between the University of Southern California and the Pennsylvania State University. The Center focuses on advancing the current understanding of the complex processes that are triggered during and after CO2 injection into geologic formations. It develops advanced modeling and other activities that improve the safety, efficiency, and economics of subsurface CO2 sequestration operations and develops a pragmatic policy framework for large scale deployment of geologic CO2 storage. The Center brings together leading faculty, researchers, and their students from across the fields of engineering, machine learning, computer science, statistics, subsurface flow, simulation, geoscience, energy policy, and economics at the two participating universities. It enables collaboration between university personnel and CO2 sequestration field operators and carbon management regulators to develop synergistic collaborations that allow key stakeholders to tackle the challenging problem of CO2 sequestration. Broader impacts have signifcant societal, energy security, and public health implications. These include development and training of the next carbon sequestration workforce, including developing engineers and scientists as technical leaders who are well-versed in implementing and managing large-scale geologic CO2 storage projects. The University of Southern California Site will develop new courses and degree programs relevant to the topic for graduate and undergraduate students. CO2 SMART's benefits can be extended to other subsurface flow systems including those related to groundwater, geothermal, and hydrocarbon resources, all of which share similar technological challenges. An Industry Advisory Board, consisting of companies, government agencies, and other interested parties will help guide Center research to ensure it is responsive to the evolving needs of the carbon economy and companies implementing geologic-carbon storage. The CO2-SMART Center aims to enable and accelerate safe, reliable, efficient CO2 storage and implementation strategies through use-inspired research relevant to the carbon sequestration economic. It's activities also include workforce development and public outreach in this important area. The Center achieves its goals through collaborative, pre-competitive, basic research in the following key thrust areas: (1) Site screening and characterization, including fluid, rock, and fracture properties; (2) Multiscale and multi-physics modeling/simulation for prediction of CO2 displacement in storage aquifers; (3) Cost-effective reservoir monitoring and characterization, including geophysical, geochemical, and geomechanical monitoring for high-resolution imaging; (4) Machine learning and data analytics for flexible and efficient workflows, including multiphysics data processing, predictive modeling, and decision support tools; (5) Risk assessment and uncertainty quantification for complex multi-physics subsurface storage systems that to informs risk mitigation strategies; and (6) Improving the economics of geologic carbon sequestration through optimization and field development planning. The University of Southern California, the lead Site of the CO2 SMART Center, contributes its expertise in multiphysics modeling, data science, data assimilation and simulation, and uncertainty quantification, risk assessment, and optimization. It also develops a range of education and public outreach activities that involves students from local community colleges and high schools in Center research. CO2 Smart will leverage university programs for increasing K-12 STEM competency via intensive summer engineering programs and distance learning activities. It will also access and/or create student and professional certificate and degree programs in the area of energy transition, energy engineering, and environmental sustainability. 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

Many systems that are central to modern society – such as web search engines, smart assistants, generative AI, and web archives – rely on the ability to automatically load (a.k.a. "crawl") large numbers of web pages quickly. However, "web crawler" software that has been traditionally used to crawl the web is now insufficient for three reasons. First, many pages require users to be logged in. As a result, a traditional crawler sees only the login page and is blind to content that actual users would see. Second, the number of web pages is ever-increasing, and interactive pages and web applications have significantly increased the amount of computation necessary for a client to identify all the resources on a typical page. In combination, these factors make it significantly more expensive than before to crawl either a large corpus of sites or to recrawl pages frequently to capture changes. Third, many pages are dynamic or interactive, and many use embedded third-party services such as maps, social media widgets, and language translation are either hampered or fail to work on crawled page copies. As a result, systems and studies that rely on content crawled from the web lack visibility into a large portion of the web, are unable to keep up with the rate at which they need to crawl pages and end up replaying crawled pages with poor fidelity. To address these challenges, this project will develop Sprinter, a modern web crawler capable of capturing the web and its rich services as seen and experienced by users. Sprinter will crawl any page such that the content crawled is representative of what users see on the page. Its overheads will grow sub-linearly with the number of pages and the frequency of monitoring. Any page crawled using Sprinter will be renderable in a manner that closely approximates the original page, both visually and functionally. To develop Sprinter, the project will make research contributions along three dimensions. First, the project will use widespread support for authentication via single sign-on (SSO) providers such as Google and Facebook and generate representative browsing profiles from privacy-preserving network traces. Second, to make Sprinter’s crawling efficient, the project will devise techniques to reuse application computations across similar pages and to identify a small representative subset of pages that Sprinter needs to measure at high frequency. Lastly, to enable high-fidelity replay of the crawled copy of a page, the project will develop methods to crawl all of the page’s resources that will be needed to serve any common load of that copy. A major broader impact is in the research and use cases that Sprinter enables for the community. Further, Sprinter and the results of its crawls will be made available to other researchers. 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 aims to build the scientific foundations for edge-to-cloud computing and catalyze research on innovative cyberinfrastructure (CI) and workflow management for the computing continuum. Edge computing is making significant inroads into our lives and is impacting the types of science and engineering we can do. Applications spanning from the edge to the cloud are used in environmental monitoring, precision agriculture, wildfire prediction, zooplankton classification, and construction. In computer science, many researchers are exploring techniques for federated learning. CI researchers are exploring edge-to-cloud solutions that can help process data at the edge and as it moves from the edge to the data center. However, efficiently executing edge-to-cloud applications is still a challenge. Edge devices are very heterogeneous and resource-limited (computing power, storage, network bandwidth, energy power). They can have intermittent network connectivity, be mobile, and are often prone to failures. However, because they are near the network’s edge, where the data is being generated, they provide low latency and quick turnaround time for latency-sensitive applications. They can also address data privacy concerns by operating on the data in place, and the overall application can be power-efficient because data does not have to move from the edge to the cloud. In this context, this project focuses on the efficient and robust management of edge-to-cloud workflows. It includes (1) the development of a set of real-world and synthetic scientific workflows that can serve as benchmarks for evaluating edge-to-cloud algorithms and systems, (2) the development and evaluation of novel task scheduling algorithms that can optimize the performance and reliability of edge-to-cloud applications, (3) the development of an experiment management framework that enables easy execution of edge-to-cloud workflows on the computing continuum, and (4) the broad dissemination of all research artifacts through online repositories with links from the project website. The project also enables and welcomes community contributions. Results will be published in peer-reviewed journals and conference proceedings. 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

Three-dimensional, large-size, and long-duration immersive content captured from real scenes will have a significant impact in the foreseeable future. Playing a critical role in holographic communication, immersive content allows viewers to exercise 6-degree-of-freedom (6DoF) motion during playback. Most existing research on immersive content delivery focuses on single-viewer scenarios. This project proposes to enable, for the first time, a large number of co-located viewers over a millimeter wave (mmWave) network that is capable of providing high bandwidth, with a single access point and edge server. It suits numerous use cases such as massive interactive demonstration and immersive classroom education. This project aims demonstrable networking and systems research with a synergy among wireless networking & sensing, multimedia systems, machine learning, and computer vision. It will help bridge the digital divide by reducing the cost of multi-user holographic communication and telepresence. It will also provide a platform to conduct various outreach activities and community services. As streaming emerging multimedia content is playing a key role in the post-COVID world, the project will have a high impact on global societies and economies. To overcome the challenge of supporting multiple users with limited network and compute resources, this project innovates in three key dimensions. First, it will develop an accurate motion prediction model that captures users' collective motion and their interactions, and study how to adapt to changes deviating from training data. Second, this project will leverage mmWave sensing based on FMCW (frequency-modulated continuous-wave) radar to directly incorporate environment reflection profiles into beamforming and mmWave throughput prediction. Assisted by 6DoF motion prediction, this will lead to proactive and fast beamforming, as well as an accurate forecast of mmWave performance that benefits upper layers. To realize environment profiling based on mmWave sensing, the project will design two techniques: collaboratively reconstructing indoor 3D reflectivity maps and building a neural representation of indoor mmWave reflections. Third, this project proposes two approaches to scale up at the application layer: hybrid streaming where certain viewers receive 3D content and others consume content live-transcoded by the edge, and allowing viewers to share a transcoded view. The team will integrate the above thrusts into a holistic framework, implement it on their mmWave testbed with heterogeneous client devices, and conduct extensive evaluations including field trials with real users. 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 2024 · 2024-09

ABSTRACT In this carefully revised application, we launch “Federate AD” - an international alliance to study Alzheimer’s disease (AD) - which combines innovative AI techniques and distributed computation to address urgent challenges in AD research. In 2023, AD and other dementias will cost the nation $345 billion; the disease kills more people than breast cancer and prostate cancer combined. With 6 million people living with AD in the U.S. alone, there is an urgent need to discover factors in the genome and environment that promote or resist dementia. The high cost of collecting neuroimaging and genomic data has led to very large numbers of small sample studies, each with limited power to detect or verify factors that influence disease onset and progression. To address this, we create a distributed artificial intelligence (AI) platform that allows researchers worldwide to compute on Alzheimer’s disease biobanks, uniting expert teams worldwide to discover factors that influence AD onset and progression in worldwide populations. Our Secure Federated Learning architecture computes on AD biobanks in India, Japan, the U.S., and Europe, numbering over 100,000 MRI and PET scans, to (1) diagnose and subtype dementia, using deep learning in large neuroimaging biobanks, (2) infer brain amyloid and tau burden from less invasive biomarkers, such as clinical data and brain MRI and DWI, and (3) predict who will decline clinically from health or mild cognitive impairment to AD. This work will facilitate clinical trial selection and personalized prognosis. Building on our work creating productive, successful worldwide imaging genomics consortia such as ENIGMA and AI4AD, our major innovations include (1) running AI methods on datasets located across the world, (2) enhancing diversity and reducing bias by also learning from diverse non-European datasets, including South Asian and East Asian datasets. Technical innovations include the use of federated deep learning and data harmonization methods that yield site-invariant predictors that generalize better and yield robust predictive models across ancestries. Building on prior DoD funding, we use secure homomorphic encryption to ensure privacy by addressing serious data leakage problems, in which private data can be deduced from AI models. As a whole, this project will produce a social network of AI innovations for AD, and a toolkit for AD researchers to use and build on for applying AI methods in a distributed or centralized setting. The vast datasets from the U.S. ADSP, UK Biobank, and Indian and Japanese repositories have not previously been collaboratively analyzed, and promise insights into common and ancestry-specific predictors of decline. By dynamically learning from worldwide data, new insights will accrue as new data is added. To maximize this work’s impact on the AD field, we carefully integrate our efforts with NIA-funded initiatives in machine learning and phenotypic harmonization, in which we participate. The resulting AI alliance will better integrate worldwide data into AD research, allowing researchers to participate and learn from each other’s data, expertise, and predictive models.

NIH Research Projects · FY 2025 · 2024-09

ABSTRACT In this carefully revised application, we launch “Federate AD” - an international alliance to study Alzheimer’s disease (AD) - which combines innovative AI techniques and distributed computation to address urgent challenges in AD research. In 2023, AD and other dementias will cost the nation $345 billion; the disease kills more people than breast cancer and prostate cancer combined. With 6 million people living with AD in the U.S. alone, there is an urgent need to discover factors in the genome and environment that promote or resist dementia. The high cost of collecting neuroimaging and genomic data has led to very large numbers of small sample studies, each with limited power to detect or verify factors that influence disease onset and progression. To address this, we create a distributed artificial intelligence (AI) platform that allows researchers worldwide to compute on Alzheimer’s disease biobanks, uniting expert teams worldwide to discover factors that influence AD onset and progression in worldwide populations. Our Secure Federated Learning architecture computes on AD biobanks in India, Japan, the U.S., and Europe, numbering over 100,000 MRI and PET scans, to (1) diagnose and subtype dementia, using deep learning in large neuroimaging biobanks, (2) infer brain amyloid and tau burden from less invasive biomarkers, such as clinical data and brain MRI and DWI, and (3) predict who will decline clinically from health or mild cognitive impairment to AD. This work will facilitate clinical trial selection and personalized prognosis. Building on our work creating productive, successful worldwide imaging genomics consortia such as ENIGMA and AI4AD, our major innovations include (1) running AI methods on datasets located across the world, (2) enhancing diversity and reducing bias by also learning from diverse non-European datasets, including South Asian and East Asian datasets. Technical innovations include the use of federated deep learning and data harmonization methods that yield site-invariant predictors that generalize better and yield robust predictive models across ancestries. Building on prior DoD funding, we use secure homomorphic encryption to ensure privacy by addressing serious data leakage problems, in which private data can be deduced from AI models. As a whole, this project will produce a social network of AI innovations for AD, and a toolkit for AD researchers to use and build on for applying AI methods in a distributed or centralized setting. The vast datasets from the U.S. ADSP, UK Biobank, and Indian and Japanese repositories have not previously been collaboratively analyzed, and promise insights into common and ancestry-specific predictors of decline. By dynamically learning from worldwide data, new insights will accrue as new data is added. To maximize this work’s impact on the AD field, we carefully integrate our efforts with NIA-funded initiatives in machine learning and phenotypic harmonization, in which we participate. The resulting AI alliance will better integrate worldwide data into AD research, allowing researchers to participate and learn from each other’s data, expertise, and predictive models.

NIH Research Projects · FY 2024 · 2024-09

PROJECT SUMMARY Our motor, cognitive, and emotional functions arise from the temporal dynamics in the activity of large populations of neurons across brain networks. Dynamic models of neural population activity describe it in terms of low-dimensional latent factors. Recent neural network models of population activity have shown the importance of capturing its nonlinearity. But despite their power, these state-of-the-art nonlinear neural networks have a non-causal inference structure and do not directly address missing neural data, which can happen in wireless naturalistic recordings. Also, these neural networks don’t enable closed-loop control. These issues have limited the utility of these neural network models to correlational validations of hypotheses/models and to experiments with constrained behaviors. But a major emerging goal in neuroscience is to enable causal validations of hypotheses/models with real-time and/or closed-loop perturbation experiments and to study naturalistic behaviors in unconstrained or long-term experiments. We will develop dynamical models that not only capture nonlinearity, but also enable two novel critical capabilities: flexible inference and closed-loop control of latent factors and behavior. Flexible inference is defined as the model’s ability to allow for all the following: a) causal/real-time inference, b) non-causal inference, c) accounting for missing neural data. Also, to extend our models for diverse neuroscience investigations, we will develop multiple novel learning methods to address 3 additional major attributes of neural population dynamics that remain challenging to capture in nonlinear models: i) these dynamics happen at multiple spatiotemporal scales, ii) relate to a mixture of behaviors and internal states that co-occur, and iii) are driven by both intrinsic dynamics and inputs. Thus, this program will provide novel neural network models of neural population activity that can capture nonlinearity, flexible inference, and closed- loop control, and can also admit multiscale data, dissociate behaviorally relevant vs. irrelevant dynamics, and disentangle input vs. intrinsic neural dynamics. We will also develop and share extensive software and documentation for these models and methods. We will comprehensively validate the models and methods on diverse existing nonhuman primate and human datasets including public datasets with multiscale neuronal spiking, field potential, and intracranial recordings, from different brain regions, with measured sensory, neural and neurostimulation input, and during various cognitive and motor behaviors. Our program includes a diverse network of investigators and end-users and outreach activities to enhance diverse perspectives. The developed models and methods will greatly impact diverse neuroscience domains by providing novel tools to causally drive and test new hypotheses about how multiscale neural dynamics control behavior, inform experimental paradigms and data-collection (e.g., closed-loop perturbations), and enhance neurotechnology for decoding/modulation.

NIH Research Projects · FY 2025 · 2024-09

Project Summary Neuropsychiatric symptoms (NPS) represent a significant but frequently overlooked facet of Alzheimer’s disease (AD). These symptoms, which encompass non-cognitive disturbances like apathy, depression, anxiety, agitation, and aggressiveness, are prevalent in patients with AD and related dementias and can potentially affect the trajectory of AD progression. While AD-NPS impose major challenges on the well-being of patients and burdens on caregivers, our understanding of their origins, interconnections and neural mechanisms in relation to pathophysiology remains limited. This proposed research seeks to understand the neural foundations of NPS-related behavioral abnormalities in mouse AD models. As a starting point, we will focus on behaviors related to apathy, and test our central hypothesis that pathological and molecular/cellular changes in the anterior cingulate cortex (ACC) are accompanied with altered excitation/inhibition (E/I) balance which leads to behavioral impairments, while rebalancing the E/I may ameliorate apathy-like deficits in two mouse AD models. By combining a suite of cutting-edge approaches, we will integrate multimodality data from behavior assessments to spatial transcriptomics, anatomy, pathology, and neural function of the brain. In Specific Aim 1, we will explore cortical pathological and molecular/cellular/transcriptional changes associated with the progression of NPS-like behavioral impairments. In Specific Aim 2, we will determine how cortical neuronal functional changes contribute to apathy-like deficits in AD. The proposed study aims to generate new mechanistic understandings of neuropsychiatric impairments in AD, which may help to develop novel strategies for patient care and therapeutic approaches to mitigate adverse impacts of NPS in AD management.

NIH Research Projects · FY 2024 · 2024-09

Abstract The purpose of this study is to examine alcohol consumption as a modifiable risk factor for Alzheimer's disease and related dementias (ADRD). Approximately 40% of risk for dementia has been attributed to modifiable risk factors, with heavy alcohol use in midlife (12+ standard US drinks weekly) added to the list of risk factors by the Lancet Commission in 2020. It is well established that more women have ADRD and men have higher rates of heavy drinking, yet alcohol-related problems occur at lower levels of consumption for women than for men. Moreover, the APOE gene, one of the strongest genetic risk factors for ADRD, begins to increase risk for women who have only one e4 allele but for men increased risk is only seen in those who have two e4 alleles. Prior research has also been mixed as to whether APOE interacts with alcohol consumption to increase risk for ADRD. This exploratory/developmental project will examine the relationships of midlife alcohol use and APOE in over 32,000 male and female twins from the Swedish Twin Registry who have lifetime alcohol consumption data and clinical or registry-based diagnoses of ADRD, plus APOE genotypes in over 8,000 twins, thus making it possible to detect phenotypic relationships as well as additive (i.e. G+E) and interactive (i.e. GxE) genetic and environmental relationships. Specifically, in Aim 1 we will delineate the relationship of mid-life alcohol consumption with late-life ADRD diagnosis and age of onset, paying particular attention to whether these relationships are similar in men and women. We will also explore whether incorporating alcohol consumption levels from earlier and later adulthood adds to our understanding of alcohol consumption as a lifecourse risk factor for ADRD. In Aim 2, we will test the alcohol consumption-APOE risk relationship with ADRD diagnosis and age of onset, again focusing on sex differences in elucidating these relationships. Finally, in Aim 3 we will leverage the genetically informative design of the Swedish Twin Registry to characterize differences in alcohol involvement within ADRD discordant and concordant monozygotic and dizygotic twin pairs of both sexes, thereby better understanding genetic contributions to alcohol-ADRD associations. This exploratory/ developmental research study will have the power to model alcohol risk for ADRD in nuanced ways and broaden our understanding of how and why alcohol may affect ADRD risk. If successful, it will set the stage for us to incorporate alcohol consumption as a risk factor for ADRD in more sophisticated models in future work.

NIH Research Projects · FY 2025 · 2024-09

Working-aged American adults spend large portions of their days engaged in work—a major non-medical correlate of health. The growth of non-standard work arrangements in the United States has had detrimental impacts on the health and well-being of workers. These evolving arrangements are defined as emergent work (EW), or contemporary work that is outside of traditional full-time employment, expected to end, lacking an implicit or explicit contract, and/or precarious. This national trend towards EW may also exacerbate health disparities in workers who engage in such work. Developing a nuanced understanding of EW and its relationship to health, well-being, and health disparities is essential for improving population health in the United States. Toward this end, this project includes a comprehensive training plan and integrative investigation of EW in California, a state at the forefront of this labor market trend. The project PI is a postdoctoral research associate in the University of Southern California’s Chan Division of Occupational Science and Occupational Therapy. The PI will work with a mentorship team of worker health and methods experts to complete a multi-method study of the state-wide California Work and Health Survey, a comprehensive study of work and health factors in 4,104 workers from the California labor force. Aim 1 will identify latent classes of EW among workers aligned with contemporary work arrangements. Aim 2 will characterize contemporary EW based on worker demographics, health, and well-being. Aim 3 will examine workers’ lived experiences of participation in EW relative to health and well-being. Completing these scientific aims will contribute to a comprehensive understanding of objective and subjective factors influencing health and well-being in EW and provide insight into personal factors informing the relationships between EW and potential health disparities. Concurrent didactic, project-based, and professional development activities will complement mentorship and pursuit of these three aims. Specifically, training in data science and survey analysis methods will support the completion of Aims 1 and 2, and instruction in qualitative methods will support the completion of Aim 3. Coursework will be completed to advance understanding of work as a non-medical correlate of health and research methods, and efforts will be directed toward professional development and dissemination. These integrative experiences with a mentored research project and a comprehensive training plan will launch the PI’s career as a junior scientist interested in understanding non-medical correlates of health in workers and addressing worker health disparities. This project aligns with the National Institute on Minority Health and Health Disparity’s research concept about the role of work in health disparities in the United States, as well as the Healthy People 2030 initiative’s focus on the work and the workplace as community-level correlates of health.

NIH Research Projects · FY 2025 · 2024-09

Project Summary Discovery of selective and potent ligands for novel targets, as well as ligands with new functionalities remains a slow and expensive process, hampering pharmacological validation of targets and discovery of new treatments for many conditions like pain, substance abuse, Alzheimer's disease and cancer. Our research aims at developing a scalable computational platform for ligand discovery, synergistically combining the advantages of structure-based and data-driven approaches. We will pursue three synergistic technological directions, combined with their experimental validation and application to clinically relevant targets. The first one builds upon our recently introduced highly scalable synthon-based approach, V-SYNTHES, which performs hierarchical structure-based virtual screening of giga- and tera-scale on-demand chemical spaces. We will further improve the performance of the method by employing complementary Machine Learning approaches to ligand selection, and optimize V-SYNTHES parameters to expand it to Tera-scale REAL libraries. The second research direction combines V-SYNTHES screening with synergistic experimental hit identification approaches like fragment- based, covalent screening and DNA-Encoded Library. Such hybrid methods build on complementary strengths of these tools, enabling ligand discovery for the most challenging targets like cryptic and allosteric pockets. Finally, building upon our extensive experience with the rational design of functionalized ligands, we will explore structure-based approaches to designing photoswitchable, irreversible, bitopic and bivalent ligands, based on both derivatives of known ligands and newly discovered chemotypes. Our broad network of experimental collaborators will allow rapid synthesis of predicted compounds, and their comprehensive experimental validating in biochemical, cellular and in vivo assays. Successful completion of this project will establish robust computational and hybrid platforms for structure-based ligand discovery in most classes of therapeutic targets, scaleable for rapidly growing REAL chemical spaces. The platform will be also thoroughly validated on a diverse set of clinical targets, yielding new chemical probes and potential leads for drug discovery.