University Of California Berkeley

NIH Research Projects · FY 2026 · 2026-04

Mycobacterium tuberculosis (Mtb) infects one-quarter of the global population and causes more deaths annually than any other infectious disease. Despite triggering a strong immune response, Mtb evades and suppresses host immunity, establishing chronic infection. Two critical virulence factors, the ESX-1 type VII secretion system and the mycomembrane lipid phthiocerol dimycocerosate (PDIM), play a key role in this immune modulation, but their precise mechanisms remain unclear. Recent data suggest that ESX-1 and PDIM may functionally intersect, as mutants in these pathways exhibit similar virulence phenotypes. This proposal aims to investigate whether ESX-1 and PDIM operate within the same virulence pathway and how PDIM influences ESX-1. Additionally, we will explore whether their roles in virulence involve suppression of protective Th17 CD4 T cell responses and induction of immunosuppressive type I interferons (IFNs). Aim 1 will test whether ESX-1 and PDIM exhibit epistasis for virulence in mice and define how PDIM impacts ESX-1 function. We hypothesize that PDIM mutants subtly affect ESX-1 activity, which will be tested through epistasis experiments and proteomic profiling of ESX-1 function in PDIM mutants. Aim 2 will investigate whether blocking IL-23 explains the suppression of Th17 differentiation by ESX-1/PDIM mutants. We have shown that these mutants induce higher IL-23 production and a robust Th17 response, suggesting that ESX-1 and PDIM suppress this protective response. We will test the role of IL-23 suppression in Th17 differentiation and perform epistasis testing for CD4 T cell orchestration. Aim 3 will determine whether type I IFN induction is necessary for the virulence of ESX-1 and PDIM mutants. Both factors induce type I IFN, which impairs host responses to Mtb. We will test if type I IFN induction is critical for virulence by evaluating mutant strains in mice lacking type I IFN signaling. These studies will provide valuable insights into Mtb's immune evasion strategies, potentially leading to new therapeutic interventions.

NSF Awards · FY 2026 · 2026-04

This award is jointly supported by the Major Research Instrumentation and the Chemistry Research Instrumentation programs. The University of California, Berkeley, is developing a next-generation tabletop ultrafast soft x-ray spectroscopy instrument to support the research of Professor Michael Zuerch and colleagues Stephen Leone and Daniel Neumark, as well as a broad community of campus and external users. By establishing a campus-based hub for ultrafast soft x-ray science, the project increases access to advanced x-ray capabilities for chemists, physicists, materials scientists, and engineers. The instrument produces high-flux, tunable soft x-ray pulses with durations reaching the few-femtosecond and attosecond regime and measures element-specific absorption changes as chemical and material systems evolve in real time. The design enables systematic, comparative studies across phases of matter and significantly expands the capabilities of existing laboratory x-ray tools. The system will be operated as a shared-use resource with structured access for internal and external collaborators. The project also incorporates a coordinated education and workforce development program that provides undergraduate and graduate students with hands-on training in high-power lasers, vacuum technology, x-ray optics, and ultrafast spectroscopy. The award addresses the development of an instrument which integrates an optical parametric chirped pulse amplification laser platform with modular sample environments spanning gases, liquids (including ultrathin liquid jets), and solids within a single shared-use environment. By extending laboratory soft x-ray spectroscopy to photon energies up to approximately 650 eV, the system enables direct observation of electron motion, charge redistribution, and bond rearrangements with sensitivity to carbon, nitrogen, oxygen, sulfur, and first-row transition metals. The instrument provides element-, site-, and oxidation-state-specific sensitivity through core-to-valence transitions and supports soft x-ray transient absorption and attosecond pump–probe measurements. Enabled research includes molecular charge migration, excited-state dynamics in solution-phase organic systems, water radiolysis at the oxygen K-edge, and ultrafast spin and charge dynamics in magnetic and photoactive materials. By enabling element- and site-specific tracking of charge, spin, and orbital dynamics on their natural time scales, the instrument provides direct access to decoherence pathways and nonequilibrium control mechanisms in quantum materials relevant to quantum information science. These capabilities support studies of ultrafast spin and valley dynamics, correlated electron phases, and light-induced symmetry control in solid-state platforms, informing the design of materials and control strategies for future quantum devices. Through research participation, workshops, and mentored projects, students gain technical expertise directly aligned with national needs in x-ray science, future quantum information science platforms, advanced manufacturing, microelectronics, and energy technologies. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NIH Research Projects · FY 2026 · 2026-04

Abstract There are 118 known elements. Nearly all of them have nuclear magnetic resonance (NMR) active isotopes and at least 39 different nuclei from 33 elements have been used in biological and biomedical NMR studies. Despite the availability of dozens of NMR active isotopes (2H, 7Li, 13C, 17O, 23Na, 31P, 35Cl, 39K, etc.), most of today’s MRI is based on one nucleus – 1H. Since its inception in the 1970s, MRI technology has made immense gains in SNR with hyperpolarization, high and ultra-high field magnets, anatomy-conforming receiver coils, improved reconstruction, and other techniques. With these SNR gains, the imaging of nuclei other than 1H, or X- nuclei, has become more clinically feasible, inspiring a variety of studies capitalizing on the essentially perfect nuclear specificity of NMR/MRI to gain information not possible with 1H alone. Notably, hyperpolarized media and deuterium imaging have made significant gains recently. These and further studies, however, are still held back by technical challenges and the low availability and high cost of the necessary tools. To overcome these bottlenecks, we aim to develop an RF system, called the ADAPT PRO system, that can be digitally programmed on the fly to image any nucleus of interest independently or simultaneously. The system will bring out the full potential of all NMR active nuclei, significantly enhancing disease knowledge, diagnoses, and treatment evaluations. X-nuclei benefits have already been shown for cancer, osteoarthritis, Alzheimer’s, and many more. The system can be mass manufactured on assembly lines without the need of highly trained coil engineers. As such, it can be produced at orders-of-magnitude lower cost, thus facilitating the clinical translation and democratization of X-nuclei spectroscopy and MRI in general. Our innovative approaches have independent transmit and receive components. The transmit side integrates high-frequency, high-power switches into the coil structure, merging the RF amplifier and coil into a single programmable device that converts DC power to any RF frequency of interest. The receive side uses high-frequency, low-noise variable capacitors (varactors) driven to convert received MRI signals from an untuned coil to the ~500 MHz range, which are then amplified by a resonant ~500 MHz circuit. These advances promise to bring MRI coils to the digital age, enabling vastly more capabilities via programmability. Any-nucleus imaging is one new capability, and more potential capabilities include magnetic field shimming for undistorted data, improving quantification by reducing coil loading effects by the patient, and being reused between scanners of different field strengths, including emerging low-field portable scanners. Our proposed work has the potential to solve a wide range of important problems all at once.

NIH Research Projects · FY 2026 · 2026-04

PROJECT SUMMARY / ABSTRACT The internal representation of space in the mammalian brain is crucial for supporting an animal's ability to navigate complex environments. Grid cells in the medial entorhinal cortex (MEC) are believed to provide a reliable and scalable representation of an animal's position through periodic firing fields arranged in a hexagonal pattern during 2D navigation. Recent experimental findings strongly support continuous attractor network (CAN) models that posit that grid cell activity is constrained to a 2D toroidal manifold which, when anchored to behavior, gives rise to structured firing patterns. However, it remains unresolved how grid cells subserve complex spatial behavior such as 3D navigation, and whether CAN models generalize across species with distinct navigational demands. Prior studies investigating grid cells in flying bats suggested that grid cell spatial responses lacked global structure, contrasting sharply with structured grid cell responses found during 2D navigation in both bats and rodents, However, these studies focused on unstructured navigation driven by human intervention, and were limited to single cell analysis of small neural populations. Spatial representations in bats are known to be modulated by non-positional aspects of behavior as well as the presence of human experimenters. Thus, how grid cells represent the bat’s natural, self-selected flight patterns is not well understood. Furthermore, MEC is intricately connected with the hippocampus, which has been shown to reflect the animal’s action plans. However, whether action plans emerge in MEC during ethological goal-directed navigation remains unexplored. By leveraging the ethological advantages of bats and breakthroughs in wireless Neuropixels recording techniques, this proposal aims to examine how grid cells underlie bats' natural, self-selected flight paths during 3D navigation and connect these findings to existing grid cell models. In Aim 1, I will characterize grid cell spatial responses and population dynamics during 3D navigation to test the hypothesis that grid cells exhibit periodic spatial responses along bats’ natural flight paths and whether grid cell population activity is constrained to a toroidal manifold. In Aim 2, I will investigate non- spatial representations to test the hypothesis that MEC reflects navigational action plans during goal-directed 3D navigation. The proposed work could uncover fundamental mechanisms that are conserved across species and may shed light on how MEC supports spatial memory functions, providing vital insights that could help advance therapeutic approaches for neurological disorders such as Alzheimer's. The proposed fellowship training plan provides a comprehensive training strategy to develop the necessary expertise to carry out this research project in the Yartsev Lab, one of the world’s leading bat neuroscience labs.

NIH Research Projects · FY 2026 · 2026-03

PROJECT SUMMARY The Song lab consists of a diverse team of computer scientists, statisticians, and mathematicians dedicated to advancing biology. Their work focuses on developing efficient computational tools, robust statistical methods, and innovative machine learning models for understanding evolution and fundamental biological processes. Their goal is to facilitate the broader biomedical community’s research while also getting deeply involved in data analysis and interpretation to help make new biological discoveries. A central goal of biology is to unravel the wealth of information contained in the genome. Achieving this would enable the integration of personal genome interpretation into healthcare, aiding in disease diagnosis, personalized therapeutic regimens based on individual genetic makeup, and improved treatments with reduced side effects. The Song lab’s research program aims to help realize this transformative vision by addressing critical computational, statistical, and modeling challenges. Recent advances in machine learning (ML) and artificial intelligence (AI) have profoundly impacted diverse sci- entific fields, transforming approaches to model development, experimental design, data analysis, interpretation, and discovery. Applying AI/ML to biology holds enormous potential, but fully realizing this potential and making it accessible to the broader biomedical community requires addressing several key challenges. For instance, in- corporating biological knowledge and insight into AI/ML models is crucial for training effective models. Achieving this requires innovation in curating suitable training data, which demands a deep understanding of the specific application domain; developing appropriate model architectures and tuning their hyperparameters; and designing effective learning objectives. Evaluating and interpreting the trained models also require a substantial amount of rigorous work. Furthermore, training advanced AI/ML models typically demands large computational resources, which can seriously limit model development, exploration, and utility; thus, novel approaches are needed to train models more efficiently. This project aims to tackle these challenges and help bridge AI/ML with basic biomedical research. Over the next five years, the Song lab will investigate several basic research problems in evolutionary biology and genomics, and develop a suite of robust, scalable AI/ML models and statistical methods to benefit the broader community. In particular, they will develop innovative models to learn complex probability distributions over biological sequences (DNA, RNA, and proteins), to decipher the intricate information contained in them and to understand the functional constraints they entail. These efforts will have wide-ranging applications, including phylogenetic tree reconstruction, viral evolution prediction, variant effect prediction, transfer learning in genomics, and protein design. In parallel, the lab will also develop novel computational methods for genomics and continue collaborating with biologists to tackle basic research questions. Lastly, this project will integrate research with education to train a generation of researchers capable of developing cutting-edge AI/ML models for biology.

NSF Awards · FY 2026 · 2026-03

With support from the Chemical Structure and Dynamics (CSD) program in the Chemistry Section, Professor Stephen Leone at the University of California, Berkeley is establishing a comprehensive framework for characterizing and controlling quantum coherence and entanglement in atomic and molecular systems. This program aims to address entanglement as a time-resolved parameter in chemical dynamics. The challenges of identifying and experimentally characterizing entanglement dynamics require precision laser measurements that can project the necessary coherence and entanglement information onto optical detectors on very short timescales. To accomplish this, Professor Leone and his students will integrate coincidence detection of dual velocity map imaging spectrometers with a tabletop attosecond extreme ultraviolet system to probe and quantify the ultrafast dynamics of electron-ion entanglement following photoionization. The research will obtain entanglement dynamics on very short timescales and directly test the interplay between coherence and entanglement in chemical processes. These studies will advance the fundamental understanding of ultrafast quantum dynamics and lay the groundwork for future applications in quantum information processing and quantum sensing. Students trained on this project learn strategic technologies that impact the economy, related to precision timing and synchronization, nanopatterning and lithography at small dimensions, optical communications, and sensing using quantum information science principles. Previous studies of chemical kinetics and dynamics produced a wealth of information about product species, state-resolved dynamics, and coherence phenomena, establishing a broad basis for understanding mechanisms. Whenever two particles are formed from one particle in a chemical process, such as photoionization or dissociation, entanglement can occur between the outgoing particles. Here entanglement is measured by probing features such as quasi-bound autoionizing states embedded in a continuum during photoionization. This imparts nontrivial dynamics to the photoionization, resulting in modifications to the degree of entanglement versus time between the resulting photoelectron and ion due to quantum path interferences of the finite lifetime autoionizing states. The ultimate timescale for the formation of the final degree of entanglement will also be accessible as a function of chemical system. When one of the particles also has two or more formal quantum states, coherence can occur among these states. The tradeoff between entanglement and coherence is anti-correlated, providing a deep understanding of the evolution of the quantum dynamics that governs the chemical process. This is made possible by simultaneous tomographic and coincidence measurements of processes that take place in a continuum. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NIH Research Projects · FY 2026 · 2026-03

ABSTRACT Animal genomes are replete with sequence added by the cDNA synthesis activity of non-LTR retrotransposon reverse transcriptases (RTs). These enzymes insert sequence directly into the genome, using a nick to initiate cDNA synthesis on an RNA template, a different strategy than the better-studied integration of an extra- genomic, double-stranded DNA intermediate in LTR retrotransposon and retrovirus life cycles. The single catalytic protein encoded by non-LTR retrotransposons evolved coordinated RNA binding, DNA binding, nicking endonuclease, and RT activities. The enzyme of the human non-LTR retrotransposon LINE-1 inserts new DNA with minimal specificity for target site sequence, allowing it to generate a third of our genome and to continue the mobility that imposes human disease. On the other hand, more ancestral non-LTR retrotransposon proteins generally have target-site specificity for safe-harbor loci in their host genome, which is enabling for their evolutionary persistence. Recently we harnessed an avian R2 non-LTR retrotransposon protein (R2p) to direct site-specific insertion of autonomously expressed transgenes at safe-harbor rDNA loci in the human genome. These rDNA loci generate the precursor of 3 ribosomal RNAs, so they are present in hundreds of copies per genome with the consequence that disruption of a few units is not deleterious. Our method, termed PRINT (Precise RNA-mediated Insertion of Transgenes), requires delivery of only two RNAs: an mRNA encoding R2p and a template RNA encoding the transgene flanked by compact 5’ and 3’ RNA modules. We exploited PRINT to determine mechanisms for the “fill-in” second-strand synthesis necessary to complete gene insertion, which had remained a mystery despite importance for general genome repair. We will use biochemistry, biophysical approaches, PRINT, and native R2 insertion assays developed in the lab, as well cryoEM structures and a newly annotated “zoo” of >300 R2s across the breadth of animal phylogeny, to fulfill 2 synergistic, impact-generating, broad goals: (i) filling knowledge gaps about non-LTR retrotransposon protein structure/function and the cellular mechanisms involved in gene insertion, and (ii) advancing PRINT as an approach for therapeutic gene delivery. Safe-harbor transgene insertion using PRINT is an approach to disease therapy that is complementary to gene disruption or nucleotide correction by CRISPR/Cas [or other method of] introduction of a DNA break, base editing, or prime editing. As additional motivations, the proposed work will (iii) elucidate biochemical and structural principles for unusual nuclease and polymerase activities and highly specific protein-nucleic acid interactions, (iv) illuminate new principles of RNP subcellular trafficking, and (v) define DNA repair pathways and transgene expression features that are specific to nucleolar rDNA versus general chromatin as a transgene location. Overall our studies will increase the efficiency, safety, and application range of a promising genome engineering therapy for human loss-of- function diseases, as well as inform mechanisms used by our deleteriously mobile retrotransposon LINE-1.

NIH Research Projects · FY 2026 · 2026-03

PROJECT SUMMARY/ABSTRACT To choreograph the many processes required for life, cells maintain precise control of gene expression. Nuclear factors regulating transcription are linked to a wide spectrum of human cancers, and therefore understanding their functions is of great interest to human health. Many of these regulators function by modulating transcription factor IID (TFIID), a component of the core transcriptional apparatus. TFIID plays a critical, fundamental role in transcriptional activation, being the molecular target of diverse transcription factors. Moreover, it interacts with the DOT1L complex, a histone modifier and reader, thus linking gene activation to chromatin state. Despite decades of study, our molecular understanding of how regulatory factors modulate TFIID structure and function remains limited. This proposal seeks to address this gap in understanding by applying structural and biochemical approaches to investigate the role of TFIID interactors in promoting transcription. Specifically, two distinct interactors will be studied: (1) E proteins, such as E2A, which are transcriptional activators that enhance TFIID promoter binding, and (2) the DOT1L complex, a chromatin regulator possessing several histone reader domains, which recruits TFIID to promoters via chromatin recognition. The structural and mechanistic insights generated by the proposed research will greatly enhance our fundamental understanding of gene regulation. Furthermore, both E proteins and DOT1L are linked to human diseases and play crucial roles in leukemogenesis, underscoring the potential of novel mechanistic insights to bolster our understanding of disease mechanisms and inspire new therapeutic approaches to benefit human health. The proposed research will be conducted in the Nogales Lab at the University of California, Berkeley, which is a leader in the field of structural biology. The Nogales Lab has deep expertise in the technique of cryo-electron microscopy (cryo-EM), which is the primary structural technique used in the proposed work. The Nogales Lab is affiliated with the California Institute for Quantitative Biosciences (QB3), which houses key facilities such as the Cal-Cryo microscopy facility and the QB3/Chemistry Mass Spectrometry Facility, which will provide crucial instrumentation and experimental support. The fellowship proposal also encompasses a training plan tailored to gaining experience critical for a future career in academia. In addition to providing the opportunity to develop rich scientific expertise in structural biology and gene regulation, the proposed work will also provide crucial training in mentorship, project management, and scientific communication.

NSF Awards · FY 2026 · 2026-02

Solar energetic particles can disrupt satellites, communications, navigation, and power systems, and they can pose risks to astronauts and spacecraft. This project seeks to improve understanding of how energetic particles are transported and accelerated from the Sun through interplanetary space, especially during events when particles stream back toward the Sun and affect Earth. The work will contribute to the scientific basis for space weather forecasting and mission planning. The project will support an early-career researcher and provide summer research opportunities for undergraduate students. This project will investigate how energetic particles from both solar and heliospheric sources form and evolve, with a particular focus on particles streaming sunward in the inner heliosphere. The team seeks to identify particle sources and formation mechanisms, including how solar activity may contribute to their solar cycle dependence; characterize the radial and longitudinal evolution of sunward-streaming energetic particle events; and examine the factors that influence differences between sunward-streaming energetic protons and electrons. The analysis will use spacecraft and ground-based datasets obtained during the past solar cycle, including the NSF-funded Expanded Owens Valley Solar Array (EOVSA). The study will leverage advances in anisotropy characterization and multi-platform observations to examine sunward-streaming energetic particles across one solar cycle. 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 · 2026-02

ABSTRACT Parkinson's Disease (PD) is an age-related progressive neurodegenerative disorder primarily characterized by the loss of dopamine (DA) neurons in the midbrain, which is crucial for motor control and balance. Current treatments, such as L-DOPA and deep brain stimulation, provide symptomatic relief but crucially, do not halt the progression of neurodegeneration. While familial PD cases have highlighted mutations in approximately 20 genes, the majority of PD cases are sporadic, likely resulting from complex interactions between aging, genetics, epigenetics, and environmental factors. Recent Genome Wide Association Studies have identified 90 risk loci and 305 nearby candidate genes that might have a significant genetic component to sporadic PD, however, the specific roles of these genes in PD pathophysiology remain to be confirmed and elucidated biologically. In this proposal, I will investigate the hypothesis that one of these genes, CTSB, plays a critical role in PD. CTSB encodes cathepsin B, a lysosomal enzyme implicated in various neurological disorders, with evidence suggesting its downregulation may be associated with PD. I will utilize a CRISPR-based functional toolkit in mouse models to downregulate Ctsb specifically in DA neurons and examine its effects on neuronal survival, morphology, physiology, and behavior over time. To achieve this, I will employ a Cre-dependent dCas9-KRAB (CRISPRi) system in combination with a DAT-IRES-Cre mouse line, allowing for targeted downregulation of Ctsb in midbrain DA neurons via intracranial injection of a Ctsb sgRNA compared to a scrambled non-targeting control sgRNA. I have already shown co-localization of Ctsb and Th mRNA in midbrain DA neurons and successful Ctsb downregulation in primary neuronal cultures. This study aims to determine whether Ctsb downregulation reveals cell autonomous alterations in dopaminergic neuronal function and behavior at 3, 6, and 12 months of age. These experiments will provide insights into the role of Ctsb in sporadic PD, potentially identifying it as a true risk gene. Understanding how Ctsb affects DA neurons may lead to the development of treatments that address the underlying pathology of PD, offering a significant advancement over current therapies that primarily manage symptoms without altering disease progression.

NIH Research Projects · FY 2026 · 2026-02

PROJECT SUMMARY Information in the brain is encoded in the membrane potential variations of neurons. To understand the neural substrates of complex behaviors we need advanced methods that can capture such variations across large volumes, with high precision and in animals behaving naturally. Voltage imaging techniques are emerging as powerful tools towards these goals. They rely on the use of voltage sensitive molecules, that change their light emission properties upon changes in transmembrane voltage, coupled with imaging devices capable of collecting these localized signals. However, since physiologically relevant voltage signals can be extremely fast and distributed across large volumes, standard frame-based imaging sensors struggle to faithfully acquire them at sufficient temporal resolution and across large fields-of-view (FOVs). Moreover, the typical data bandwidth during voltage imaging with conventional sensors is hardly compatible with wireless applications. Here, to overcome these limitations, we propose to use a radically new type of sensors: event-based cameras (ECs). ECs are inspired by biological systems: rather than capturing images at a fixed rate, they asynchronously detect changes in brightness at each pixel, generating a continuous stream of events – analogous to action potentials - that encode the time, location, and polarity of these changes. Compared to conventional cameras, ECs have exceptional temporal resolution, low latency (both at the µs scale), extensive dynamic range (~120 dB), minimal power consumption (few mW) and reduced data output. However, the potential of ECs for imaging biologically relevant signals, such as calcium or voltage in neurons, remains virtually unexplored, as new methods are needed to acquire and process their output and embed ECs in current microscopes. Our proposal aims at pioneering the use of ECs for ultrafast voltage imaging across large FOVs. We will first investigate the capabilities of ECs for recording single action potentials across large FOVs (1 mm2) in neuronal cultures expressing genetically encoded voltage indicators (GEVIs). Second, we will set up and optimize the methods for imaging voltage signals with ECs in vivo, in the cortex and hippocampus of head-immobilized bats. Third, we will embed an ultra-compact event-based sensor into an existing miniaturized microscope for wireless one-photon voltage imaging. We will showcase the applicability of this new tool for untethered imaging by monitoring neural activity in the hippocampus of freely flying Egyptian fruit bats. Supporting the feasibility of this proposal, we have extensive expertise in wireless neural recordings in freely flying bats and we have collected preliminary data suggesting that ECs can capture voltage signals in vitro. Despite the highly ambitious nature of this project, we believe it will make several groundbreaking contributions to the rapidly expanding fields of event-based sensing and biomedical fluorescence imaging.

NSF Awards · FY 2026 · 2026-02

With support from the Chemical Measurement and Imaging Program in the Division of Chemistry, Professor Evan Williams of the University of California, Berkeley, with his group, will investigate the chemistry and physics of charged droplets and the chemistry that occurs within evaporating droplets. Charged water droplets are ubiquitous in the environment, where they are generated by waves, waterfalls, and in thunderclouds, and have significant importance in industrial processes, including protective coatings, microencapsulation to protect pharmaceutical compounds, and generation of nanoparticles. Electrospray ionization produces highly charged droplets that are used in many of these processes, including as thrusters for positioning spacecraft. It is also an essential analytical method that is used in thousands of laboratories worldwide when combined with mass spectrometry for detailed chemical analysis of complex mixtures, including environmental, synthetic, and biological materials. The project goals are to understand processes that occur with highly charged droplets. This includes droplet fission, whereby droplets spontaneously break up into smaller charged droplets. Experiments aimed at understanding how crystallization occurs in evaporating nanodrops will be pursued to better understand factors that affect both the rates and structures of the initially formed aggregates. Electrospray ionization at reduced pressure will be explored to better understand how droplets are initially formed and what factors affect their initial sizes and charges. This research will take advantage of unique charge detection mass spectrometers that have been developed in the laboratory to mass analyze individual nanoparticles and will likely lead to improved analytical methods, a better understanding of aerosol chemistry that affects the environment and human health, as well as the potential to improve the efficiency of spacecraft thrusters. Students involved in this research learn important skills in chemical analysis, solving complex problems and how to communicate results. These skills are important for the biotechnology, pharmaceutical, and chemical industries. The Williams Lab is studying the chemistry and physics of charged nanodrops with diameters between ~10 nanometers and 1 micron formed using a variety of different methods, including electrospray ionization. Analysis is performed using custom-designed and constructed charge detection mass spectrometers capable of measuring analytes and droplets with masses ranging from a few thousand Daltons to GigaDaltons. Investigations will include measuring the fission processes and dynamics of charged droplets using unique analysis methods to track individual droplet mass and charge that will be developed as part of this project. Effects of solvents and constituents inside the droplets will be investigated to learn how gaseous ions are formed. A reduced-pressure electrospray apparatus will be constructed and evaluated for the potential to extend the current mass range of these instruments and to investigate the initial charged droplet formation process to better understand factors that affect ionization efficiency and selectivity in mass spectrometry. This may also lead to more efficient electrospray-based thrusters. The ability to analyze highly heterogeneous mixtures at high molecular mass makes it possible to investigate the early onset of cluster formation in crystallization. A better understanding of equilibria governing solubility and crystallization has important implications for industrial-scale chemical synthesis. 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 · 2026-01

Abstract Successful navigation requires that our brains continuously maintain and update a sense of our orientation in space. This internal sense, called head direction (HD), is computed in dedicated neural circuits found across diverse organisms. The network mediating this representation is “dynamically stable.” In new environments, rapid synaptic plasticity between sensory neurons and HD neurons enables the HD representation to be updated by available sensory cues, so that the animal will know the direction that it is facing based on its view of the current surroundings. Conversely, the HD network must also be able to stably hold the heading representation even when environmental cues are sparse, such as when the sun becomes obscured on a cloudy day. Local excitatory loops in the network support a “bump” of high activity that reinforces the location of the HD representation, while broad inhibition is required to ensure that only a subset of HD neurons is active at once making the HD representation unique and stable. In the Drosophila HD network, a class of sensory neurons called ring neurons have been proposed to play a role in both sensory plasticity and widespread inhibition. Ring neurons form inhibitory synapses onto all HD neurons and these synapses undergo associative plasticity to link heading angle with features in the environment, though the synaptic mechanisms mediating this plasticity are not understood. My preliminary data has revealed that a subset of ring neurons that respond to visual cues fire large voltage events called “bursts” in addition to sodium spikes. I hypothesize that visual ring neuron bursts and spikes encode for strong or weak visual inputs, respectively, and that bursts are required for plasticity to link HD representation to visual cues while spikes are important for global inhibition to stabilize the network. I will address this hypothesis using cell-specific molecular perturbations, whole-cell patch-clamp electrophysiology, and in vivo voltage or calcium imaging during the presentation of visual stimuli. This project presents a rare opportunity to gain a molecular-level understanding of parallel neural codes and its importance in HD network dynamics. This work will inform how visual input entering the HD network are transformed by ring neuron spikes and bursts. I anticipate that the mechanism I discover will inform models of how HD networks in invertebrate and vertebrate systems balance flexibility and stability.

NIH Research Projects · FY 2025 · 2026-01

Project Summary Organic synthesis is a rate-limiting factor in drug discovery, and the chemical space that can be accessed by medicinal chemists is limited by the synthetic methods available to access targets of interest. The development of novel synthetic methods for constructing organic molecules with distinct architectures has the potential to enhance human health by enabling the synthesis and discovery of new small molecule drugs. Supramolecular host catalysis has proven to be a unique synthetic approach which can enable novel chemical reactions that are not observed in bulk solution. In this approach, substrates encapsulated within supramolecular hosts can engage in host-guest binding interactions that induce distinct reactivity and product selectivity in both intra- and intermolecular transformations. Significantly, supramolecular host-guest binding has been found to promote size-, site-, regio-, and enantioselective reactions. Building off work by the Toste, Raymond, and Bergman groups studying the reactivity of supramolecular host assemblies, the objective of this proposal is to design and develop supramolecular hosts that function as photocatalysts to induce novel photochemical reactions that afford access to oxygen- and nitrogen-containing heterocycles in a highly selective manner. Prior work by the Toste lab has established that stoichiometric quantities of a dodecanionic GaIII-centered tetrahedral host assembly can engage in photoinduced single-electron transfer events with encapsulated substrates to enable a radical rearrangement reaction. This established photoactivity of tetrahedral host assemblies will be explored in two orthogonal approaches for the development of new host-catalyzed photochemical reactions. The first approach develops supramolecular photocatalysts that perform single-electron transfer to encapsulated substrates. Through the design of supramolecular assemblies with modulated redox potentials and tailored microenvironments, new supramolecular hosts will be developed and applied as photocatalysts to facilitate radical cyclization reactions that afford oxygen- and nitrogen-based heterocycles. The second application will apply pyrene-based supramolecular hosts as photocatalysts to perform triplet energy transfer to azides, generating encapsulated triplet nitrene intermediates. Through coencapsulation of triplet nitrenes with suitable coupling partners, these hosts will be applied to promote intermolecular reactions of nitrenes to afford nitrogen- based heterocycles and other nitrogenated molecules with chemoselectivity and site selectivity. As nitrogen- and oxygen-based heterocycles are privileged scaffolds in medicinal chemistry, it is expected that these proposed methods will have a positive impact on human health by aiding chemists in the synthesis of new small molecules for drug discovery efforts.

NSF Awards · FY 2026 · 2026-01

Grasses are beneficial to human society by creating habitat for bees, and other pollinators, that ensure crops produce fruit and seeds, improve the quality of water, trap carbon, and provide food for animals upon which people rely for nutrition. However, grasses are a large group of over 11,000 species, which cover ~50% of the earth’s surface, and there are differences in their ability to perform beneficial ecological functions. For example, they can differ in the time of year they grow (also called phenology), how fast they grow, and their ability to tolerate and survive droughts. Importantly, there are often trade-offs between these characteristics, such that species that only grow in the spring may not be very tolerant of drought, and species that only grow during the summer generally do not grow very fast. Therefore, environmental changes during different seasons may prevent some species from thriving in their current locations, altering the ecosystem services they provide. To effectively manage resilient grasslands for the future, we need better information on the phenology, growth rates, and drought tolerance of a broader range of grass species. Most projects that measure these characteristics have focused on trees, leaving major gaps in our understanding of these traits in grasses. Using novel techniques to observe processes occurring inside the leaf and new mapping methods, our project will provide critical information about plant traits and tradeoffs in different environments to help predict how grass distributions will respond to changing weather patterns and environmental conditions. Changes to plant communities are continually occurring as plants disappear, appear, and re-arrange in ecosystems across the globe as rising temperatures and changing precipitation patterns reduce the available water for plant growth. Plant responses to these dynamic conditions dictate whether a species can persist in a region or must shift distributionally. Modern approaches to modeling species distributions rarely include the mechanistic underpinnings of organismal responses but, instead, rely on bivariate relationships between individual traits and annual summaries of abiotic conditions. This approach ignores the fact that networks of traits, rather than any single trait, generate different drought-coping strategies and that drastic differences in grass phenology decouples plant growth conditions from annual summaries of abiotic conditions. To improve predictions of future species distributions and inform restoration projects of ideal seed-mixes, the overall objective of our study is to improve the accuracy of species distribution models through a better understanding of grass species resilience by including trait networks and growth phenology. Using a set of species that spans the entire grass family, the investigators will identify mechanistic trait networks leading to different drought-coping strategies, including mechanisms leading to embolism formation, a key drought-coping trait rarely studied in grasses. Integrating these key traits will provide information on species responses and distribution shifts and the experimental design will also provide information on how these traits may evolve independently or in unison within the grass family. 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 · 2026-01

PROJECT SUMMARY Neural circuits derive most of their computational power from inhibitory interneurons, which balance excitation, enhance contrast, and provide temporal-spatial precision. In the retina, amacrine cells (ACs) are the largest class of inhibitory interneurons, providing critical inhibition and modulation to bipolar cells and ganglion cells through their release of GABA or glycine. ACs are incredibly heterogeneous, comprising as many as 60 types that perform distinct computations and expand the repertoire of visual features relayed to the brain. However, over 80% of these AC types have not been functionally characterized. This is largely due to a lack of molecular markers to genetically target these types for morphological and physiological experiments in animal models. Furthermore, since AC types are also morphologically complex, they often lack clear orthologs in other species, making it difficult to identify their conserved biological function. The overarching objective of this proposal is to map AC types across species and identify conserved molecular markers with which to target these types for functional experiments. Recent work in our lab showed that bipolar and ganglion cell types in the retina are highly conserved across mammals. Although this study omitted AC types due to their complexity, our preliminary computational analysis suggests that AC types can be molecularly aligned between primates and rodents. Consistent with these findings, more recent experimental and genomic studies revealed that the direction- selective and scotopic circuits – key AC-driven circuits – are present in fish. Based on these observations, we hypothesize that AC diversity is evolutionarily ancient and that AC types are broadly conserved across vertebrates. To test this hypothesis, we will use single-cell RNA-sequencing atlases of ACs recently generated in our lab to guide a series of computational and experimental studies. Aim 1 will investigate the molecular conservation of ACs across 20 vertebrates by aligning single-cell atlases and building upon existing comparative transcriptomic methods. Aim 2 will identify novel, conserved molecular markers of AC types and label them in primates, rodents, and chicken using histology. Experiments in Aim 3 will investigate the circuitry and function of a highly conserved, novel AC type marked by PDGFRA. Together, these data will describe the molecular and structural conservation level of ACs and reveal new aspects about their cellular properties. Importantly, we will identify and test novel molecular markers of specific AC types that will be useful for targeting them in future studies. This proposed work has broad implications for uncovering contributions of inhibitory interneurons to retinal processing.

NSF Awards · FY 2026 · 2026-01

Modern society depends on analyzing massive amounts of personal information to improve healthcare, enhance national security, and drive economic growth. However, current practices for storing and sharing sensitive data have led to major data breaches exposing millions of people's personal information, including medical records, financial details, and government secrets, undermining public trust and threatening national security. This project addresses this critical challenge by developing new computer systems that allow organizations to gain insights from large datasets while keeping individual information completely private. This project will bring privacy protections to each of the three steps of the data-management lifecycle: data collection, data processing, and data retrieval. By protecting privacy during data collection, analysis, and retrieval, this research serves the national interest by enabling continued technological advancement while safeguarding citizen privacy, supporting economic competitiveness in data-driven industries, and strengthening cybersecurity infrastructure. This project investigates three fundamental research areas to advance privacy-preserving data systems. First, the research team will develop new protocols for privacy-preserving data collection that enable servers to compute aggregate statistics over client data without accessing individual records, with emphasis on reducing computational costs and expanding the class of computable functions compared to existing systems. Second, the project will design privacy-preserving machine learning algorithms for training recommender systems, clustering algorithms, and decision trees that operate on encrypted data while maintaining model accuracy. The third and last component of the project will be to develop new techniques that let clients privately query server-side datasets. In this thrust, the project will develop a relational database that supports private queries. A key design component will be new data structures, optimized to work with cryptographic privacy-protecting protocols. 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.

CIHR Grants and Awards · FY 202526 · 2026-01

The global rise of multidrug-resistant bacteria has created an urgent demand for new treatments. Phage therapy, which uses viruses to neutralize bacteria, is a promising alternative to antibiotics; however, bacteria often resist phage infection using sophisticated immune systems, complicating the use of phage-based treatments. Despite their clinical importance, most bacterial immune systems remain poorly understood, particularly those prevalent in pathogens. Understanding these immune systems will uncover strategies that bacteria use to resist phage infection, leading to new clinical solutions to combat antibiotic resistance. Furthermore, these insights may reveal innovative molecular tools that can be repurposed to advance human health. My research focuses on a novel broad-spectrum bacterial immune system called Hachiman, which is widespread in pathogenic strains of bacteria including Escherichia coli and Bacillus cereus. Our prior studies have shown that Hachiman triggers antiviral immunity by recognizing host DNA damage, an immune checkpoint conserved across domains of life. This breakthrough has opened new research directions on exploring the unexpected diversity of Hachiman systems and their bioengineering applications. Specifically, I will identify and characterize functionally diverse Hachiman systems found in bacterial pathogens that possess new anti-phage immune mechanisms. Furthermore, I will leverage Hachiman to engineer a minimal genome editing platform enabling improved high-throughput functional analysis of non-coding regions in the human genome, critical for understanding gene regulation and disease. Successful implementation of this proposal will significantly enhance our understanding of bacterial immune systems, paving the way for improved phage therapies and novel molecular tools that can advance human health. Keywords: BACTERIAL IMMUNE SYSTEMS; BIOINFORMATICS; BIOCHEMISTRY; STRUCTURAL BIOLOGY; BIOENGINEERING; PROFILE HIDDEN MARKOV MODELS; CRYO-ELECTRON MICROSCOPY; STRUCTURE-GUIDED ENGINEERING; BACTERIOPHAGE; PHAGE THERAPY

NIH Research Projects · FY 2025 · 2026-01

PROJECT SUMMARY/ABSTRACT: This proposal aims to address a critical challenge in treating anaplastic lymphoma kinase (ALK)-driven neuroblastoma (NB) and other cancers: namely the development of resistance to targeted therapies, particularly tyrosine kinase inhibitors (TKIs). ALK is a receptor-tyrosine kinase that plays a crucial role in the development of the central and peripheral nervous systems, and its dysregulation, through mutations, fusions, and overexpression, has been implicated in several cancers, most notably NB. Despite the success of TKIs in treating ALK-driven cancers, resistance mutations in ALK often emerge, leading to disease relapse and limited long-term efficacy. One promising alternative approach to bypass resistance is targeted protein degradation through PROteolysis TArgeting Chimeras (PROTACs), small molecules that harness the ubiquitin-proteasome system to degrade specific target proteins. However, the current landscape of protein degradation therapies is limited to the near-exclusive use of two E3 ubiquitin ligases – CRBN and VHL – for PROTAC development. These ligases, while effective for certain targets, may not be optimal for all targets, and the emergence of resistance to these degrader-based therapies highlights an overlap in limitations between TKIs and PROTACs that necessitates recruitment of more diverse and effective degraders. Additionally, while at least a third of the proteome transits the endoplasmic reticulum (ER) through its membrane or lumen, almost all current degraders target cytosolic proteins leaving a large swathe of the proteome unexplored. Given ALK mutant proteins in NB often mislocalize to the ER, where they become constitutively activated and contribute to oncogenic signaling, the ER presents an untapped resource for therapeutic development through targeted protein degradation. This project intends to leverage ER-associated degradation (ERAD), a posttranslational quality control system within the ER that aids in protein homeostasis by degrading misfolded and immature proteins, to target mutant ALK. The ERAD system, which consists of membrane-embedded E3 ligases and their associated cofactors, provides a unique platform and specificity to target a previously under-explored pool of disease- related proteins, including ALK mutants. The research strategy will focus on several key innovations: (1) identifying selective protein degraders that can efficiently target ALK mutants and overcome resistance mutations, (2) expanding the therapeutic landscape of targeted protein degradation through ORFeome-based screening approaches in search novel target-degrader pairs, (3) developing genetic proof-of-principle tools to rapidly test and validate these novel degradation strategies, and (4) leveraging the specificity of organellar- based degradation to selectively degrade disease-relevant proteins localized to the ER. Ultimately, this research seeks to develop a more diverse set of tools for targeted degradation therapies, offering promising new strategies for overcoming drug resistance in ALK-driven NB and improving clinical outcomes in cancer treatment as a whole.

NSF Awards · FY 2026 · 2026-01

In this project, the researchers will use genomics, genetics, and cell biology to understand the function of coral genes during heat stress. Thermal stress often causes the “bleaching” (whitening) of corals that is harming reefs worldwide and is a threat to marine biodiversity. Genes are the functional unit of DNA and how the coral reacts to heat stress is called its phenotype. There are many genes involved in achieving the heat response phenotype and they will be studied in this project. This is important because the function of many of these and other genes and how they influence the phenotype of corals are not understood. In addition, many of the coral stress genes are unknown in other organisms (i.e., they arose in corals), and therefore present the opportunity to discover the function of novel genes. This research is important because it will help in the understanding of how both known and novel genes function, which can aid in the conservation of coral reefs- a resource that is important for national security and economic growth. Lastly, this project will fund early career scientists and lead to broader impacts through the training and education of a diverse workforce in science. The overarching goal of this proposal is to determine the mechanistic drivers of the stress response in multicellular organisms, with a focus on the ecologically important corals. To achieve this goal, application of an end-to-end approach that extends from genes to phenotypes, to cell biology, using multi-omics and network methods with both holobiont and single cell data will be applied. This work will elucidate the coral stress phenome in species with differing reproductive strategies and geographical origins to determine the functions of both known and unknown (dark) genes. Generating this knowledge base will transform coral biology and more broadly, provide a research platform for testing hypotheses about the resilience of other ecosystems, marine and terrestrial, which is particularly important under accelerating climate change. This project aims to increase the understanding and predictive capability of how key properties of living systems emerge from the interaction of genomes. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2026 · 2026-01

To use a web service today, Internet users must often reveal their private information to the web-service provider. For example, Internet users upload their photographs to online photo albums, reveal their interests to web-search engines, and disclose their favorite websites to Internet service providers. Sending sensitive data to web-service providers is a serious privacy risk: the provider could lose the user's data in a data breach, decide to sell it later on, or be forced to disclose it to a foreign government. At the same time, web services are indispensable. Thus, computer users currently have no choice but to hand over their sensitive data to web-service providers and to suffer the accompanying privacy risks. This project will develop a new suite of privacy-protecting web services that never see or process any unencrypted user data. This project's goal is to make it possible for everyday Internet users to enjoy the tremendous benefits of today's web while shielding them from the accompanying privacy risks. In addition, the educational aspects of the project will focus on the development of undergraduate content and an openly available textbook to teach security and systems in tandem. This project consists of three parts, each dedicated to the development of a different private web service. The first part focuses on private machine-learning inference: allowing a client to evaluate a large server-side machine-learning model on its private data (e.g., the client's photos) while revealing no information about the client's private input data to the machine-learning service. The second part focuses on private search: allowing a client to search over a server-side corpus of billions of documents (e.g., web pages) while revealing no information about its search query to the search engine's servers. The third part focuses on private web browsing: allowing a client to browse a web of hundreds of millions of text-based pages while revealing no information about which pages it is reading. Building each of these three private web services will require new technical tools. In particular, this project will develop a suite of new low-level cryptographic primitives, including new high-speed protocols for private matrix multiplication, new protocols for private nearest-neighbor search in high-dimensional vector spaces, and a new cryptographic primitive, distributional private information retrieval, which allows a client to privately fetch data from a remote database server at relatively low cost. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2026 · 2026-01

This award funds a research project examining the factors influencing exchange rates, with a focus on international asset prices and capital flows. Understanding the drivers of exchange rates is a central objective of international macroeconomics and finance. Exchange rates are key prices affecting the relative cost of imports and exports, as well as the returns to investing in assets abroad. Yet exchange rates have long puzzled economists: they are inconsistent with benchmark models in economics and are very difficult to predict. This project makes progress on these puzzles by exploring the relationship between exchange rates and other international asset prices, such as bonds and stocks in different countries, as well as macroeconomic quantities, such as capital flows across borders. This research helps market participants use exchange rates to better understand macroeconomic conditions in the global economy. It also helps decision-makers better understand how monetary and fiscal policies affect exchange rates and thus the broader economy. The research findings could inspire innovative international policy frameworks, improve the well-being of households, businesses, and investors, and reinforce global market activity. This award project explores exchange rates through the lens of general equilibrium models with international capital market frictions. These capital market frictions are reflected in a central role for global arbitrageurs to intermediate capital flows across borders. Such models thus allow the researchers to accommodate classic forces that have been thought to drive exchange rates, such as the supply or demand for goods, as well as drivers that have been proposed in more recent research, such as changes in the intermediation capacity of global arbitrageurs. The project includes research to decompose the drivers of the dollar exchange rate versus other advanced economies through the lens of such a model. The co-movements of exchange rates with bond yields, spreads in financial markets, and capital flows suggest a nuanced variance decomposition of exchange rates: at high frequencies, shocks to intermediation capacity play an important role in driving exchange rates, but at low frequencies and overall, demand shocks play the most important role. The project includes further research to understand the sources of demand shocks driving variation in these prices and capital flows. The researchers study the co-movements of exchange rates, bonds, and stocks to differentiate between potential sources of demand shocks. The project also includes research to understand how heterogeneous exposures to demand or intermediation shocks can explain differences in exchange rate behavior across countries. The researchers study how patterns in trade linkages and in net foreign assets give rise to heterogeneous effects of these shocks. The project finally includes research on the propagation of macroeconomic policies through exchange rates. The researchers study whether state-dependence in the effects of foreign exchange rate intervention can account for mixed evidence on its effectiveness in the data. This research could help researchers and decision-makers gain deeper insights into exchange rates and their relationship with international asset prices, capital flows, and other macroeconomic factors, ultimately improving the well-being of the U.S. population and strengthening the country's leadership in international markets. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Awards · FY 2026 · 2026-01

This project will contribute to the national need for well-educated scientists, mathematicians, engineers, and technicians by supporting the retention and graduation of high-achieving, low-income students with demonstrated financial need at the University of California, Berkeley. A total of 32 scholars pursuing Bachelor of Arts degrees in Astrophysics, Physics, Mathematics, Applied Mathematics, Atmospheric Science, Environmental Earth Science, Geology, Geophysics, Marine Science, and Planetary Science will receive scholarships averaging $11,200 per year for up to five years. Scholars will receive mentoring from faculty, graduate students, peers, and alumni, and the project will build strong scholar cohorts through orientation events and resources for new students, events focused on skill-building for academic and career advancement, and participation in research and internship opportunities. Additional activities for scholars include career seminars, coaching, and leadership training to develop students' skills and professional networks. The overall goal of this Track 2 Scholarships in STEM project is to increase STEM degree completion of academically talented, low-income undergraduates with demonstrated financial need. There is a significant national need to grow the STEM workforce and nurture key talent that will ensure economic competitiveness and provide domestic leadership across critical sectors. This project directly speaks to this need by supporting STEM student success, which will strengthen the workforce in software development, data science, cybersecurity, actuarial science, nuclear energy, and other key areas of need. The project will be assessed by an experienced evaluator who will examine participant outcomes, including retention, graduation, academic performance, and preparation for graduate study and STEM careers, while also tracking changes in students' self-efficacy, disciplinary identity, and career aspirations and outcomes. The data generated will contribute to the knowledge base regarding effective strategies to support talented, low-income students in STEM. This project is funded by NSF's Scholarships in Science, Technology, Engineering, and Mathematics program, which seeks to increase the number of academically talented, low-income students with demonstrated financial need who earn degrees in STEM fields. It also aims to improve the education of future STEM workers and to generate knowledge about academic success, retention, transfer, graduation, and academic/career pathways of low-income 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 2026 · 2026-01

As autonomous systems like self-driving cars and delivery robots become more common, making sure that they operate safely is critical. These systems often learn how to act using reinforcement learning (RL). While powerful, RL methods typically do not guarantee safety, which limits their use in the real world. A common approach towards capturing safety in RL is ensuring the satisfaction of safety constraints. However, verifying that a learned RL controller never leads to any constraint violation is in general a nontrivial problem due to the black-box nature of such controllers. In this project, we will develop new techniques that allow autonomous systems to learn effective behaviors while always respecting strict safety constraints. This research tackles a core challenge in making intelligent machines reliable and trustworthy. As such, the results of this project could impact areas like transportation, aerospace, and healthcare, where safety is non-negotiable. We will develop a formal framework for learning control policies that provably satisfy hard safety constraints, even when the system dynamics are unknown or treated as black boxes. We will develop RL-based methods that guarantee constraint satisfaction throughout training and deployment in three research thrusts. Our approach begins with the challenge of 1) enforcing a single affine constraint of relative degree one, where we embed an appropriate structure into the policy network to ensure constraint satisfaction by design. 2) We will then extend this framework to handle multiple constraints and more general nonlinear safety specifications by lifting them into an augmented state space. 3) Finally, we will expand the methodology to handle constraints with higher relative degrees, which require greater anticipatory control, and explore how to generalize the approach to cyber-physical systems with hybrid or non-smooth dynamics. This progression allows us to systematically tackle increasingly complex safety requirements, enabling practical and reliable RL deployment in real-world 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.

NIH Research Projects · FY 2025 · 2025-12

PROJECT SUMMARY/ABSTRACT Enzymes are the primary functional molecules in cells, providing enormous rate enhancements, specificity and regulation to the diverse chemical reactions that are necessary for life. Enzymes, like all biological macromolecules, are the products of evolution: all enzymes have evolved to operate within the complex environment of the organism/cell in specific environmental niches(s). Thus, an understanding of enzyme function and evolution is fundamental to biology. Enzymes also have tremendous potential in medicine (e.g., as targets for anti-cancer, antimicrobial and antiviral drugs and as therapeutics for metabolic disorders) and in industry (e.g. to make important commodity chemicals and as catalysts for bioremediation). Our central premise is that a quantitative, mechanistic understanding of enzyme function and its relationship to organism fitness is critically needed to precisely manipulate enzymes and to deeply understand biology. To generate this level of understanding, we need: (1) a quantitative, chemical, and physical knowledge of enzyme function, and (2) mechanistic data describing how and when these physical principles contribute to enzyme function within the complex environments where enzymes operate. An enhanced understanding of the relationships between protein sequence, protein function and cellular/organismal fitness will have profound impacts across biology and medicine, from improving our ability to predict how mutations will influence the virulence and drug susceptibility of human pathogens, to enhancing precision medicine by accurately predicting the consequences of allelic variants, to enabling the design of next-generation protein and cellular therapeutics. Achieving this understanding requires new tools and a new conceptual paradigm. Enzymes are highly interconnected, their functions are multifaceted, and their cellular environments are complex. Traditional biochemistry is enormously powerful, allowing for the intensive study of a few individual enzymes in vitro (10s) and providing detailed knowledge of their chemical mechanisms. But identifying the many residues that matter for enzyme function requires investigation of residues beyond the active site at a scale far beyond that of traditional biochemistry. Furthermore, this biochemical information then needs to be translated to organism fitness in vivo in a quantitative manner. Here we will overcome these challenges. We will first use evolutionary sequence information to direct enzyme variant design towards functionally important areas of sequence space. We will adapt high-throughput microfluidic technologies to quantitively measure the biochemical properties (e.g., kcat, Km, Ki, and ∆GFold) of this library of 104 enzyme variants in vitro (Aim 1). Then we will determine how each of these variants affects organismal fitness in vivo using pooled competition and barcode sequencing assays (Aim 2). Finally, we will use this sequence-function-fitness map to test long-standing models in biochemistry and evolution and reveal the biochemical determinants of fitness important for industry and medicine (Aim 3). Such a comprehensive and quantitative mapping of biochemical function to fitness has never been achieved.