Grants

NIDA - National Institute on Drug Abuse

PROJECT SUMMARY Vocalizations and other bioacoustic signals convey an individual’s identity, location, and behavioral state and these sounds are used to guide social-acoustic interactions. While much stellar research has been done on the neural basis of vocal communication, it has largely focused on brief, stereotyped interactions between unfamiliar adults in featureless test cages. Here, we propose to study auditory communication within a cohesive social unit, the family, of a highly social species, Mongolian gerbils, in a stable, naturalistic setting. By integrating audio, video, and neural data, we aim to uncover the behavioral meaning of distinctive vocal bouts and test how cortical mechanisms support the use of these vocalizations to make social decisions. In Aim 1, we will create and disseminate general-purpose machine learning tools for studying auditory communication within family groups living in undisturbed naturalistic environments. Aim 1A will establish multi-animal 3D body pose tracking in complex spacious environments. Aim 1B will develop deep learning methods to noninvasively localize and attribute vocalizations to specific family members, focusing on the multi-second vocal bouts that dominate auditory communication. Aim 1C will integrate body pose, bioacoustic signals, and neural data with a 3D environment model using Gaussian splatting, enabling us to estimate how auditory signals are combined with line-of-sight information. In Aim 1D, we collaborate with a team at Princeton to extend and validate sound attribution methods in a different species, environment, and behavioral paradigm (mouse courtship) to establish general-purpose and open-source tools for the field. Aim 2 develops a machine learning approach to extract low- dimensional behavioral descriptors from this large data set that are used as variables in our predictive models. In Aim 2A, we extract recurring bouts of acoustically driven behavior using a mixture of expert-supervised and data-driven, unsupervised approaches to annotate continuous streams of natural behavior. Computational innovations are proposed, and we employ playback and hearing attenuation manipulations to test causal relationships between vocal bouts and behavior. In Aim 2B, we use these latent features and contextual variables to build linear Bayesian models that predict future actions, permitting us to ask how individual behavioral traits are explained by social (e.g., kinship, sex, age) and environmental factors. Aim 2C will characterize how auditory cortex activity represents vocal bouts, social and environmental factors, and subsequent behavioral decisions. Aim 3 will test the cortical network mechanisms that give rise to social and contextual modulation during auditory communication. In Aim 3A we will silence auditory or frontal cortices, as well as projections between the two, and measure how these perturbations influence behavior. In Aim 3B, we will record wirelessly from frontal cortex during sound-driven social interactions.

NIGMS - National Institute of General Medical Sciences

Abstract NMR is one of the most versatile analytic tools for investigating biomolecules. It can probe structure and dynamics in solution at atomic resolution, even for disordered molecules, and it can quantify the affinity and kinetics of biomolecular interactions and the sites of interaction. It is capable of quantifying individual components of complex mixtures of small molecules, with applications to metabolomics, drug discovery, and clinical diagnostics. The broad goal of this project is to develop computational methods and tools that enhance the application of NMR in these biomedical applications, by improving sensitivity, resolution, lowering the barriers to implementing complex experimental and analytic workflows, and making them more reproducible and the resulting data easier to survey and share. The approach will leverage and enhance the open, international repository for biomolecular NMR data, the Biological Magnetic Resonance Data Bank, and the NMRbox computational platform, comprised of >250 NMR-related software packages and high performance computational and storage resources. Special emphasis is given to the application of NMR for investigating dynamics and disorder in biological macromolecules, which are inaccessible to other methods, and to the quantitative analysis of mixtures of small molecules, for NMR-based screening, metabolomics, and biomarker discovery.

U.S. National Science Foundation

Computational and Data-Enabled Science and Engineering

NHGRI - National Human Genome Research Institute

Computational and experimental approaches to decode domain-specific protein-RNA interactions Project Summary RNA-binding proteins (RBPs) play central roles in post-transcriptional gene regulation by diversifying the types and levels of protein products expressed in specific cellular contexts. This is achieved through interactions of RBPs with specific sequence or structural elements in their target transcripts. Disruption of these regulatory elements accounts for a substantial fraction of human disease associations. However, since most of these elements are embedded in the noncoding genomic regions, they are currently annotated poorly in the human genome. While CLIP-seq and its many variants can map RBP binding footprints on a genome-wide scale, such maps remain sparse in coverage concerning both the number of RBPs and cell types. Predictive computational models can potentially complement experimental data and provide powerful tools to interpret the functional impact of genetic variants, but the success of this approach is still limited. We realize that a major challenge in the precise mapping and prediction of protein-RNA interactions is a critical lack of technologies that can delineate the specificity of individual RNA-binding domains (RBDs) of multi-domain RBPs, which account for about half of all RBPs in humans. Since the current CLIP methods pull down all RNA fragments crosslinked to any RBDs in a mixed population, with each individual RBD recognizing short and degenerate motif sites, we are unable to deconvolute the binding sites of individual RBDs. Such resolution is required to precisely understand the mechanisms conferring the specificity of protein-RNA interactions and interpret the functional significance of genetic variants. In this study, we propose two complementary strategies to overcome the fundamental challenge and develop new methods that will enable one to map domain-specific protein-RNA interactions in the native cellular context, at single-nucleotide resolution, on a genome-wide scale. This project builds on the tight integration of complementary expertise of the Zhang lab in RNA and computational biology and Wang lab in chemical biology. If successful, this study will produce platform technologies that will find impactful applications in studies of gene expression regulation, genotype-phenotype relationships, and development of RNA-based precision genetic medicine.

NHGRI - National Human Genome Research Institute

PROJECT SUMMARY ABSTRACT Genome-wide association studies (GWAS) have been remarkably successful at mapping genomic loci for complex human diseases. However, most studies use case-control designs based on disease occurrence at a specific time point, such as diagnosis. overlooking genetic factors that influence the course of disease over time, including initiation, onset, progression, severity, and therapeutic response. Time-to-event (TTE) phenotypes capture both the occurrence and timing of disease activity. GWAS of TTE phenotypes can identify genetic variants associated with disease onset and progression, providing insights for early prevention and therapies aimed at halting disease progression. Large biobanks like UK Biobank, All of Us, and FinnGen, which combine longitudinal electronic health records (EHR) with genomic data, offer unprecedented opportunities to study genetics of TTE phenotypes. However, limitations remain in current computational and statistical tools, especially for rare variants and admixed populations. Additionally, biobank heterogeneity— differences in sampling strategies, follow-up times, and baseline hazards—poses challenges for meta-analysis of GWAS on TTE phenotypes. This project aims to overcome these barriers by developing innovative computational tools and statistical methods for GWAS of TTE phenotypes in large biobanks and cohorts. Aim 1 focuses on creating scalable methods for rare variant association tests on TTE endpoints, accounting for sample relatedness, population substructure, and high censoring. Aim 2 develops novel approaches to include admixed individuals in GWAS of disease onset and progression over time, enhancing inclusivity and reducing disparities in genetic discovery. Aim 3 improves GWAS and meta-analysis methodologies for TTE endpoints by addressing biases like left censoring, sampling bias, and collider bias, enabling robust integration of data across biobanks. The proposed methods will be evaluated through extensive simulation studies and applied to multiple biobanks. Successful completion of this project will provide new critical tools to advance our understanding of the genetic basis of complex diseases, reduce health disparities, and fully harness the potential of biobank resources for uncovering genetic factors influencing disease onset, progression, and treatment response. These tools and results will be made available as open-source software and public datasets to ensure broad accessibility to the research community. Additionally, we will continue to develop, distribute, and support open- source software packages for the proposed methods.

NIGMS - National Institute of General Medical Sciences

Proteins and their interaction networks play crucial roles in virtually all cellular processes. As such, knowledge of their functioning is essential for understanding the molecular basis of life. While high-throughput experimental methods are routinely used to characterize proteins and catalog their interactions, the overwhelming amount of genomic variation both within and across organisms—and in healthy and disease states—necessitates the development and use of efficient computational methods. Our overarching goal is to devise algorithms and machine learning methods that yield a predictive understanding of proteins and their specificities, interactions, and networks, and of how these attributes are altered by genetic variation. This is an especially exciting time to develop methods for analyzing protein sequences, as in recent years the field has been transformed by new artificial intelligence technologies, including protein language models that, akin to the progress in natural language processing, learn the “language” of proteins. In the next five years, we will combine these groundbreaking technologies with our years of domain expertise on proteins and their networks to tackle fundamental problems in three important areas. First, we will develop powerful new machine learning methods to predict DNA-binding specificities for broad classes of transcription factor proteins; such methods will newly enable the inference of regulatory interactions mediated by uncharacterized transcription factors or those mutated in disease, significantly advancing upon current work that is focused just on specific transcription factor families. Second, we will develop highly scalable, sequence- based approaches to predict the specific functional effects of variants within proteins; these approaches will be a great aid in interpreting the millions of coding variants observed across human populations and will be a step towards obtaining a mechanistic understanding of disease mutations. Third, we will develop novel algorithmic and machine learning approaches to predict the targets of kinase proteins and the pathways they regulate; these methods will elucidate kinase signaling networks and help decipher the growing body of complex phosphoproteomics datasets. A final, cross-cutting goal of our research is to rigorously evaluate current protein language models in order to uncover their strengths and limitations, and to develop innovative strategies to improve their capacity to capture the syntax, grammar, and semantics of protein sequences. We will release open source software for all developed methods.

National Institutes of Health

Computational Approaches to Curation at Scale for Biomedical Research Assets (R01 Clinical Trial Not Allowed)

NSF

With support from the Chemical Structure and Dynamics (CSD) program in the Division of Chemistry, Professors Evangelos Miliordos, Marcelo Kuroda, and Konstantin Klyukin of Auburn University are employing computational methods to a novel class of electronic materials known as solvated electron precursor electrides (SEPEs). Electrides are materials comprised of atomic or molecular layers with diffuse electrons occupying the interstitial space between the layers. These diffuse electrons could give rise to new electronic and magnetic properties that could be harnessed for a range of technologies, including advanced quantum computing, but their behavior is difficult to describe, making it difficult to connect it with the underlying molecular framework. Professor Miliordos, Kuroda, and Klyukin and their students will develop state-of-the-art computational simulations based on molecular and periodic approaches to model these diffuse electrons. Their work could clarify the structure-property-function relationships in these systems and advance the rational design and experimental realization of SEPE-based materials. The project will also provide research opportunities for graduate and undergraduate students and thus contribute to the development of STEM workforce. Solvated electron precursors are made of a positively charged metal centers fully coordinated by ligands which can sustain at least one peripheral diffuse electron, not bound to a specific atom or molecule. The aggregation of multiple SEPs leads to the formation of nanoparticles and materials known as SEPEs. The work of Professors Miliordos, Kuroda, and Klyukin will analyze the properties of structures where SEPs are connected via molecular bridges (linked-SEPEs) or are deposited on surfaces (surface immobilized SEPEs). The tunable composition of the electride-based material platforms offers enhanced control of the topological and electronic features owing to the variety of metals, ligands, materials, and linkers that can be used. Employing high-level multi-reference and density functional studies, this study will elucidate the underlying mechanisms defining their properties. This work will also provide design guidelines for a range of applications, including the development of chemical catalysts with low activation energy barriers, the reduction of overpotentials in electrochemical processes, the facilitation of multi-electron redox reactions, and the creation of qubits with extended coherence times and controlled entanglement. 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

NON-TECHNICAL SUMMARY A polymer is a long chain created by bonding together many building blocks known as monomers. Polymers are essential to daily life, from technologies such as plastics to biopolymers such as DNA. A diblock copolymer is created by bonding two chemically different polymers at their ends. In most cases, different polymers do not want to mix, so they phase separate similar to the way that oil and water separate in salad dressing. In contrast, block polymers cannot phase separate because the two blocks are bonded at their ends. Rather, block polymers spontaneously undergo a process known as “self-assembly” to create an ordered structure with nanometer-sized domains that minimizes the frustration created by their inability to phase separate. These domains can have different physical properties, e.g., regularly spaced, rubbery spherical inclusions in an otherwise glassy material, creating multifunctional materials. Self-assembly occurs in a wide range of soft materials, from soaps to biomaterials, and block polymers serve as an important model system to understand the physics that drives self-assembly. A long-standing goal in this area is predicting the ordered structure that will result from self-assembly based on the chemistry of the different blocks, the number of monomers in each block, and the temperature. Such a predictive ability would remove the need for tedious trial-and-error experimentation. The theory of block polymer self-assembly is remarkably powerful but suffers from a “chicken or the egg” conundrum: To perform the required calculations, one needs to know the structure that will be formed in advance. While this need poses no problems for understanding what has already been observed in experiments, it makes it extremely difficult to discover new block polymer materials via theory. This project will overcome this challenge, using generative artificial intelligence methods, similar to those that create new pictures from a library of images, to generate new possible block polymer structures by learning from a library of known structures. Machine learning will then be used to identify block polymers with appropriate chemistries and processing temperatures that will spontaneously form these new structures. This project will turn the theory of block polymer self-assembly from a tool primarily used for explanation to one that can be used for exploration and, ultimately, to design new materials with novel properties. The development of these computational techniques will provide graduate and undergraduate researchers with advanced training in polymer physics and high-performance computing, contributing to the US workforce, and the computer codes developed for this work will be released to the public so that others can use them for their research. High school students will be involved in the generative artificial intelligence part of the project, developing projects for the Minnesota Science Fair and sparking their interest in pursuing materials science as a future career path. TECHNICAL SUMMARY Block polymers are amphiphilic materials created by bonding two or more different polymers at their ends. Below the order-disorder temperature, a block polymer melt selects an ordered state that balances chain stretching and interfacial tension, subject to the constraint of filling space at constant density. Self-consistent field theory (SCFT) has proven to be a remarkably successful approach for understanding the block polymer thermodynamics governing order state selection. However, the non-linear governing equations in SCFT require high quality initial guesses for convergence. These computational challenges pose an obstacle when using SCFT to discover new block polymer phases. To overcome this obstacle, this project will use a generative adversarial network that learns from existing SCFT solutions to propose starting points for subsequent SCFT calculations, which are expected to converge to novel solutions. Those solutions are anticipated to be metastable states. As a result, a hierarchical computational method will be developed to efficiently solve the inverse problem of identifying block polymer formulations that stabilize those new phases. This computational workflow will be used to identify novel phases in triblock terpolymers. In addition to exhibiting a wide variety of ordered states, triblock terpolymers are synthetically tractable with fast ordering kinetics, making experimental tests of the predictions from this project realistic. Block polymers are the model system for understanding self-assembly in amphiphilic soft matter owing to their relatively simple thermodynamics, and SCFT’s ability to capture those thermodynamics is a major success in soft matter theory. However, SCFT has largely been relegated to an explanatory role in block polymer self-assembly; discovery typically results from serendipitous experimentation and then theory provides a physical rationale. This project turns the paradigm on its head by developing a powerful computational approach to discover new candidate phases and then identify block polymer formulations to produce those materials. The tools developed through this research will be made available to the community as open-source software. This project will provide both graduate and undergraduate students with advanced training in polymer physics, materials science, computer science, and numerical methods. K-12 education will be enhanced through an outreach effort with the Breck School to develop projects for the Minnesota Science Fair. STATEMENT OF MERIT REVIEW 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.

NIDCR - National Institute of Dental and Craniofacial Research

The Human Virome Program (HVP) seeks to advance our understanding of the human virome’s role in health and disease, but the field faces persistent challenges in virus discovery and characterization due to inconsistencies in analytic methods and uneven access to tools and benchmarking datasets. Currently, a range of bioinformatic tools exist for viral genome assembly, taxonomic classification, and novel virus identification, although many require specialized computing environments or expertise. Moreover, there is a lack of standardized benchmarks for assessing tool performance, leading to fragmented evaluations that make it difficult for researchers to identify the most effective methods for their specific needs. This absence of uniform evaluation criteria undermines reproducibility and reduces the efficiency of methodological innovation. To address these challenges, we propose the development of a platform to provide access to: 1) high-priority analytic tools, 2) standardized benchmarking datasets, and 3) an environment for continuous tool evaluation. Leveraging our expertise in genomic data analysis and interoperable computing, we will first establish a common HVP tool registry, prioritizing bioinformatic methods critical for virome research, including viral genome assembly, taxonomic assignment, and novel RNA virus identification. These tools will be ported to the Dockstore tool registry to maximize availability to all HVP researchers. This collection will be continuously maintained based on feedback from the HVP Bioinformatics Working Group and tool authors. To support standardized evaluation, we will construct comprehensive benchmarking datasets that reflect the complexity and diversity of human virome samples. These datasets will include both real and synthetic data, spanning a range of virome compositions, host backgrounds, and sequencing technologies. Synthetic datasets will be designed to test specific tool capabilities under controlled conditions, incorporating known viral genomes, engineered mutations, and varying levels of host and microbial contamination. Real datasets will be curated from validated sources, providing examples for benchmarking common analytic tasks. These resources will allow for standardized, reproducible comparisons of tool performance across multiple metrics, including accuracy, sensitivity, precision, recall, computational efficiency, and resource utilization. Finally, we will develop a benchmarking infrastructure using shared workspaces on the Terra cloud compute platform that integrate benchmarking datasets with analytic workflows, enabling users to test and compare tools directly. Researchers will have the flexibility to customize benchmarking parameters and compare performance across different use cases. To foster community engagement and continuous improvement, we will coordinate two benchmarking challenges officiated by the HVP Bioinformatics Working Group, the first involving a single use case of known virus identification, and the second spanning three additional priority use cases: 1) novel virus identification, 2) phage profiling, and 3) host prediction.

NINDS - National Institute of Neurological Disorders and Stroke

1 Malignant gliomas are the most common and deadly form of primary brain tumors in adults. Despite decades of 2 research, their formation and progression mechanisms remain poorly understood, and survival rates have 3 remained unchanged over the past 30 years. These glial tumors include isocitrate dehydrogenase (IDH1) mutant 4 (IDH1mut) and IDH wild-type (IDH1WT) subtypes, each of which presents with unique clinical and histopathological 5 correlates. Prognostic outcomes for IDH1WT tumors are poor, conferring a median survival of less than 14 6 months. In contrast, IDH1mut tumors confer significantly better prognoses, with a median survival of 31–65 7 months after diagnosis. Molecular and genomic studies have revealed that the disparity in survival outcomes 8 between glioma subtypes is primarily attributed to differences in tumor cell proliferation and invasiveness. 9 Emerging evidence, including our own studies, suggests that glioma progression is closely linked to neuronal 10 activity, where interactions between glioma cells and neurons create a vicious cycle that drives tumorigenesis. 11 Despite recent findings on tumor-neuron interactions, there remains a key knowledge gap in glioma biology 12 regarding the electrophysiological profiles of tumor cells and their role in tumor progression. Therefore, the 13 overarching goal of this proposal is to develop computational methods to integrate multi-modal data and 14 systematically link the molecular and electrophysiological states of glioma cells, enabling a deeper understanding 15 of the mechanisms underlying electrophysiological function that contribute to tumor progression. 16 Patch-sequencing (Patch-seq), which couples electrophysiological recordings, morphological analysis, 17 and single-cell RNA-sequencing (scRNA-seq), has unveiled various neuronal subtypes that feature distinct 18 properties. In our preliminary studies, in situ Patch-seq on surgically resected human gliomas to 19 ascertain whether glioma cells exhibit neuronal features. We have identified a unique subset of glioma cells, 20 termed neuronal-like tumor (NLT) cells, which exhibit partial neuronal characteristics. These cells are capable of 21 firing action potentials, mimicking the electrophysiological behavior of neurons, while retaining the morphological 22 features of glial cells. While Patch-seq provides paired electrophysiological and transcriptomic data, technical 23 limitations restrict profiling to hundreds of cells. In contrast, scRNA-seq alone allows the sequencing of millions 24 of cells but lacks corresponding electrophysiological profiles. To bridge this gap, we aim to develop computational 25 models to predict electrophysiological profiles within large scRNA-seq datasets. These models will identify cell 26 types and gene programs underlying electrophysiological function, uncover postsynaptic partners, and reveal 27 spatially colocalized cells of electrophysiologically active NLT cells. Collectively, our preliminary findings have 28 led us to our central hypothesis that understanding how glioma cells disrupt normal brain physiology by mimicking 29 neuronal functions will provide critical insights into their contribution to tumor progression and guide the 30 repurposing of approved neurological and psychiatric drugs that target neural-cancer signaling. we performed

NLM - National Library of Medicine

PROJECT SUMMARY Spatial biology techniques, such as multiplex immunofluorescence (mIF) imaging, enable detailed molecular profiling of tissue microenvironments while preserving spatial context. These techniques hold tremendous potential in basic research and personalized medicine, but the complexity of the resulting data necessitates carefully-designed computational methods to fully leverage this potential. Such approaches are especially needed for studying the tumor microenvironment, where both the cellular composition and spatial structure may play a critical role in patient outcomes. Our project will combine a unique mIF dataset with biologically-informed computational methods to further our basic understanding of the tumor microenvironment and identify patterns that can ultimately inform new treatment strategies. The dataset is derived from a clinically validated mIF assay that was performed prospectively across a real-world cohort of 2,032 patients encompassing over 20 cancer types. We will develop several complementary computational strategies tailored to multiplex images, ranging from spatial statistical methods to uncover tumor phenotypes in an unsupervised fashion, to deep learning models trained to directly predict patient outcomes. Importantly, the methods are designed with interpretability in mind, offering opportunities to identify spatial biomarkers associated with tumor, genomic, and clinical factors. Towards improved scalability, we will also develop deep learning approaches to predict mIF-based phenotypes from routine histopathology slides. The strategies developed in this project will be broadly applicable across spatial biology applications, and we will make these tools freely available to support other researchers.

NHGRI - National Human Genome Research Institute

Project Summary Next-generaƟon biobanks will contain sequence data for millions of individuals. This project will develop genotype phasing methods for these immense data sets. Genotype phasing esƟmates an individual's haplotypes, which are the two sequences of alleles that are inherited from the parents. Genotype phasing is necessary in order to perform powerful haplotype-based analyses. This project will develop computaƟonal methods that substanƟally reduce the cost of genotype phasing, so that it is possible to phase large biobanks for an acceptable cost. In addiƟon, the project will develop methods that allow marker filtering, sample filtering, and haplotype-based analyses to be performed directly on highly compressed data. This will remove the need to decompress phased genomic data prior to data filtering or haplotype-based analyses. This project will develop two methods that significantly improve genotype phase accuracy. The first method will idenƟfy difficult-to-phase heterozygous genotypes at run Ɵme and apply a special phasing algorithm to these heterozygotes. The second method will increase the number of geneƟc markers available for haplotype-based analyses by allowing less stringent quality control filters to be applied without harming genotype phase accuracy. Finally, the project will develop an open-source pipeline for phasing All of Us Research Program sequence data, and it will apply this pipeline to phase each sequence data release. The phased sequence data will be a shared resource that enables researchers to perform powerful haplotype-based analyses of All of Us genomic data, which will increase the power to detect genomic variants that influence heritable traits and diseases.

NLM - National Library of Medicine

ABSTRACT The gut microbiome has shown immense clinical potential for human disease as non-invasive biomarkers and new therapeutic interventions such as fecal microbiota transplants. A lack of replicable taxonomic associations across cohorts is currently limiting the clinical potential of the gut microbiome. Most microbiome studies focus on taxa-level associations because methods to taxonomically profile microbiomes are more mature and less computationally complex than those for functional profiling. However, the taxonomic composition of the gut microbiome is not conserved across individuals in different populations, resulting in the need for approaches that can explain the persistent relationships between different gut microbes and human health phenotypes. Given that the functional composition of the gut microbiome is more consistent across individuals and microbial proteins with the same function can originate in different species, my central hypothesis is that protein biomarkers underlie taxonomic gut microbial associations with human disease and health outcomes. There are no existing methods to quantify and efficiently map metagenomic short reads to microbial proteins, which is a necessary first step in identifying protein and peptide-level biomarkers. My objectives are to develop taxa-free frameworks for gut microbial biomarker discovery with a goal of understanding the role of gut microbial proteins in human disease. In Aim 1 I will develop and comprehensively benchmark a new tool to efficiently map whole-genome metagenomic shotgun sequencing reads to groups of homologous microbial proteins originating in different taxa. In Aim 2 I will develop non-invasive clinical risk stratification methods based on gut microbial proteins and introduce a pipeline for prioritizing specific microbial peptides for experimental validation. I will demonstrate the clinical utility of the risk stratification methods I develop by predicting risk from immune-checkpoint inhibitor therapy from metagenomic sequencing of fecal samples in published cohorts and validating the methods on a newly sequenced cohort. Successful completion of these aims will enable non-invasive clinical risk stratification and microbial biomarker discovery for a wide range of human health outcomes. Application of the developed methods will demonstrate their clinical utility and help clarify the roles of gut microbes in responses to anti-cancer immunotherapy.

NEI - National Eye Institute

Project summary This project aims to investigate the process and mechanisms of echoacoustic information accumulation in blind individuals by comparing their behavioral performance and brain activity with that of sighted controls, and a consequent creation and training of a dynamic computational model. Echolocation, a skill that allows blind individuals to navigate their environment using auditory cues, presents a unique opportunity to study neuroplasticity and sensory substitution. The study will employ behavioral tasks and EEG recordings to examine how auditory information is processed and accumulated in the brain, focusing on the temporal dynamics and neural correlates of this ability. Participants will be separated into two groups: a test group of blind individuals proficient in echolocation and a control group of sighted people naïve to echolocation skills. Behavioral tasks will involve locating and identifying objects using auditory clicks in a controlled environment, while EEG will be used to record brain activity during these tasks. The study will also incorporate computational modeling, including a dynamic neural network to simulate the accumulation of sensory information over time. This model will be trained on and validated against the collected data to understand how neural processes correspond to behavioral performance. The research will address key questions about how the brain adapts to the loss of vision by enhancing other senses, specifically auditory processing. By comparing the performance and neural activity of blind and sighted individuals in a dynamic system, the study aims to uncover differences in sensory processing strategies and the underlying neural circuitry. The findings could have broad implications for the development of training programs for echolocation and other sensory substitution techniques, as well as for our understanding of neural plasticity in general. The study is designed to minimize risks to participants, with appropriate protections in place for recruitment, informed consent, and data confidentiality. The expected outcomes include a detailed characterization of the neural mechanisms of echolocation, insights into the temporal dynamics of echoacoustic sensory processing, and a validated computational model that can predict neural responses based on behavioral performance. This research has the potential to advance the field of sensory neuroscience and contribute to the development of assistive technologies for the blind community.

NSF

Neonatal brachial plexus palsy (NBPP) is an injury that can occur during childbirth when a baby's head is stuck during delivery, which can cause pulling on the baby's neck and their nerves. How this stretching and pulling affects the nerves is unknown, which makes it difficult for a doctor to predict and prevent NBPP and delays treatment for injured newborns. This project will investigate how the nerves in the neck respond when subjected to various levels of stretch and for different durations. The team will use an animal model that has similarities to human newborns to understand NBPP. By stretching the nerves to different degrees and for varying lengths of time, the team will understand the response of these nerves and the limit when injury occurs. The project will also develop computer simulations that will mimic the conditions during complicated deliveries and allow researchers to estimate the stress the nerves experience during stretching. This study will enable the team to develop a practical tool that doctors can use to better predict NBPP and make informed decisions that reduce the risk of nerve injuries in newborns. The project will also help train the future science and engineering workforce by engaging students in research, fostering interest in STEM fields, and encouraging future contributions to healthcare advancements. NBPP untreated has serious long term impacts including muscle atrophy, impaired bone development, and osteoarthritis. There is a critical knowledge gap correlating the effects of over-extended neck, prolonged delivery, and mechanical forces imposed on the neonatal brachial plexus (NBP) during obstructed delivery. This limits understanding of the injury mechanism and ultimately delays prognosis and appropriate intervention in the affected newborn. This study will address this critical gap by investigating the stress relaxation behavior of the NBP when subjected to clinically relevant magnitude and durations of pre-stretch as observed during prolonged head-to-body delivery. The study will leverage a neonatal animal model to provide biomechanical failure data and structural changes in the NBP that are pre-stretched to a range of stretch magnitudes (10%, 15%, and 20% strains) and duration (90 and 300 seconds) as observed during prolonged delivery in the clinic. The study will also integrate experimental data with computational models of the maternal pelvis and neonate that will serve as a surrogate and help overcome the existing ethical limitations in accessing NBP biomechanical information. By using experimentally obtained viscoelastic properties of the NBP, the computations will evaluate NBP risk factors that include endogenous maternal forces, exogenous clinician-applied forces, and fetopelvic disproportion. This integrated approach will enable a highly translational predictive tool that can simulate prolonged delivery in an over-extended neck and predict NBP strains during birthing. The project will also enable hands-on experience for undergraduate students through design courses and summer research programs and for K-12 students through engineering summer camp programs. 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

Tao Li of University of Delaware is supported by an award from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry to develop new computational methods for modeling molecular polaritons—hybrid states arising from the interaction between molecules and confined light inside optical cavities. These hybrid states have shown promise for controlling chemical reactions and energy transfer in unconventional ways. However, current theoretical models fail to capture the complexity of real-world experiments, which involve millions of molecules interacting with a complicated photonic environment. To bridge this gap, this project will develop innovative simulation tools by integrating molecular dynamics, first-principles electronic structure methods, and computational electrodynamics. These tools will enable more accurate modeling of polariton chemistry in realistic experimental conditions, ultimately deepening our understanding of how strong light-matter interactions influence molecular processes. In addition to scientific advancements, the team will make their computational tools openly available to the general public. Dr. Li’s research will focus on developing three theoretical frameworks for modeling collective strong coupling in molecular ensembles. First, the team will implement the mesoscale molecular dynamics simulation approach to describe vibrational strong coupling in Fabry–Pérot cavities by explicitly accounting for multimode photonic environments. Second, this project will develop a first-principles simulation approach to study electronic excited-state dynamics under both vibrational and electronic strong couplings, thus enabling a unified description of nonadiabatic processes under strong coupling. In addition, a semiclassical computational electrodynamics approach will be developed which treats the bulk molecular ensemble as a dielectric medium while explicitly simulating the quantum dynamics of impurity molecules. By combining these approaches, the aim is to advance the theoretical modeling of polariton chemistry and provide powerful computational tools for understanding strong light-matter interactions at experimentally relevant scales. Beyond the polariton study, the developed computational tools can also advance research in fields such as plasmonic catalysis and plasmon-enhanced spectroscopy. 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

With support from the Chemical Measurement and Imaging Program in the Division of Chemistry, Steffen Lindert and his group at Ohio State University are working to understand and improve mass spectrometry (MS) measurements of covalently labeled proteins. Chemical reactions of accessible sites on a protein with specific labels followed by the use of MS to identify the areas that were labeled helps illuminate protein conformation. Dr. Lindert will perform molecular dynamics simulations to understand how labeling reagents interact with proteins and whether they might induce unwanted conformational changes. Additionally, a web server will be developed to identify the optimal covalent labeling reagents for a given protein sequence. These studies are designed to improve MS covalent labeling measurements, and in this way lead to a better understanding of protein conformation, with potentially broad long term scientific impact in protein conformational/folding studies. If successful, these studies will support work with minimally invasive covalent labeling reagents and support the design of new labeling reagents that do not distort the probed structure. The performance and utility of MS covalent labeling measurements would be greatly improved by a better understanding of how covalent labeling reagents interact with proteins and through a systematic understanding of which labels are most suited for a particular protein under investigation. To address these needs, continued computational work that advances covalent labeling measurement science is required. Dr. Lindert’s research is expected to further improve covalent labeling measurements by addressing these current limitations. MD simulations will be used to develop better models of how different covalent labels interact with proteins. The interaction of several commonly used covalent labels with proteins in solution will be simulated before and after covalent attachment, and the Lindert group will explore if and potentially how certain labels distort protein structure or dynamics. Knowledge gained from these simulations has the potential to elevate understanding of covalent labeling measurements. The protein sequence defines possible structural conformations, illuminating which residues may be optimal for labeling in structure determination studies. Subsequently, a web server will be developed to identify the optimal covalent labeling reagents for a given protein sequence. The projected development of this web server stands to benefit the entire protein covalent-labeling community. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NIGMS - National Institute of General Medical Sciences

Abstract This R35 Maximizing Investigators’ Research Award (MIRA) proposal aims to develop computational systems to address critical challenges in single-cell and spatial transcriptomics experiments. Single-cell and spatial transcriptomics experiments allow us to measure genome-wide gene expression in tens of thousands of individual cells (e.g. from blood, healthy tissue, tumors, or other diseased tissue) or across thousands of tissue spots. As a result, they have emerged as revolutionary tools that allow us to address scientific and clinical questions that were elusive just a few years ago. Computation has likewise been transformed by developments in artificial intelligence (AI). In spite of incredible advances, our ability to obtain genomic measurements continues to outpace our ability to derive useful information from them. This MIRA proposal addresses some of the most critical challenges that are currently limiting the pace at which the scientific community can turn valuable data from high-throughput genomic experiments into meaningful results. In particular, we plan to develop powerful and efficient computational methods to improve the downstream analysis of data from single-cell and spatial transcriptomics experiments. To reduce barriers for non-experts, the methods we develop will be implemented in accessible, interactive software that exploits the power of AI foundation models. Comprehensive benchmarking frameworks will also be developed so that our models and others can be appropriately evaluated. The proposed methods, software, and benchmarking frameworks are required to improve the use of AI single-cell foundation models in genomics, to maximize the information obtained from powerful single-cell and spatial transcriptomics experiments, and to enable biologists and/or clinicians to perform complex analyses without expertise in programming or data science.

NSF

This project will develop advanced computational tools to simulate and optimize the locomotion in a fluid medium of microscopic particles, such as bacteria or specially-designed micro-robots. Understanding how these microswimmers move is crucial for a wide range of applications, from improving our understanding of how biological systems function at the cellular level to creating innovative technologies. For instance, this research could lead to breakthroughs in targeted drug delivery, where microscopic robots deliver medicine precisely where it is needed in the body, or in developing new ways to transport materials within lab-on-a-chip or organ-on-a-chip devices. By making these simulations more efficient and accurate, this project aims to provide fundamental insights that will benefit fields such as biotechnology and materials science, ultimately contributing to scientific discovery and technological advancements that impact people's daily lives. Despite advancements in modeling and simulations of microswimmers, performing control and optimization in complex environments is still daunting, primarily owing to the lack of computational tools that scale to realistic problem sizes and work on arbitrary moving geometries. The primary goal of this project is to develop accurate, computationally efficient, and scalable numerical algorithms necessary for large-scale simulations, as well as shape and functional optimization of microswimmers. The research will address key computational difficulties, including the accurate evaluation of near-singular integrals and periodization schemes in three dimensions that support adaptivity. Adjoint formulations will be derived for various practical scenarios, such as microswimmers in confined flows or in the presence of other active or passive particles. Building on existing work with integral equation methods, the project will specifically develop adjoint-based shape optimization schemes, new high-order nearly-singular integration schemes to simulate flows through confined geometries, and spectrally-accurate, adaptive three-dimensional periodization schemes that can be accelerated by existing fast N-body algorithms. 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.

NIGMS - National Institute of General Medical Sciences

PROJECT SUMMARY/ABSTRACT Advances in molecular simulation models and methods have made it possible in many cases to simulate biological macromolecules and their environments over time scales of physical relevance with near-quantitative accuracy. In particular, such approaches are now used directly in the pharmaceutical industry to guide synthesis of new compounds using computed changes in small molecule binding affinities. However, despite substantial work by many researchers over decades, the majority of protein-ligand binding phe- nomena related to human health remain prohibitively expensive to model with sufficient accuracy using current simulation capabilities. Such therapeutic modalities include targeting protein-protein interfaces, kinases and phos- phatases inhibitors, and partners of short peptide interaction motifs. One significant reason for this difficulty is that these targets favor bigger and more flexible binders, with both larger scale motions and less frequent transi- tions between metastable states, requiring significantly more sampling than is currently computationally feasible. Additionally, existing force fields have been developed in a piecemeal fashion, do not generalize over all of rel- evant chemistry, and the proper trade-offs between increased complexity and computational cost of force field improvements are not clear. We propose to address these challenges building on our extensive expertise in methods development and simu- lation software for improved biophysical modeling and drug design. In collaboration with other researchers in the Open Force Field Initiative, my group will work to build systematically improvable force fields that achieve high accuracy and broad coverage of NIH-relevant chemistry, that self-consistently model heterogeneous biomolecular systems, and that can be broadly applied across a range of high-performance software packages. In particular, we will use new sources of thermophysical data to optimize and validate behavior in complex chemical environments and will develop statistical approaches to better gauge the utility of increasing force field complexity. In addition, we will develop novel, well-validated molecular simulation sampling methodologies that can capture the full configurational ensembles needed to accurately compute binding free energies of large, flexible ligands. Such algorithms will be designed to support new and anticipated computing architectures such as heterogeneous clusters and cloud computing via asynchronous approaches, and will augment molecular dynamics with recent developments in machine learning structural prediction. This research is significant and transformative as the computational concepts and technology developed via this research will enable accurate prediction of larger, flexible ligands from across all of bioorganic chemistry. The effort has the potential to assist drug developers break away from existing design paradigms of relatively small, rigid molecules bound to relatively rigid well-defined binding sites, accessing new targets and mechanisms of action and as well as therapeutics targeting them.

NIDA - National Institute on Drug Abuse

SUMMARY: Dopamine, serotonin, and norepinephrine neurotransmitters are known to be critically involved in process underlying substance use disorder and psychiatric illness, as well as healthy motivated behavior, decision-making, and learning. However, little is known about how these signals coordinate and modulate subjective feeling and motivate behavior as mammals (including humans) navigate the world. Progress has been hindered by a lack of technology that permits fast, real-time, measurements that can discriminate and track dopamine, serotonin, and norepinephrine release simultaneously in areas of the brain where two or more of these neurotransmitters are co-released. A major challenge to current methods (e.g., fast scan cyclic voltammetry) is that the calibration models use to interpret in vivo data are trained in vitro and it is unclear how the background signal changes between these environments and how this affects the measured responses. This proposal capitalizes on (and seeks to radically improve) a technological innovation developed by the principal investigator, which resulted in the first successful colocalized measurements of dopamine and serotonin release with sub-second temporal resolution from the brains of consciously behaving humans. Here, we pursue two specific aims, which seek to develop a computational approach to extend these kinds of measurements to include simultaneous detection of norepinephrine and make these methods available for a larger area of preclinical animal model research and human clinical neuroscience research. In both aims we will be testing the overarching hypotheses that 1) the ‘background’ signal present in fast scan cyclic voltammetry measurements can be quantitatively characterized, mathematically modeled, and therefore subtracted using a model-based approach in in vivo research paradigms; and 2) that the “in vitro bias” in the mathematical models used in model-based electrochemistry can be corrected for if we can obtain a better characterization of the background signals in each of the in vivo, ex vivo, and in vitro conditions. The experiments and analyses proposed will begin to provide much needed clarity on the impact biological ‘interferents’ have on interpreting in vivo fast scan cyclic voltammetry data – currently the only approach amenable to sub-second multi-neurotransmitter detection in humans. We expect to develop mathematical models and calibration methods that can be used to predict and control for unwanted interfering signals while significantly improving detection methods for multi-neurotransmitter detection. Notably, these advances – to be shared via open-source online repositories – would accelerate ongoing efforts in the field aimed at understanding how dopaminergic, serotonergic, and noradrenergic systems coordinate to motivate behavior in humans and pre-clinical model organisms, and thereby provide insight into mechanisms underlying human mental health.

NSF

With the support of the Chemical Synthesis (SYN) program in the Division of Chemistry Professor Timothy Newhouse of Yale University is studying the development of synthetic routes to diterpenoids. The synthesis of small molecules is one of the rate-limiting steps across disciplines from materials science to medicinal chemistry. To overcome this bottleneck in the discovery process, we need methodological advancements and improvements to the synthesis design process. The long-term goal of this proposal is to apply computational strategies to synthetic planning to access structurally complex natural products, and in this proposal these efforts are focused on synthesis of diterpenoids. The approach to model development described in this proposal can be applied to any synthetic transformation and would be enabling and thus broadly impactful whenever that transformation’s short-term experimental evaluation is not possible. Moreover, strategic partnerships within and around the Yale community will bring science to K-12 audiences. The design of a synthetic pathway to a desired molecule is generally conducted by human analysis although computational approaches are beginning to emerge. This proposal outlines the development of several artificial intelligence-based tools to predict the yield of common carbon-carbon bond forming cyclization reactions. Additionally, generative modelling is proposed to design ligands, substrates, and routes to natural products. These computational methods will enable the planning and synthesis of natural products and analogs. High-risk yet high-reward plans are de-risked through the use of machine learning models, thereby allowing efficient and expedient laboratory access to synthetically challenging molecular scaffolds. 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.

NIAID - National Institute of Allergy and Infectious Diseases

Summary. The prevalence of antibiotic resistant (AR) bacterial infections continues to grow. Although the development of novel antimicrobial compounds is one approach to combat AR infections, phage therapy or using lytic phage to treat bacterial infections, offers another solution that has many attractive benefits, including the specificity of infection, leaving the healthy microbiome intact, low toxicity, and the diversity of phages available. However, before phage therapy can be widely used in the clinic, one of the significant challenges that must be addressed is the selection of which phage to use for a given bacterial pathogen. Phage infection specificity is complicated by the fact that bacteria encode a diverse array of phage defense systems that block phage infection, typically expressed by horizontally transferrable DNA elements. The bacterial pathogen must also encode and express the phage receptor and any bacterial host factors that the phage requires for successful replication and phage production. Currently, bacterial pathogens are manually screened against large phage biobanks to select phage cocktails that can provide effective in vivo killing. Although this has been effective, such an approach is costly and, more importantly, time-intensive, and it will be challenging to scale up as phage therapy becomes more widely used. The field, therefore, needs rapid and cost-effective approaches to identify effective phages for any bacterial pathogen, given the genome sequence of the bacteria and phages. The MPIs of this proposal, Ravi and Waters, will use their diverse expertise in bacterial pathogenesis, phage biology and defense, AR, microbial genomics, and computational biology, to develop an ML-based prediction model that can identify effective phage and phage resistance-associated molecular features for any given E. coli strain. Another critical outcome of this work will be the gold-standard data set generated in Aim 1 that will define the successful infection of 69 dsDNA E. coli phage in the well-characterized BASEL phage collection with ~600 sequenced pathogenic and non-pathogen E. coli strains generating ~42,000 unique data points. Aim 2 will first define all known phage defense, AR, and virulence elements in this collection of E. coli and merge these annotated features with the phage host infection phenotypes generated in Aim 1 using (un)supervised ML-based approaches (e.g., logistic regression, random forest) to generate models that can predict effective phage infections of any given E. coli host, along with the underlying molecular features (genes, proteins, domains) culminating in resistance/susceptibility. This model will be validated with 50 new E. coli strains. Successful completion of this proposal will generate a clinically useful predictive model for E. coli and lay the framework for generating such predictive models for phage therapies against other bacterial pathogens. Moreover, the model will lead to the discovery of novel phage defense elements and bacterial factors that impact phage infection, and a deeper understanding of how bacterial pathogens evolve resistance to phage infection, knowledge, which can be used to effectively tailor phage therapy to prevent widespread emergence of resistance.