Biology subjects

White, A. D.

Publications and source records attributed to White, A. D..

15 recordsLinked to original sources

bioRxiv2026

Small molecule structure elucidation from NMR spectra with generative and reasoning models

Elucidating the structures of unknown small molecules remains a major challenge in chemistry, pharmacology, and drug discovery. While nuclear magnetic resonance (NMR) spectroscopy is the gold standard for structure elucidation, interpreting spectra remains labor-intensive and requires expert knowledge to navigate the vast space of possible structures. Here, we present an AI system for rapid, automated de novo structure elucidation that combines a molecular generative model with a scientific reasoning agent. Its generative component, ChefNMR2, is an atomic flow-matching model that predicts small-molecule structures from molecular formulae and one-dimensional 1H and 13C NMR spectra. ChefNMR2-Agent augments our generative model with large language models (LLMs) to reason over and refine candidate structures. Across diverse benchmarks, ChefNMR2 substantially outperforms existing tools, offering calibrated confidence estimates and two to three orders of magnitude faster inference than frontier LLMs. We use ChefNMR2-Agent to elucidate the structures of four novel natural products, matching human-chemist accuracy in a fraction of the time. Our work demonstrates how specialized generative models and general-purpose reasoning agents can unite to solve challenging scientific inverse problems and accelerate NMR structure elucidation, thus opening new possibilities for AI-driven small-molecule discovery across diverse disciplines.

bioinformatics↗

bioRxiv2026

Super-Resolution Visual ProteomEx for Hard Tissues and Clinical Samples

Correlating nanoscale tissue architecture with unbiased molecular composition within the same specimen remains a fundamental challenge in spatial biology. Current spatial proteomics methods either lack the imaging resolution required to resolve nanoscale tissue architecture or achieve only limited proteome coverage, and untargeted approaches capable of deep protein identification have yet to be integrated with super-resolution imaging within a single workflow. Here, we present microProteomEx, an integrated platform that combines hydrogel-assisted tissue expansion with mass spectrometry-compatible fluorescence imaging, laser capture microdissection, and bottom-up proteomics to achieve simultaneous three-dimensional super-resolution imaging (effective lateral resolution down to [~]47 nm on conventional diffraction-limited microscopes) and spatially resolved proteomics at a scalable lateral resolution of 29-100 {micro}m (0.02-0.28 nL volumetric resolution). By optimizing fixation, protein anchoring, and a secondary re-embedding strategy, we extend the method to mechanically resilient tissues, including mouse kidney and heart, as well as to formalin-fixed, paraffin-embedded clinical specimens. We demonstrate single-glomerulus and single-plaque proteomics (450-1,000 proteins per structure), resolve proteomic differences between malignant melanoma and giant congenital melanocytic nevus in a rare pediatric case, and characterize the morphology-resolved molecular heterogeneity of cored versus diffuse amyloid-beta plaques in a mouse model of Alzheimers disease across two disease stages. microProteomEx establishes a broadly accessible framework for correlating tissue ultrastructure with deep spatial proteomics, with direct implications for disease biology, biomarker discovery, and precision medicine.

biochemistry↗

bioRxiv2026

Ultrasensitive voltage imaging reveals distinct electrical microdomains in neurons

For the brain to compute, electrical signals must propagate over the membranes of individual neurons, connecting synaptic inputs to synaptic outputs1. Complex neuronal morphologies coupled with the spatial organization of synaptic inputs and outputs enable diverse voltage transformations that underlie cell-type specific computations2,3. However, measuring these transformations in vivo has remained challenging, leaving a crucial gap in our mechanistic understanding of single neuron computation. Here, we develop ASAP7y, a genetically encoded voltage indicator with unprecedented subthreshold sensitivity and expanded excitation compatibility in both mice and flies. We leveraged ASAP7y combined with two-photon random-access microscopy to record sensory stimulus-evoked voltage dynamics with millisecond, subcellular, and subthreshold resolution along the neurites of individual neurons in Drosophila. We found remarkable heterogeneity in voltage propagation across cell-types, delineating a fundamental axis of electrical diversity. Leveraging a nanoscale EM reconstruction of the visual system4, we modeled the electrotonic properties of single neurons spanning 717 cell types, revealing how morphology shapes voltage transformations. Finally, we demonstrate that confined voltage propagation creates substrates for local computation, producing subcellular domains with distinct feature selectivity across multiple cell types. These results provide mechanistic insight into how critical single neuron computations arise and reveal parallel processing in single neurons.

neuroscience↗

bioRxiv2026

Intrinsic space-time couplings governing multi-scale cortical dynamics

The neocortex covers a vast expanse of the mammalian brain and represents the principal target of clinical neuromodulation; however, the global principles of neocortical operation have been challenging to identify. In this regard, a limitation has been tracking activity with cortex-wide spatial coverage while maintaining access to the millisecond temporal resolution of neuronal firing-- a crucial combination not achievable with existing recording technologies. Here we introduce and apply conformal immersion microscopy, enabling activity tracking across the entire dorsal cortex with millisecond temporal resolution and 100 {micro}m spatial resolution (thus spanning five orders of magnitude in time and four in space), at sufficient sensitivity to resolve single-trial activity beyond 100 Hz. Drawing on physics-based frameworks, we apply multiscale analysis to identify a fundamental frequency-dependent coherence length that partitions the neocortex into discrete dynamical elements with well-defined propagation speeds, boundaries, and scale-invariant dynamics. These dynamical elements were found to be conserved from sub-threshold to suprathreshold (neuronal firing) regimes of neural activity, and were robust to diverse pharmacological, optogenetic, and genetic interventions. However, it was possible to identify and establish conditions allowing elemental boundaries to be selectively overridden, and to allow perturbation of specific elements even while conserving global dynamical architecture. Together, these findings enable measurement of intrinsic spatiotemporal parameters governing the dynamical organization of neocortex, which may provide a foundation for mechanistically-informed basic and translational understanding.

neuroscience↗

bioRxiv2026

Causal and directional elements of global brain dynamics

Mammalian cognition appears to involve coordinated neural activity within and across brain-spanning networks. However, the governing principles of large-scale integration remain largely unknown. Here we developed an approach to record spontaneous cortex-wide neural activity (including with novel genetically encoded activity sensors), over very long timescales, and applied unbiased computation to capture conserved spatiotemporal regularities in these data. Initial screening extracted a set of directional elements: consistent spatiotemporal patterns of cortical activity that generalized across excitatory and inhibitory neuronal cell types as revealed by genetic targeting of activity sensors. In particular, novel genetically encoded voltage-sensing strategies revealed that the directional elements were not only highly conserved, but also were represented across the full neural activity frequency spectrum including the fastest timescales (gamma rhythms) enabled by voltage-sensing. Exploiting the spatiotemporal structure revealed by these directional elements, we tested for causal rules governing cortical network activation, using patterned optogenetic stimulation combined with activity imaging. We found that the directional propagation structure of these elements encoded a causal control hierarchy, as source regions (but not sink regions) sufficed to drive full element recruitment--a principle that held across all tested directional elements. Employing a panel of psychotropic drugs, we showed that directional element structure and excitability were robust to manipulations of neural and behavioral state, even as spontaneous network dynamics were reshaped in compound-specific ways. Finally, we developed and applied an all-optical sensing/control approach targeting the directional elements in a behavioral visual detection paradigm, revealing contributions of these conserved large-scale dynamics to elevated sensorimotor behavioral performance.

neuroscience↗

bioRxiv2025

Drug-induced phospholipidosis as an artifact in antiviral drug repurposing

Drug repurposing in principle can speed antiviral drug discovery. Among the molecules most frequently advanced in such repurposing efforts are a group of structurally diverse cationic amphiphilic drugs (CADs). While CADs have shown micromolar to mid-nanomolar antiviral activity in cell-based assays, they can induce phospholipidosis, confounding in COVID-19 repurposing. A barrier to the identification of phospholipidosis inducers has been the involved nature of the microscopy assays used to characterize them. To ease the identification of these artifacts, we describe a rapid microplate-based assay to detect phospholipidosis. Leveraging this assay, we quantified the prevalence of phospholipidosis-inducers across several cell-based antiviral repurposing screens. We selected 40 drugs reported to have micromolar antiviral activities and found that 26 of them (65%) induced phospholipidosis within the same concentration range as their reported antiviral activities. Intriguingly, we identified four non-CADs that also induce phospholipidosis, revealing a new group of drugs that can lead to this toxic event and highlighting the importance of facile experimental assays to detect it. Understanding how phospholipidosis can confound antiviral drug discovery, and its rapid detection, will help prevent what is an apparently general artifact, active across viruses, from distracting investigators from potentially more useful candidates.

cell biology↗

bioRxiv2025

Differential role of phosphorylation in glucagon family receptor signaling revealed by mass spectrometry

In response to extracellular ligands, G protein-coupled receptors (GPCRs) undergo conformational changes that induce coupling to intracellular effectors such as heterotrimeric G proteins that trigger various downstream signaling pathways 1. These events have been shown to be highly regulated by the concerted effects of post-translational modifications (PTMs) that occur in a ligand-dependent manner. Most notably, phosphorylation of residues in the C-terminal cytoplasmic tail of GPCRs has been strongly implicated in promoting receptor interactions with {beta}-arrestins ({beta}arrs) 2, 3 which are cytosolic adaptor proteins that modulate G protein coupling 4, receptor internalization 5, and perhaps also serve as signaling modules in their own right 6. Here, we use proteomic methods to identify C-tail residues that are phosphorylated in the glucagon family of class B1 GPCRs (GLP-1R, GCGR, GIPR) upon agonist addition. Mutagenesis studies reveal unique effects of phosphorylated residues on {beta}arr recruitment and cyclic AMP (cAMP) production that occur in a receptor-dependent manner. We demonstrate that phosphorylation of GLP-1R and GIPR is a critical determinant in driving the formation of GPCR-{beta}arr complexes. However, our results suggest that ligand-induced {beta}arr recruitment to the GCGR proceeds in a phosphorylation-independent manner. These findings highlight the importance of recognizing phosphorylation as a component in the regulation of class B1 GPCR signaling, but also the need to consider how such phenomena may not necessarily yield identical effects on intracellular signaling cascades.

pharmacology and toxicology↗

bioRxiv2024

Prodrug activation in malaria parasites mediated by an imported erythrocyte esterase, acylpeptide hydrolase (APEH)

The continued emergence of antimalarial drug resistance highlights the need to develop new antimalarial therapies. Unfortunately, new drug development is often hampered by poor drug-like properties of lead compounds. Prodrugging temporarily masks undesirable compound features, improving bioavailability and target penetration. We have found that lipophilic diester prodrugs of phosphonic acid antibiotics, such as fosmidomycin, exhibit significantly higher antimalarial potency than their parent compounds (1). However, the activating enzymes for these prodrugs were unknown. Here, we show that an erythrocyte enzyme, acylpeptide hydrolase (APEH) is the major activating enzyme of multiple lipophilic ester prodrugs. Surprisingly, this enzyme is taken up by the malaria parasite, Plasmodium falciparum, where it localizes to the parasite cytoplasm and retains enzymatic activity. Using a novel fluorogenic ester library, we characterize the structure activity relationship of APEH, and compare it to that of P. falciparum esterases. We show that parasite-internalized APEH plays an important role in the activation of substrates with branching at the alpha carbon, in keeping with its exopeptidase activity. Our findings highlight a novel mechanism for antimicrobial prodrug activation, relying on a host-derived enzyme to yield activation at a microbial target. Mutations in prodrug activating enzymes are a common mechanism for antimicrobial drug resistance (2-4). Leveraging an internalized host enzyme would circumvent this, enabling the design of prodrugs with higher barriers to drug resistance. SignificanceRising antimalarial drug resistance threatens current gains in malaria control. New antimalarial drugs are urgently needed. Unfortunately, many drug candidates have poor drug-like properties, such as poor absorbability in the gastrointestinal tract, or poor accumulation at the site of action. This can be overcome by prodrugging, the addition of prodrug groups which mask poor drug features until they are removed by an activating enzyme. Here, we show that a red blood cell enzyme, acylpeptide hydrolase, is taken up by malaria parasites and serves as the activating enzyme for multiple lipophilic ester prodrugs. Our findings highlight a novel mechanism for prodrug activation, which could be leveraged to design novel prodrugs with high barriers to drug resistance.

microbiology↗

bioRxiv2024

Mtb specific HLA-E restricted T cells are induced during Mtb infection but not after BCG administration in non-human primates and humans

Novel vaccines targeting the worlds deadliest pathogen Mycobacterium tuberculosis (Mtb) are urgently needed as the efficacy of the Bacillus Calmette-Guerin (BCG) vaccine in its current use is limited. HLA-E is a virtually monomorphic unconventional antigen presentation molecule and HLA-E restricted Mtb specific CD8+ T cells can control intracellular Mtb growth, making HLA-E a promising vaccine target for Mtb. In this study, we evaluated the frequency and phenotype of HLA-E restricted Mtb specific CD4+/CD8+ T cells in the circulation and bronchoalveolar lavage fluid of two independent non-human primate (NHP) studies and from humans receiving BCG either intradermally or mucosally. BCG vaccination followed by Mtb challenge in NHPs did not affect the frequency of circulating and local HLA-E/Mtb CD4+ and CD8+ T cells, and we saw the same in humans receiving BCG. HLA-E/Mtb T cell frequencies were significantly increased after Mtb challenge in unvaccinated NHPs, which was correlated with higher TB pathology. Together, HLA-E/Mtb restricted T cells are minimally induced by BCG in humans and rhesus macaques (RMs) but can be elicited after Mtb infection in unvaccinated RMs. These results give new insights into targeting HLA-E as a potential immune mechanism against TB.

immunology↗

bioRxiv2023

Nonspecific membrane bilayer perturbations by ivermectin underlie SARS-CoV-2 in vitro activity

Since it was proposed as a potential host-directed antiviral agent for SARS-CoV-2, the antiparasitic drug ivermectin has been investigated thoroughly in clinical trials, which have provided insufficient support for its clinical efficacy. To examine the potential for ivermectin to be repurposed as an antiviral agent, we therefore undertook a series of preclinical studies. Consistent with early reports, ivermectin decreased SARS-CoV-2 viral burden in in vitro models at low micromolar concentrations, five-to ten-fold higher than the reported toxic clinical concentration. At similar concentrations, ivermectin also decreased cell viability and increased biomarkers of cytotoxicity and apoptosis. Further mechanistic and profiling studies revealed that ivermectin nonspecifically perturbs membrane bilayers at the same concentrations where it decreases the SARS-CoV-2 viral burden, resulting in nonspecific modulation of membrane-based targets such as G-protein coupled receptors and ion channels. These results suggest that a primary molecular mechanism for the in vitro antiviral activity of ivermectin may be nonspecific membrane perturbation, indicating that ivermectin is unlikely to be translatable into a safe and effective antiviral agent. These results and experimental workflow provide a useful paradigm for performing preclinical studies on (pandemic-related) drug repurposing candidates. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=61 SRC="FIGDIR/small/563088v1_ufig1.gif" ALT="Figure 1"> View larger version (18K): org.highwire.dtl.DTLVardef@a21f94org.highwire.dtl.DTLVardef@1c76751org.highwire.dtl.DTLVardef@500930org.highwire.dtl.DTLVardef@8b6c05_HPS_FORMAT_FIGEXP M_FIG Graphical abstract C_FIG

pharmacology and toxicology↗

bioRxiv2023

Learning Peptide Properties with Positive Examples Only

AO_SCPLOWBSTRACTC_SCPLOWDeep learning can create accurate predictive models by exploiting existing large-scale experimental data, and guide the design of molecules. However, a major barrier is the requirement of both positive and negative examples in the classical supervised learning frameworks. Notably, most peptide databases come with missing information and low number of observations on negative examples, as such sequences are hard to obtain using high-throughput screening methods. To address this challenge, we solely exploit the limited known positive examples in a semi-supervised setting, and discover peptide sequences that are likely to map to certain antimicrobial properties via positive-unlabeled learning (PU). In particular, we use the two learning strategies of adapting base classifier and reliable negative identification to build deep learning models for inferring solubility, hemolysis, binding against SHP-2, and non-fouling activity of peptides, given their sequence. We evaluate the predictive performance of our PU learning method and show that by only using the positive data, it can achieve competitive performance when compared with the classical positive-negative (PN) classification approach, where there is access to both positive and negative examples.

bioinformatics↗

bioRxiv2022

Now What Sequence? Pre-trained Ensembles for Bayesian Optimization of Protein Sequences

Pre-trained models have been transformative in natural language, computer vision, and now protein sequences by enabling accuracy with few training examples. We show how to use pre-trained sequence models in Bayesian optimization to design new protein sequences with minimal labels (i.e., few experiments). Pre-trained models give good predictive accuracy at low data and Bayesian optimization guides the choice of which sequences to test. Pre-trained sequence models also remove the common requirement of having a list of possible experiments. Any sequence can be considered. We show significantly fewer labeled sequences are required for three sequence design tasks, including creating novel peptide inhibitors with AlphaFold. These de novo peptide inhibitors require only sequence information, no known protein-protein structures, and we can predict highly-efficient binders with less than 10 AlphaFold calculations.

bioinformatics↗

bioRxiv2022

Serverless Prediction of Peptide Properties with Recurrent Neural Networks

We present three deep learning sequence prediction models for hemolysis, solubility, and resistance to non-specific interactions of peptides that achieve comparable results to the state-of-the-art models. Our sequence-based solubility predictor, MahLooL, outperforms the current state-of-art methods for short peptides. These models are implemented as a static website without the use of a dedicated server or cloud computing. Web-based models like this allow for accessible and effective reproducibility. Most existing approaches rely on third-party servers typically that require upkeep and maintenance. That trend leads to a relatively longer lifetime of web-based models. These predictive models do not require servers, require no installation of dependencies, and work on across a range of devices. The models are bidirectional recurrent neural networks. This serverless prediction model is a demonstration of edge machine learning that removes the dependence on cloud providers. The code and models are accessible at https://github.com/ur-whitelab/peptide-dashboard.

bioinformatics↗

bioRxiv2020

Comparison of Rhesus and Cynomolgus macaques as an authentic model for COVID-19.

A novel coronavirus, SARS-CoV-2, has been identified as the causative agent of the current COVID-19 pandemic. Animal models, and in particular non-human primates, are essential to understand the pathogenesis of emerging diseases and to the safety and efficacy of novel vaccines and therapeutics. Here, we show that SARS-CoV-2 replicates in the upper and lower respiratory tract and causes pulmonary lesions in both rhesus and cynomolgus macaques, resembling the mild clinical cases of COVID-19 in humans. Immune responses against SARS-CoV-2 were also similar in both species and equivalent to those reported in milder infections and convalescent human patients. Importantly, we have devised a new method for lung histopathology scoring that will provide a metric to enable clearer decision making for this key endpoint. In contrast to prior publications, in which rhesus are accepted to be the optimal study species, we provide convincing evidence that both macaque species authentically represent mild to moderate forms of COVID-19 observed in the majority of the human population and both species should be used to evaluate the safety and efficacy of novel and repurposed interventions against SARS-CoV-2. Accessing cynomolgus macaques will greatly alleviate the pressures on current rhesus stocks.

microbiology↗

bioRxiv2020

Predicting Chemical Shifts with Graph Neural Networks

AO_SCPLOWBSTRACTC_SCPLOWInferring molecular structure from NMR measurements requires an accurate forward model that can predict chemical shifts from 3D structure. Current forward models are limited to specific molecules like proteins and state of the art models are not differentiable. Thus they cannot be used with gradient methods like biased molecular dynamics. Here we use graph neural networks (GNNs) for NMR chemical shift prediction. Our GNN can model chemical shifts accurately and capture important phenomena like hydrogen bonding induced downfield shift between multiple proteins, secondary structure effects, and predict shifts of organic molecules. Previous empirical NMR models of protein NMR have relied on careful feature engineering with domain expertise. These GNNs are trained from data alone with no feature engineering yet are as accurate and can work on arbitrary molecular structures. The models are also efficient, able to compute one million chemical shifts in about 5 seconds. This work enables a new category of NMR models that have multiple interacting types of macromolecules.

biophysics↗