Biology subjects

Bender, A.

Publications and source records attributed to Bender, A..

At least 19 recordsLinked to original sources

bioRxiv2026

Longitudinal diffusion tensor imaging identifies coordinated white matter remodeling during cocaine self-administration in mice

White matter microstructural alterations are increasingly recognized as a feature of cocaine use disorder (CUD), yet the longitudinal progression and biological underpinnings of these changes remain poorly understood. Although rodent intravenous self-administration (IVSA) provides a translational model of voluntary cocaine use, longitudinal diffusion tensor imaging (DTI) has not previously been combined with cocaine IVSA in mice. Here, we established a longitudinal IVSA-DTI framework to characterize white matter microstructural remodeling during cocaine self-administration and early withdrawal. Male C57BL/6 mice underwent DTI at baseline, following 11 consecutive days of cocaine (0.5 mg/kg/infusion) or saline IVSA, and after a 7-day withdrawal period immediately following a cocaine-seeking test (n = 4/group). Fractional anisotropy (FA) was evaluated longitudinally using atlas-based region-of-interest analyses and whole-brain diffusion maps. Cocaine-treated mice acquired robust operant responding and exhibited coordinated longitudinal FA alterations across interconnected hippocampal-associated white matter pathways. The most consistent changes occurred within the bilateral dorsal hippocampal commissure, fornix, and fimbria, with cocaine-treated mice demonstrating increased FA during self-administration relative to baseline while saline controls exhibited smaller or opposing longitudinal changes. Difference-in-differences maps further demonstrated cocaine-associated FA alterations within hippocampal pathways and localized regions of the corpus callosum after accounting for longitudinal changes in saline controls. These findings establish the feasibility of combining longitudinal DTI with cocaine IVSA in mice and demonstrate coordinated white matter microstructural remodeling during voluntary cocaine exposure. This framework provides a translational platform for future studies integrating longitudinal neuroimaging with genetically modified mouse models, tractography, and histological validation to investigate the mechanisms underlying cocaine-associated white matter plasticity.

neuroscience↗

bioRxiv2026

Model Validation Protocols for Machine Learning in Small Molecule Drug Discovery

Machine learning (ML) models for molecular property prediction are increasingly deployed in drug discovery, yet their adoption in real-world scenarios requires an understanding of the conditions in which a model succeeds or fails. While standardized benchmarks are powerful instruments to measure and unlock progress in ML research, they should not be blindly treated as the end goal. Especially static and retrospective benchmarks, in which no true unknown test set is employed, limit our ability to robustly validate a model's performance. Building on the collective expertise of a cross-industry consortium, we present a model validation framework consisting of five recommendations that would enable the community to move beyond aggregate metrics toward understanding where and why molecular property prediction models fail. We connect evaluation choices to real-world applications and case studies encountered in pharmaceutical research. The framework proposes splitting strategies that mimic realistic distribution shifts and expose common failure modes. We apply the recommended framework to a recently released dataset of absorption, distribution, metabolism, and excretion (ADME) properties. Across two complementary model algorithms, our case studies reveal four distinct failure modes (extrapolation, interpolation, representation, and evaluation), showing that model errors arise not only from distribution shift but also from limitations in molecular representations. Our results show that commonly used evaluation protocols can significantly overestimate performance and may not detect important model failure modes. All software and data are released via https://github.com/srijitseal/polaris.

bioinformatics↗

bioRxiv2026

A novel IL-1β reporter assay identifies tolcapone as a caspase-1 suppressor targeting pyroptosis in endotoxemia

Unconventional protein secretion mediated by gasdermin D (GSDMD) pores is essential for the release of pro-inflammatory cytokines such as interleukin-1{beta} (IL-1{beta}), a major driver of many inflammatory pathologies. Despite extensive investigation over the years, discovery of IL-1{beta} secretion modulators has been hindered by the lack of robust, scalable experimental platforms. To date, IL-1{beta} related studies largely rely on primary cells and animal models, suitable for mechanistic studies but not readily scalable for high-throughput applications. Here, we engineered a CRISPR-based reporter cell line that allows quantitative monitoring of endogenous IL-1{beta} secretion while preserving the physiologically relevant inflammasome signaling. This platform faithfully recapitulated the response of primary macrophages to pathogen-associated molecular pattern (PAMPs) stimulation and supported the screening of an FDA-approved drug library comprising 1,398 compounds. Form this screen, we identified tolcapone as a potent inhibitor of IL-1{beta} secretion, reducing cytokine release by more than 80% across the screening pipeline. Mechanistically, tolcapone suppressed caspase-1 activation, thereby limiting GSDMD cleavage, pore formation, and the downstream maturation and secretion of IL-1{beta} and IL-18 in vitro. In vivo, tolcapone administration attenuated the acute inflammatory response in a lipopolysaccharide-induced endotoxemia model. Together, these findings establish our reporter platform as a robust tool for discovery of endogenous IL-1{beta} secretion modulators and identify tolcapone as a promising inhibitor of inflammasome-driven immune pathology.

immunology↗

bioRxiv2026

Semi-quantitative Classification of HIV-1 Nucleic Acids Using ResNet Image Analysis of Discretized Isothermal Amplification Reactions in a Microfluidic Chip

Isothermal nucleic acid amplification tests enable rapid and decentralized molecular diagnostics but often lack robust quantitative readouts compared to quantitative PCR. Here, we present a semi-quantitative nucleic acid measurement approach using machine learning to extract spatiotemporal features from real-time fluorescence imaging of rapid isothermal amplification reactions in microfluidic chips. A convolutional neural network was trained on multiple images sampled throughout a chip-based recombinase polymerase amplification reaction to classify samples into clinically relevant or logarithmically spaced concentration ranges spanning five orders of magnitude. The clinical classification model achieved 94.6% accuracy, and the logarithmic model achieved 92.7% accuracy, with most errors occurring between adjacent concentration categories. By learning spatiotemporal patterns of fluorescence development rather than relying on explicit feature extraction, the model remained accurate at both high and low nucleic acid concentration regimes where other quantitative isothermal molecular tests struggle. This approach enables automated interpretation of amplification reactions and extends the usable dynamic range of the assay. These results demonstrate that integrating machine learning with image-based amplification methods can support rapid semi-quantitative molecular testing and may facilitate broader deployment of nucleic acid diagnostics outside centralized laboratory settings. Author summaryMany rapid nucleic acid testing methods for infectious diseases are simple to run but struggle to measure how much genetic material is present, which limits their usefulness in clinical decision-making. In our work, we study a technique that produces visible fluorescent patterns during nucleic acid amplification reactions. Traditionally, the amount of nucleic acids present are measured by counting individual bright spots, but this becomes difficult when the target nucleic acid concentration is high and the spots merge together. We developed a machine learning approach that models how the fluorescence pattern changes over time. By analyzing a sequence of images from each reaction, our model can assign samples to concentration ranges across a wide span. This allows us to extract meaningful information even when traditional analysis methods break down. Because this approach works with simple imaging systems and does not require complex equipment, it could help support more informative and accessible diagnostic testing in point-of-care and low-resource settings.

bioengineering↗

bioRxiv2026

Data-Efficient Exploration of Enzyme Function Using Family-Specific Machine Learning

Enzymes are essential biocatalysts across diverse industries, driving demand for high-performing variants. Foundation models are attractive for guiding enzyme discovery, but often lack the resolution to model subtle variations driving function within homologous families. Navigating these rugged functional landscapes to identify elite variants remains challenging and experimentally costly, even when guided by such models. Here we show that coupling dense, family-specific experimental screening with targeted, sequence-based deep learning provides a data-efficient discovery strategy. We experimentally screened 1,513 natural homologues from an esterase superfamily (>7,500 assays) and used this functional landscape to train task-specific models that predict activity, thermostability, and substrate specificity from sequence alone. Prospective experimental validation of previously untested sequences demonstrated that these task-specific models significantly outperformed generalist pre-trained and physics-based models in enriching for target traits. Residue-level attribution further indicated that the models captured sequence patterns consistent with underlying structural features. Finally, retrospective simulations showed that iterative retraining compresses the search space, discovering 60% of top-tier hits using nearly half the samples required by pre-trained baseline models. Together, these results highlight that machine learning can provide mechanistic insight, and that integrating targeted data acquisition with iterative machine learning provides a more data-efficient discovery strategy than relying on generic model scale.

bioengineering↗

bioRxiv2025

Unified Generation of Regionalized Neural Organoids from Single-Lumen Neuroepithelium

Human brain development begins with the formation of the neural tube, a neuroepithelium organized around a single continuous lumen that is patterned to specify distinct brain regions. Human pluripotent stem cell (hPSC)-derived neural organoids offer powerful models to study this process and its disruption in disease. However, most existing protocols rely on stochastic self-organization of hPSC aggregates, leading to high variability in tissue architecture and cell type composition, which limits reproducibility and fidelity to natural development. This variability has also hindered the broader adoption of brain organoid technology in translational and pharmaceutical research, where robustness and standardization are critical. Here, we present a scalable and reproducible platform that uses uniform, single-lumen neuroepithelium (SLN) as a standardized starting point for generating diverse, regionally specified neural organoids. SLNs form with high efficiency and reproducibility, can be patterned along dorso-ventral and anterior-posterior axes, and mature in suspension culture into organoids representing forebrain, midbrain, hindbrain, and neural retina. This unified and scalable approach provides a reproducible foundation for modeling human brain development, offering broad translational potential for mechanistic studies, disease modeling, and drug discovery.

developmental biology↗

bioRxiv2025

An extremophilic Nocardiopsis strain from Great Salt Lake expands the taxonomic range of mycolic acid biosynthesis

Mycolic acids, long-chain fatty acids that form the characteristic and relatively impermeable mycomembrane, have long been considered a defining chemotaxonomic feature of the order Mycobacteriales. Here, we report that Nocardiopsis bonnevillensis, a new type strain isolated from the hypersaline environment of Great Salt Lake, is the first organism outside of this order known to produce mycolic acids. Lipid profiling, acid-fast staining, isoniazid sensitivity, and genome mining confirmed hallmark features of mycolic acid biosynthesis. Evaluation of the strains metabolic capabilities led to the isolation of bonnevanoside, a thiophenyl nonulopyranoside reported here for the first time from a natural source. Moreover, additional Great Salt Lake-derived Nocardiopsis isolates exhibited acid-fast staining, suggesting that this trait may be more widespread than previously recognized. Altogether, these findings expand the taxonomic distribution of mycolic acid biosynthesis, challenge long-standing chemotaxonomic boundaries, and highlight the potential ecological significance of mycolate-containing envelopes in supporting bacterial survival in extreme environments.

microbiology↗

bioRxiv2025

Transfer learning enables discovery of sub-micromolar antibacterials for ESKAPE pathogens from ultra-large chemical spaces

The rise of antimicrobial resistance, especially among gram-negative ESKAPE pathogens, presents an urgent global health threat. However, the discovery of new antibiotics is hampered by sparse publicly available antibacterial data, complex bacterial defenses, and weak economic incentives. Here, we introduce a transfer learning framework using deep graph neural networks (DGNNs) to identify antibacterials from ultra-large chemical libraries. DGNNs were first pre-trained on large molecular datasets of protein-ligand simulations, binding affinities, and physicochemical properties to learn generalizable chemical features, and then fine-tuned on limited antibacterial screening data. Compared to classical methods, transfer learning significantly improved enrichment factors and predictive performance in cross-dataset benchmarks. Applying this strategy to the ChemDiv and Enamine libraries, we virtually screened over a billion compounds and prioritized 156 candidates. Experimental testing against Escherichia coli revealed that 54% of compounds exhibited antibacterial activity (MIC [≤]64 {micro}g/ml), with several demonstrating sub-micromolar potency and broad-spectrum efficacy against gram-positive and gram-negative pathogens, including three ESKAPE species. Of 18 broad-spectrum candidates, 15 showed minimal cytotoxicity and no hemolytic activity. These results validate our approach for navigating underexplored chemical space and identifying potent, low-toxicity compounds with antibiotic activity. We release open-source models and a scalable workflow to accelerate antibacterial discovery in the face of data scarcity. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=137 SRC="FIGDIR/small/669197v1_ufig1.gif" ALT="Figure 1"> View larger version (39K): org.highwire.dtl.DTLVardef@d1e793org.highwire.dtl.DTLVardef@ba6c85org.highwire.dtl.DTLVardef@12ab303org.highwire.dtl.DTLVardef@4ea019_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2025

A cold-inducible phospholipid--protein interaction in brown fat mitochondria optimizes thermogenic capacity

Cold stress elicits dynamic remodeling of the mitochondrial lipidome in brown adipose tissue (BAT), marked by an increase in arachidonoyl-phosphatidylethanolamine (AA-PE). However, the function of membrane lipid rewiring in thermoregulatory physiology has been a longstanding mystery. Here, we identify LPCAT3 as a cold-regulated O-acyltransferase driving the highly selective accrual of AA-PE in BAT mitochondria. Lipid-based proteomics, molecular dynamics simulations, and bioenergetic analyses reveal that AA-PE partitions at the COX4I1 interface of the Cytochrome c oxidase complex, enhancing electron transport chain (ETC) efficiency. Accordingly, fat-specific Lpcat3-knockout mice have defects in respiratory-dependent BAT thermogenesis and cold tolerance, despite intact {beta}-adrenergic signaling and UCP1 function. Under cold acclimation, Lpcat3-/-BAT exhibits ETC dysfunction and activation of the integrated stress-response. Thus, our study illuminates a cold-regulated lipid-protein interaction as a gating factor in UCP1-dependent thermogenesis.

physiology↗

bioRxiv2025

Small molecule bioactivity benchmarks are often well-predicted by counting cells

Phenotypic profiling methods, such as Cell Painting and gene expression, have been widely used to predict compound bioactivity, often showing improvement over predictive models based on chemical structures alone. We discovered that a large subset of assays in widely-used benchmark datasets either directly relate to cell health and cytotoxicity or are assays intending to capture a more specific phenotype but whose active compounds impact cell count, while inactives do not. As a result, counting cells can achieve similar predictive performance as Cell Painting or gene expression data. Filtering benchmarks to include only assays relating to protein targets reveals that Cell Painting can capture information that cannot be predicted by mere cell counting. We re-evaluated three benchmark datasets used with Cell Painting data and observed that, in many cases, cell count models produced an AUC comparable to models using the full Cell Painting profiles. However, in protein-target-specific benchmarks across 17 distinct protein targets, Cell Painting features demonstrated unique predictive power, outperforming mean balanced accuracy from cell count models with a relative improvement of 19.6%. We propose five practical recommendations for benchmarking machine learning models for predicting bioactivity, including using cell count as a baseline feature. Although multi-class classification applications (such as matching samples based on their morphological profile) are less likely to be predictable by cell count than bioactivity benchmarks, these recommendations are broadly applicable to machine learning for drug discovery.

bioinformatics↗

bioRxiv2025

Identification of novel candidate regulators of human naive pluripotency by means of a genetic switch utilizing the chimeric receptor G-CSFR:GP130

This study investigated the molecular mechanisms underpinning the transition from primed to naive pluripotency in human pluripotent stem cells (hPSCs) by utilizing a dual genetic switch combining hormone-dependent STAT3-ERT2 and a chimeric GCSFR:gp130 receptor. Upon activation with tamoxifen and G-CSF, hPSCs demonstrated upregulation of naive markers, and sustained self-renewal independent of LIF. Transcriptomic analyses revealed distinct clusters of early response genes, with a subset of 26 genes significantly enriched in the human epiblast. Knockdown experiments further delineated essential regulators that modulate differentiation and proliferation during naive reprogramming including interferon gamma inducible protein (IFI)16 and interferon-induced transmembrane (IFITM) proteins. Moreover, functional assays using mutant chimeric GCSFR:gp130 receptor highlighted the critical role of JAK kinase recruitment in orchestrating epigenomic remodeling and early gene activation. This approach establishes a robust platform to dissect the signaling dynamics governing the accession of hPSCs to naive pluripotency.

cell biology↗

bioRxiv2025

Differential representations of spatial location by aperiodic and alpha oscillatory activity in working memory

Decades of research have shown working memory (WM) relies on sustained pre-frontal cortical activity and visual extrastriate activity, particularly in the alpha (8-12 Hz) frequency range. This alpha activity tracks the spatial location of WM items, even when spatial position is task-irrelevant and there is no stimulus currently being presented. Traditional analyses of putative oscillations using bandpass filters, however, conflate oscillations with non-oscillatory aperiodic activity. Here, we reanalyzed seven different human electroencephalography (EEG) visual WM datasets to test the hypothesis that aperiodic activity-which is thought to reflect the relative contributions of excitatory and inhibitory drive-plays a distinct role in visual WM from true alpha oscillations. To do this, we developed a novel, time-resolved spectral parameterization approach to disentangle oscillations from aperiodic activity during WM encoding and maintenance. Across all seven tasks, totaling 112 participants, we captured the representation of spatial location from total alpha power using an inverted encoding model (IEM), replicating traditional analyses. We then trained separate IEMs to estimate the strength of spatial location representation from aperiodic-adjusted alpha (reflecting just the oscillatory component) and aperiodic activity, and find that IEM performance improves for aperiodic-adjusted alpha compared to total alpha power that blends the two signals. We also discover a novel role for aperiodic activity, where IEM performance trained on aperiodic activity is highest during stimulus presentation, but not during the WM maintenance period. Our results emphasize the importance of controlling for aperiodic activity when studying neural oscillations while uncovering a novel functional role for aperiodic activity in the encoding of visual WM information. Significance statementWorking memory is a crucial component of cognition, yet its neural mechanisms are not fully understood. Research shows that alpha activity - presumed to reflect neural oscillations - tracks the location of items we hold in memory. However, these analyses assume that all alpha power is oscillatory, even though oscillations are mixed with non-oscillatory, aperiodic activity that may be physiologically and functionally distinct. Here, we use a novel analytical approach for separating alpha oscillations and aperiodic activity dynamically across time. Our results reveal distinct roles for each in human visual working memory: aperiodic activity encodes the spatial location of information whereas alpha oscillations maintain the location of that information.

neuroscience↗

bioRxiv2025

Divergent granulopoiesis at extramedullary sites safeguards host defense

Extramedullary organs such as the spleen can assume granulopoiesis as a supportive mechanism to cope with the demands during persistent inflammation. However, the quantitative output of extramedullary granulopoiesis is limited, thus raising the question if the spleen in fact provides neutrophils of a qualitative difference rather than merely contributing to neutrophil numbers. Here we report splenic stress granulopoiesis with distinct production and differentiation trajectories. Myeloid progenitors in the spleen engage in accelerated production of neutrophils with an immature phenotype. Yet, neutrophils generated during persistent stress granulopoiesis are fully competent to exert antimicrobial functions and are necessary to contain bacterial invasion. Activation of type I interferon signaling in the spleen is required for splenic neutrophil production and its loss impairs host defense. Thus, the spleen provides an immunological environment for stress-induced rapid production and priming of highly active neutrophils to meet the demands during infection.

immunology↗

bioRxiv2025

A multiomic network approach uncovers disease modifying mechanisms of inborn errors of metabolism

For many inborn errors of metabolism (IEM) the understanding of disease mechanisms remains limited in part explaining their unmet medical needs. We hypothesize that the expressivity of IEM disease phenotypes is affected by the activity of specific modifier pathways, which is controlled by rare and common polygenic variation. To identify these modulating pathways, we used RNA sequencing to generate molecular signatures of IEM in disease relevant tissues. We then integrated these disease signatures with multiomic data and gene regulatory networks generated from animal and human populations without overt IEM. We identified and subsequently validated glucocorticoid signaling as a candidate modifier of mitochondrial fatty acid oxidation disorders, and we re-capitulated complement signaling as a modifier of inflammation in Gaucher disease. Our work describes a novel approach that can overcome the rare disease-rare data dilemma, and reveal new IEM pathophysiology and potential drug targets using multiomics data in seemingly healthy populations. Take-home message: Multiomic data and gene regulatory networks derived from animal and human populations without overt IEM can be used to study disease mechanisms in inborn errors of metabolism.

genetics↗

bioRxiv2025

Cryo-EM reveals a new allosteric binding site at the M5 mAChR

The M5 muscarinic acetylcholine receptor (M5 mAChR) represents a promising therapeutic target for neurological disorders. However, the high conservation of its orthosteric binding site has posed significant challenges for drug development. While selective positive allosteric modulators (PAMs) offer a potential solution, a structural understanding of the M5 mAChR and its allosteric binding sites has remained limited. Here, we present a 2.8 [A] cryo-electron microscopy structure of the M5 mAChR complexed with heterotrimeric Gq protein and the agonist iperoxo, completing the active-state structural characterization of the mAChR family. To identify the binding site of M5-selective PAMs, we implemented an integrated approach combining mutagenesis, pharmacological assays, structural biology, and molecular dynamics simulations. Our mutagenesis studies revealed that selective M5 PAMs bind outside previously characterized M5 mAChR allosteric sites. Subsequently, we obtained a 2.1 [A] structure of M5 mAChR co-bound with acetylcholine and the selective PAM VU6007678, revealing a novel allosteric pocket at the extrahelical interface between transmembrane domains 3 and 4 that was confirmed through mutagenesis and simulations. These findings demonstrate the diverse mechanisms of allosteric regulation in mAChRs and highlight the value of integrating pharmacological and structural approaches to identify novel allosteric binding sites.

pharmacology and toxicology↗

bioRxiv2024

SimMS: A GPU-Accelerated Cosine Similarity implementation for Tandem Mass Spectrometry

Untargeted metabolomics involves a large-scale comparison of the fragmentation pattern of a mass spectrum against a database containing known spectra. Given the number of comparisons involved, this step can be time-consuming. In this work, we present a GPU-accelerated cosine similarity implementation for Tandem Mass Spectrometry (MS) with approximately 1000-fold speedup compared to the MatchMS reference at a rate of 0.005% incorrect matches and a rate of 0.002% incorrect scores. We describe the underlying reasons for these errors and provide means to avoid them.

bioinformatics↗

bioRxiv2024

Step Forward Cross Validation for Bioactivity Prediction: Out of Distribution Validation in Drug Discovery

Recent advances in machine learning methods for materials science have significantly enhanced accurate predictions of the properties of novel materials. Here, we explore whether these advances can be adapted to drug discovery by addressing the problem of prospective validation - the assessment of the performance of a method on out-of-distribution data. First, we tested whether k-fold n-step forward cross-validation could improve the accuracy of out-of-distribution small molecule bioactivity predictions. We found that it is more helpful than conventional random split cross-validation in describing the accuracy of a model in real-world drug discovery settings. We also analyzed discovery yield and novelty error, finding that these two metrics provide an understanding of the applicability domain of models and an assessment of their ability to predict molecules with desirable bioactivity compared to other small molecules. Based on these results, we recommend incorporating a k-fold n-step forward cross-validation and these metrics when building state-of-the-art models for bioactivity prediction in drug discovery.

bioinformatics↗

bioRxiv2024

SLC7A5 is required for citrulline-dependent growth in arginine limited conditions

Tumor cells must optimize metabolite acquisition between synthesis and uptake from their surroundings. The tumor microenvironment is characterized by hypoxia, lactate accumulation, and depletion of many circulating metabolites, including amino acids such as arginine. We performed a metabolism-focused functional screen using CRISPR/Cas9 in a melanoma cell line to identify pathways and factors that enable tumor growth in an arginine-depleted environment. Our screen identified the SLC-family transporter SLC7A5 as required for growth, and we hypothesized that this protein functions as a high-affinity citrulline transporter. Citrulline, an essential precursor to arginine synthesis, is present in human serum at 40 M and supports localized arginine synthesis across diverse tissues. Using isotopic tracing experiments, we show that citrulline uptake and metabolism are dependent upon expression of this transporter. Pharmacological inhibition of SLC7A5 blocks growth in low arginine conditions across a diverse group of cancer cell lines. Loss of SLC7A5 reduces tumor growth and citrulline import in a mouse tumor model. Overall, we identify a conditionally essential role for SLC7A5 in arginine metabolism as a mediator of citrulline uptake, and we propose that SLC7A5-targeting therapeutic strategies in cancer may be especially effective in the context of arginine limitation. Key PointsO_LISLC7A5 is required for proliferation in arginine-free conditions when citrulline is present. C_LIO_LISLC7A5 loss impairs arginine metabolism. C_LIO_LICitrulline import is uniquely dependent on SLC7A5. C_LIO_LISmall molecule inhibitors of SLC7A5 can be paired with senolytic drugs to drive apoptosis. C_LIO_LISLC7A5 knockout decreases citrulline import in a xenograft model. C_LI

biochemistry↗