Biology subjects

Yu, G.

Publications and source records attributed to Yu, G..

At least 19 recordsLinked to original sources

bioRxiv2026

Centered cross-context perturbation transfer on the public 2025 Virtual Cell Challenge benchmark

Predicting transcriptional responses across cellular contexts requires transferring both expression direction and differential-expression priority. We developed a compact numerical method that combines public perturbation summaries with a differential-expression kernel fitted to H1 CRISPRi training targets. The method centers an independent KOLF2.1J source on known training targets, transfers its response signs and magnitude ranks, and uses isotonic regression to render predicted expression means. We evaluated the frozen model retrospectively on the released public TEST100 of the 2025 Virtual Cell Challenge, using an external KOLF source released in 2026. The model achieved DES 0.371834, PDS 0.8614, and MAE 0.302206. With reconstructed historical anchors, its composite score was 0.341344, compared with the archived champion score of 0.340052, a difference of +0.001292. The comparison showed higher DES and lower PDS, with zero normalized MAE contributions for both methods. Source centering also improved development PDS across all three TRAIN folds and official VALIDATION. These results demonstrate a practical numerical approach to cross-context perturbation transfer and differential-expression ranking.

bioinformatics↗

bioRxiv2026

An immune world model with multi-agent orchestration for peptide vaccines

Peptide-vaccine design requires a connection between vaccine-induced priming and recognition of the antigen naturally displayed by a tumor. We present an immune world model that preserves this connection through separate vaccine and tumor branches, endpoint-specific evidence and multi-agent orchestration. Native-recognition constraints held an altered peptide despite its higher presentation score, and targeted perturbations removed antigen-dependent benefit when required tumor-recognition conditions were absent. Public dendritic-cell data distinguished donor from antigen-source transfer: a context ranker achieved Recall@20 of 0.483 and 0.333 under the respective holdouts. Structure-informed receptor prediction added candidate-ranking evidence while preserving functional constraints and matched scenario trajectories. An assay observation model further concentrated a cross-presentation coefficient around its known value in a synthetic conditioning experiment. Together, these results establish a computational framework for linking molecular predictions, native-target requirements and conditional immune dynamics. The framework provides a reproducible basis for choosing peptide-vaccine candidates and the experiments needed to resolve their biological dependencies.

bioinformatics↗

bioRxiv2026

Hidden grassland productivity from overlooked herbivore impacts

Grasslands cover approximately 40% of Earth's land surface and contribute substantially to terrestrial primary production. Despite widespread grazing, field estimates of grassland aboveground net primary productivity commonly overlook two grazing-related processes: biomass consumed by herbivores and post-grazing compensatory growth. The magnitude and direction of the resulting biases remain unclear. Here we quantified these biases using a 3-year yak grazing experiment on the Tibetan Plateau, a meta-analysis of 152 grazing experiments across China, and a global synthesis of 212 paired observations from 58 studies. Overlooking biomass consumed by herbivores consistently underestimated productivity by 34-51%. By contrast, overlooking grazing-induced compensation generated smaller but more variable biases, ranging from 29% underestimation to 19% overestimation, mainly reflecting contrasting compensatory responses under rotational versus continuous grazing. These findings show that overlooking grazing-mediated processes introduces substantial, process-dependent biases in grassland productivity estimates, underscoring the need to incorporate these processes into ecosystem and climate assessments.

ecology↗

bioRxiv2026

BioMod: Biological Module Tokenization for Single-Cell Foundation Modeling

Single-cell foundation models typically represent cells as flat sequences of individual genes, requiring higher-order biological organization to be inferred implicitly from sparse expression profiles. Similar to the evolution from character-level to semantic tokenization in natural language processing,we hypothesize that cellular states can be more efficiently represented using biologically meaningful intermediate units. Here we present BioMod, a 50-million-parameter single-cell foundation model that introduces biological modules as representation units between genes and cells. BioMod constructs an overlapping vocabulary of 660 biological modules derived from curated pathways and data-driven gene programs, replacing thousands of gene-level inputs with compact module representations. Pretrained on approximately 73 million human cells using high-ratio module masking with an asymmetric decoder, BioMod achieves state-of-the-art performance across cell-type annotation, zero-shot batch integration, perturbation-response prediction and drug-response forecasting while maintaining a compact architecture that enables efficient inference on a single GPU. Module-level analyses further demonstrate that BioMod captures biologically meaningful cellular programs and preserves pathway-level organization.

bioinformatics↗

bioRxiv2026

An astrocytic AMPAR-dependent parity signal preserves synaptic strength

Astrocytic processes closely associate with excitatory synapses and express AMPARs, positioning them to sense the same glutamate signals as neurons. Whether a mismatch between neuronal and astrocytic glutamate sensing affects synaptic physiology is unknown. Using AstroHTP, an extension of Drug Acutely Restricted by Tethering (DART) for cell-type-specific pharmacology, we found that acute astrocytic AMPAR inhibition spared neuronal AMPARs and basal transmission yet converted ongoing glutamatergic activity into a remarkably potent form of long-term depression (LTD) in mouse striatal projection neurons (SPNs). Unlike canonical presynaptic LTD, this form of LTD manifests postsynaptically, requires non-ionotropic NMDA receptor signaling, and arises when glutamate transmission occurs without a matching astrocytic signal. Astrocytic glutamate sensing through AMPARs, therefore, acts as a parity signal that verifies transmission and stabilizes active synapses.

neuroscience↗

bioRxiv2026

e^-2 Threshold for Bistability in Non-equilibrium Sorting of Biomolecular Condensates

Cells use ATP-driven chaperones to extract clients from biomolecular condensates selectively, yet no closed-form criterion links molecular parameters to condensate quality control. We derive a minimal non-equilibrium sorting reactor model that reduces the reaction-diffusion dynamics to one algebraic equation. In the idealized low-load limit, a concentration-dependent partition coefficient produces a saddle-node bifurcation between functional clearance and pathological storage. The critical Damk\"ohler number is \(\mathrm{Da}_{0c}=e^{-2}\) and is independent of the feedback strength. Corrections shift the value but preserve the topology. The model predicts a Kramers-type functional lifetime and recovery costs that scale as \(R^3\). Critical slowing down near the bifurcation provides early-warning signatures for FCS and FRAP.

biophysics↗

bioRxiv2026

NSD2 deficiency disrupts the epigenetic landscape of cochlear hair cells to enhance ECM expression and to perturb cochlear hair cell function

Mammalian cochlear hair cells (HC) are crucial for hearing, and histone modifications may play an important role in HC maturation and functions. However, whether and how NSD2, a histone H3 lysine 36 (H3K36) dimethyltransferase, affects cochlear HC maturation remains unclear. Here, we established an HC-specific Nsd2 knockout mouse model and discovered that the loss of NSD2 results in severe structural defects in stereociliary bundles and profound hearing impairment. Integrated multi-omics profiling revealed that NSD2 deficiency leads to a decreased expression of H3K36me2, which in turn alters genome-wide chromatin accessibility and gene expression. Mechanistically, NSD2 deficiency triggers the aberrant upregulation of extracellular matrix (ECM) genes at the transcriptional level, causing excessive collagen accumulation and compromising tissue structure. Crucially, ECM intervention can mitigate NSD2-deficiency mediated hearing damage. Our study reveals an epigenetic mechanism by which NSD2 regulates cochlear HC maturation through inhibition of chromatin remodeling-induced ECM accumulation, and provides a potential new therapeutic direction for hearing impairment.

cell biology↗

bioRxiv2026

Neuronal processing of facial expressions in the primate superior colliculus

Fast processing of emotionally salient facial expressions is widely believed to depend on a subcortical visual circuit from the superior colliculus through the thalamus to the amygdala, but the specific processing steps remain unclear. Here we identified two largely nonoverlapping groups of neurons in the superior colliculus significantly modulated by facial expressions. One group exhibited a short-latency preference for threat and the other exhibited a delayed preference for fear.

neuroscience↗

bioRxiv2026

Ageas enables time-agnostic cell fate inference from single-cell and spatial multi-omics data

Understanding cell fate decisions is fundamental to developmental biology and disease research. However, experimental lineage tracing requires genetic manipulation, which is impractical in many systems, particularly in humans. Computational approaches often rely on time-resolved measurements, which single-cell and spatial omics studies rarely provide. Here, we present Ageas, a time-agnostic transfer learning framework for cell fate inference from single-cell and spatial multi-omics data. Ageas overcomes these limitations by learning fate memory from terminal cell populations and transferring this information to progenitor or intermediate cells, enabling fate bias inference from static molecular snapshots. To enable robust generalization across molecular modalities, Ageas employs a data-adaptive ensemble strategy with automated model selection. In benchmark datasets with lineage-traced single-cell transcriptomic and epigenomic profiles, as well as spatial transcriptomics, Ageas achieves strong performance compared to existing methods. Applying Ageas to a 3D human embryo reveals a spatially organized anterior-posterior gradient of epiblast fate priming, with anterior epiblast cells biased toward ectodermal fates and posterior cells toward primitive streak-derived lineages, accompanied by regionally graded fate-associated regulatory programs. Together, these results establish Ageas as a general framework for decoding cell fate decisions from static molecular snapshots.

bioinformatics↗

bioRxiv2026

FOCUS-3D: Robust, generalizable volumetric cell segmentation for three-dimensional fluorescence microscopy

Understanding how cells establish spatial organization within tissues is a fundamental question in life sciences. While modern three-dimensional fluorescence microscopy captures large-volume tissue architecture, extracting quantitative cellular insights from complex volumetric datasets remains a major barrier. Here, we introduce FOCUS-3D, a robust, broadly generalizable volumetric cell segmentation framework built on a large, diverse manually annotated cell resource and advanced AI designs. Integrating volumetric representation learning, multi-scale feature extraction, and query-based mask prediction, FOCUS-3D achieves state-of-the-art performance across diverse species, tissues, fluorescent reporters and imaging modalities. During zebrafish (Danio rerio) development, FOCUS-3D uncovers three successive phases of notochord morphogenesis. We disentangle early motility-driven rearrangements from later cell shape remodeling and tissue repacking, and further link these morphological states to spatial and developmental transcriptional programs across independent datasets.

bioinformatics↗

bioRxiv2026

Expression and Engineering of Conductive Cytochrome Nanowires

Electrically conductive protein nanowires produced by metal-reducing bacteria have attracted interest as sustainable electronic materials, but their study and engineering have been limited by difficulties in expression, purification, and genetic manipulation. Here we establish Shewanella oneidensis as a heterologous host for production of Geobacter sulfurreducens OmcZ nanowires and progress a complementary in vitro assembly strategy that yields highly pure nanowire preparations. This platform enabled systematic engineering of OmcZ, revealing extensive tolerance to sequence variation and faciliating the design of enhanced-conductivity variants. Guided by comparative analysis of environmental OmcZ homologs, we generated a chimeric nanowire, OmcZ+, that exhibited a [~]3.5-fold increase in bulk conductivity while retaining the overall structure of the parent nanowire. Cryo-electron microscopy revealed unexpected architectural plasticity in OmcZ+ wires, including non-linear dendritic and pentameric assemblies mediated by the solvent-exposed heme VII, suggesting previously unrecognized modes of cytochrome nanowire organization. Incorporation of engineered OmcZ variants into water evaporation-induced electricity generators produced power densities up to [~]25.3 W cm-{superscript 2} and enabled high-performance operation in saline environments, including seawater and human sweat. Together, these results establish a versatile platform for the production, structural analysis, and engineering of cytochrome nanowires, providing a foundation for the development of programmable biological electronic materials.

bioengineering↗

bioRxiv2026

Integrated Metagenomics and Metabolomics Studies Reveal Core Bacterial Guild Regulating Carbohydrate Metabolism in Pediatric MASLD

BackgroundMetabolic dysfunction-associated steatotic liver disease (MASLD) is a prevalent pediatric disorder with limited treatment options, primarily due to an incomplete understanding of its molecular drivers. Recent research underscores the role of microbial guilds in metabolic health, but the mechanisms by which dysbiosis driven by core species and co-abundant symbionts disrupt metabolic homeostasis in pediatric MASLD remain unclear. ResultsHere, we conducted integrated metagenomic and metabolomic analyses on 285 pediatric subjects including MASLD patients, obese and healthy controls. The gut dysbiosis in MASLD was characterized by a depletion of Phocaeicola vulgatus, Bacteroides uniformis, Parabacteroides distasonis, and Bacteroides thetaiotaomicron. Co-abundance network analysis, integrating our cohort with four public datasets, identified these species as core guild members associated with MASLD. Microbial enrichment analysis showed significant disruptions in carbohydrate metabolism, particularly the downregulation of the tricarboxylic acid (TCA) cycle, fructose and sucrose metabolism, and pentose and glucuronate interconversions. P. vulgatus and B. uniformis were identified as dominant species linked to the downregulation of KEGG orthologs (KOs) in these disrupted pathways that were inversely correlated with hepatic injury biomarkers. CAZyme database analysis further emphasized P. vulgatus as the primary contributor to glycoside hydrolases involved in monosaccharide utilization. Finally, both untargeted and targeted metabolomics analysis validated a disrupted metabolic network centered on the TCA cycle and monosaccharide metabolism in pediatric MASLD. ConclusionOur findings suggest the core guild species P. vulgatus and B. uniformis may serve as critical regulators of carbohydrate metabolism in pediatric MASLD, offering potential mechanistic targets for gut microbiome-based interventions.

microbiology↗

bioRxiv2026

Visual activity in primate superior colliculus requires geniculostriate input

The superior colliculus (SC) is an ancient visual structure whose principle source of visual drive comes directly from the retina. In primates, however, the SC also receives geniculostriate input from primary visual cortex (V1) via the lateral geniculate nucleus (LGN), making it unclear which pathway normally drives visually evoked spiking. Here we tested whether visually evoked spiking in the primate SC depends on retinal signals routed through LGN and V1, rather than on direct retinal input alone. We recorded from macaque SC neurons before and during reversible inactivation of the ipsilateral LGN and found that LGN inactivation nearly abolished visually evoked spiking. This loss was not due to nonspecific suppression of SC, because saccade-related bursts were spared and was observed across SC layers. Magnocellular-biased stimuli, designed to reveal any potential direct retinal drive, failed to produce visually evoked spiking during LGN inactivation. Interhemispheric inhibition (i.e., a "Sprague effect") was ruled out, because contralateral SC silencing during LGN inactivation did not restore SC visual responses. Consistent with a geniculostriate input route to SC, V1 inactivation also reduced SC visual responses, with effects proportional to the overlap between the V1-induced scotoma and the stimulus representation. Together, these results show that visually evoked neural responses in primate SC depend on retinal signals routed through LGN and V1, and that direct retinotectal input is insufficient to drive SC spiking in the absence of geniculostriate input. These findings revise current models of visual drive to the primate SC and constrain theories of SC-dependent visual behavior. HighlightsO_LILGN inactivation nearly eliminates visual but spares movement-related activity in primate SC C_LIO_LINo retinotectal input is detected, even with magnocellular-biased stimuli or in SC input layers C_LIO_LIThe elimination of visual activity cannot be explained by interhemispheric inhibition C_LIO_LIDirect retinal projections are not sufficient to drive visual responses in primate SC C_LI

neuroscience↗

bioRxiv2026

STAnalyzer: Transparent Spatial Transcriptomics Analysis via an Agentic Architecture

Spatial transcriptomics enables high-resolution profiling of gene expression within spatial contexts, yet its potential is often hindered by fragmented toolchains, intricate parameters, and cognitive bottlenecks of interpreting high-dimensional data. While recent Large Language Model agents have attempted to automate this process, they remain constrained by rigid execution logic, lack multimodal feedback for self-correction, and operate in epistemic isolation from established biological knowledge. Here, we present STAnalyzer, an intelligent multi-agent framework designed to automate the end-to-end analytical lifecycle--from raw data processing to biological hypothesis generation. Transcending traditional pipelines, STAnalyzer employs a collaborative intelligence architecture to achieve three core capabilities: (1) Intent-Driven Orchestration, which dynamically translates natural language queries into rigorous bioinformatics workflows; (2) Multi-Modal Self-Refinement, which autonomously ensures analytical robustness through closed-loop synthesis of evidence from visual patterns and statistical metrics; and (3) Evidence-based Cross-Validation, which bridges the gap between data-driven correlations and biological causation by anchoring findings in ground-truth literature and structured databases. By eliminating manual analytical bottlenecks and ensuring rigorous evidentiary traceability and transparency, STAnalyzer makes high-resolution spatial omics more accessible to a broader research community. It provides a robust and scalable framework for cross-platform automated analysis and accelerated biological discovery, translating massive spatial datasets into verifiable biological insights.

bioinformatics↗

bioRxiv2026

High-Field Multinuclear MRI Reveals Sodium Relaxation Heterogeneity in Cortical Organoids

Sodium magnetic resonance imaging (23Na MRI) provides a unique opportunity to probe ionic microenvironments in neural tissue because sodium ions play central roles in membrane electrophysiology, ion transport, and cellular homeostasis. Unlike conventional proton ({superscript 1}H) MRI, which primarily reflects water distribution and tissue structure, {superscript 2}3Na MRI is sensitive to ionic compartmentation and quadrupolar interactions arising from the spin-3/2 nature of the sodium nucleus. However, sodium MRI remains technically challenging due to intrinsically low signal sensitivity and rapid biexponential relaxation, particularly when imaging small biological systems. Here, we establish a high-field multinuclear MRI platform for imaging human cerebral organoids at 14 Tesla. Cerebral organoids derived from human induced pluripotent stem cells provide a simplified three-dimensional neural tissue model that enables investigation of ionic microenvironments without vascular or systemic confounds. Using a dual-tuned {superscript 1}H/{superscript 2}3Na radiofrequency coil, we performed co-registered structural, diffusion, and sodium imaging of individual fixed organoids. High-resolution {superscript 1}H MRI (33-100 m) revealed pronounced microstructural heterogeneity, while multi-echo {superscript 2}3Na MRI (300-400 m) enabled voxel-wise characterization of quadrupolar relaxation behavior. Bi-exponential analysis of the sodium signal decay identified distinct relaxation components (T2*short {approx} 1 ms and T2*long {approx} 12 ms) and revealed spatial heterogeneity in sodium microenvironments across the organoid tissue. These results demonstrate the feasibility of quantitative sodium relaxometry in cortical organoids and establish a multinuclear imaging platform for investigating ionic microenvironment dynamics in three-dimensional neural tissue models.

bioengineering↗

bioRxiv2026

High-Fidelity Long-term Whole-embryo Lineage and Fate Reconstruction by Iterative Tracking with Error Correction

Reconstructing the complete cell lineages and fate maps of a living embryo has remained a central challenge in developmental biology. Here we introduce ITEC (Iterative Tracking with Error Correction), a fully unsupervised method that automatically reconstructs the lineage of every cell in the embryo with high fidelity. ITEC was validated with manually annotated lineages on four cross-species datasets including zebrafish, mouse, and Drosophila. We reconstructed the developmental lineages of a zebrafish embryo from terabyte-scale data of total 18.5 million cells with an estimated accuracy of over 99.7%. With ITEC, we revealed spatiotemporal dynamics of key morphogenetic processes such as somite boundary formation, demonstrated the various patterns of spatial sorting among adjacent organs or anatomical regions, and identified associations between cellular movement and spatial transcriptomics. ITEC provides a powerful platform for retrospective fate mapping and systematic exploration of developmental dynamics at the cellular scale and single-embryo level.

bioinformatics↗

bioRxiv2026

AQuA2-Cloud: a web platform for fluorescence bioimaging activity analysis

BackgroundAdvanced biological imaging analysis platforms such as Activity Quantification and Analysis (AQuA2) enable accurate spatiotemporal activity analysis across diverse cell populations within many species. These tools are increasingly important for investigating cellular signaling dynamics and behavior. However, despite advances in the accuracy and species capability of AQuA2, it remains computationally demanding for analysis of long time-series datasets and requires all users to maintain a MATLAB license, which may limit accessibility and large-scale deployment. ResultsTo address these limitations, we have designed and made available AQuA2-Cloud, a portable software stack and web platform developed as an improvement and further evolution of AQuA2. This container-deployable system permits multi-user cloud-based high accuracy activity quantification with intuitive workflows, export of analysis data and project files, and comparable processing times. The platform offers integrated features such as in-browser analysis control interfaces, asynchronous program state control, multiple users and user management, support for unreliable connections, file uploading and downloading via web browsers and File Transfer Protocol, and centralized organization of analysis output. ConclusionAQuA2-Cloud constitutes a cloud-native solution for laboratories or research groups seeking to centralize analysis of spatiotemporal biological imaging datasets while reducing software installation and licensing barriers for end users. The platform enables researchers with minimal technical expertise to perform advanced bioimaging analysis through standard web browsers while maintaining the analytical capabilities of AQuA2. AQuA2-Cloud source code, deployment procedures, and documentation are freely available at (https://github.com/yu-lab-vt/AQuA2-Cloud).

bioinformatics↗

bioRxiv2026

scDynOmics: An Optimized Transformer Model for Representation Learning from Single-Cell Multiomics

As foundation models have become increasingly prevalent in several fields for multiple purposes, pretraining models with single-cell transcriptomic data has gained significant interest. Although existing single-cell foundation models have demonstrated that transformer-based designs can be applied to various biological tasks, they do not show consistently competitive performance compared to much simpler approaches while requiring much more resources in terms of data and compute. Here, we introduce scDynOmics, a pretrainable multiomics-capable transformer for representation learning from single-cell data. The model is motivated by gene regulatory networks without excluding unknown interactions between genes and adopts a Linformer-style attention mechanism to scale to coding-genome wide multimodal inputs. Pretraining on paired single-cell transcriptomic and chromatin accessibility profiles yields compact high-fidelity embeddings that represent cellular states and developmental dynamics. For versatile application, scDynOmics employs low-rank adaptation modules, enabling parameter-efficient fine-tuning for downstream tasks. We demonstrate that scDynOmics outperforms existing single-cell foundation models by a large margin and achieves or surpasses state-of-the-art performance compared to simpler approaches, while revealing interpretable factors driving developmental trajectories and perturbation responses that simpler approaches cannot provide. Overall, scDynOmics is an efficient, scalable, flexible, and interpretable framework for cellular representation learning and deciphering cellular heterogeneity and dynamics.

bioinformatics↗