Biology subjects

Ding, J.

Publications and source records attributed to Ding, J..

At least 55 records · Page 3Linked to original sources

bioRxiv2025

High-Fat Diet-Induced Obesity Impairs Endothelium-Dependent Relaxation via MLCK Activation: Reversal by ML-7 in Rabbits

ObjectiveObesity is an independent risk factor for CVD(cardiovascular diseases) and elevated cardiovascular mortality, though the precise mechanisms remain incompletely elucidated. Impaired endothelium-dependent relaxation is an early manifestation of cardiovascular diseases. This study investigated the impact of HFD(high-fat diet)-induced obesity on endothelium-dependent relaxation function and explored the potential role of MLCK(myosin light chain kinase) signaling. MethodsForty-five New Zealand White rabbits were randomized into three groups: a control group (normal diet), HFD(high-fat diet) group, and an ML-7 group (HFD plus the MLCK inhibitor ML-7). After 8 weeks of feeding, vascular endothelial function was assessed in vivo by measuring FMD(flow-mediated dilation). Serum lipids and blood glucose levels were determined. Endothelium-dependent vasodilation was evaluated ex vivo using aortic ring assays. The protein expression levels of MLCK and eNOS(endothelial nitric oxide synthase), as well as the phosphorylation level of MLC(myosin light chain), in rabbit arteries were analyzed by Western blot. ResultsHFD-fed rabbits developed obesity, dyslipidemia (elevated TC[total cholesterol], LDL-C[low-density lipoprotein cholesterol], and ox-LDL[oxidized low-density lipoprotein]), perirenal adipose tissue deposition, and significant endothelial dysfunction, evidenced by impaired endothelium-dependent relaxation function and morphological endothelial damage. Western blot analysis revealed upregulated MLCK and eNOS expression, along with increased phosphorylation of MLC in the arterial walls of obese rabbits, despite no significant change in NO(nitric oxide) levels compared to controls. Treatment with ML-7 ameliorated endothelium-dependent relaxation function impairment. Furthermore, the inherent limitations of FMD as a non-invasive technique in terms of accuracy and sensitivity for evaluating endothelium-dependent dilation. ConclusionsObesity-induced endothelial dysfunction arise from visceral fat accumulation, dyslipidemia, oxidative stress, and MLCK overexpression. Notably, MLCK inhibition via ML-7 restores vascular function, highlighting its therapeutic potential in obesity-related vascular injury.

physiology↗

bioRxiv2025

The mevalonate pathway couples lipid metabolism to amino acid synthesis via ubiquinone-dependent redox control

The mevalonate pathway produces sterols and isoprenoids that support cancer cell growth, yet its broader metabolic functions remain incompletely defined. Here, we show that this pathway sustains amino acid biosynthesis by promoting mitochondrial NAD regeneration through ubiquinone-dependent electron transport. Statin-mediated inhibition of the mevalonate pathway impairs oxidative phosphorylation, lowers the NAD/NADH ratio, and suppresses de novo serine and aspartate synthesis, thereby activating the GCN2-eIF2-ATF4 amino acid deprivation response. The resulting depletion of serine-derived glycine and one-carbon units, together with reduced aspartate availability, limits purine and pyrimidine nucleotide production. Expression of the bacterial NADH oxidase LbNOX or the alternative oxidase AOX restores NAD levels and rescues statin-induced growth inhibition. These findings suggest that impaired NAD regeneration is a key mechanism contributing to the anti-proliferative activity of statins, linking the mevalonate pathway to mitochondrial electron transport- dependent control of amino acid metabolism. SignificanceThis study identifies the mevalonate pathway as a regulator of amino acid biosynthesis through mitochondrial electron transport-dependent NAD regeneration and reveals redox disruption as a key mechanism contributing to the anti-proliferative effects of statins.

cancer biology↗

bioRxiv2025

Nuclear polysaccharides maintain H3K9me3-heterochromatin and genomic stability

Polysaccharides are known to be synthesized by enzymes in the endoplasmic reticulum and Golgi apparatus and transported through the secretory pathway to the cell surface or extracellular space1, where they mediate essential biological processes2-6. While classical localization and functions of polysaccharides are well established, their presence and potential roles in the nucleus remain unclear. Here, we demonstrate that N-glycans, a type of polysaccharides, are present in the cell nucleus and modify inner nuclear membrane (INM) proteins across diverse cell types--a modification referred to as N-linked glycosylation (N-glycosylation). N-glycosylation is enriched in chromatin regions marked by H3K9me3 and long interspersed nuclear element-1 (LINE-1) retrotransposons. N-glycosylation inhibition and INM protein N-glycosylation site mutation both downregulate H3K9me3 within lamina-associated domains (LADs) and lead to genomic instability. Mechanistically, N-glycosylation regulates the interaction between the histone H3K9 methyltransferase SETDB1 and INM proteins, promotes the association of SETDB1 with the INM, and maintains H3K9me3. Moreover, we reveal that canonical N-glycan biosynthetic machinery in ER contributes to the N-glycosylation of INM proteins. These findings uncover a previously unrecognized nuclear role for polysaccharides, broadening our understanding beyond their traditional subcellular distributions and functional profiles.

molecular biology↗

bioRxiv2025

DANCE 2.0: Transforming single-cell analysis from black box to transparent workflow

Preprocessing is a critical step in single-cell data analysis, yet current practices remain largely a black-box, trial-and-error process driven by user intuition, legacy defaults, and ad hoc heuristics. The optimal combination of steps such as normalization, gene selection, and dimensionality reduction varies across tasks, model architectures, and dataset characteristics, hindering reproducibility and method development. We present DANCE 2.0, an automated and interpretable preprocessing platform featuring two key modules: the Method-Aware Preprocessing (MAP) module, which discovers optimal pipelines for task-specific methods via hierarchical search, and the Dataset-Aware Preprocessing (DAP) module, which recommends pipelines for new datasets via similarity-based matching to a reference atlas. Together, MAP and DAP execute over 325,000 pipeline searches across six major tasks - clustering, cell type annotation, imputation, joint embedding, spatial domain identification, and cell type deconvolution - yielding robust and generalizable recommendations. MAP-recommended pipelines consistently outperform original method defaults, with substantial gains across all tasks. Beyond automation, DANCE 2.0 reveals interpretable preprocessing patterns across tasks, methods, and datasets, transforming preprocessing into a transparent, data-driven process. All resources are openly available at https://github.com/OmicsML/dance to support broad community adoption and future methodological advances.

bioinformatics↗

bioRxiv2025

Self-limiting population suppression gene drive in the West Nile vector mosquito, Culex quinquefasciatus

Culex mosquitoes transmit major pathogens including West Nile virus, encephalitis, filariasis, and avian malaria, threatening public health, poultry, and ecosystems. We engineered a CRISPR-based population suppression gene drive targeting a conserved exon of the doublesex (dsx) gene. The drive incorporates a recoded dsxM segment to preserve male function while converting genetic females into sterile intersexes, enabling male-biased propagation and removal of fertile females. It achieves super-Mendelian inheritance ([~]71%) and generates partially dominant sterile resistance alleles via end-joining, resulting in intersex phenotypes with reduced fertility and hatchability. Modeling predicts that this RIDD (Release of Insects carrying a Dominant-sterile Drive) system can suppress populations at low intrinsic growth rates and release ratios, outperforming SIT and fs-RIDL strategies in persistence and efficiency, with further gains achievable through improved cleavage rates. This study establishes a self-limiting gene drive framework for Culex suppression, highlightling the potential of targeting conserved sex-determination pathways for sustainable vector control.

genetics↗

bioRxiv2025

Alveolar epithelial cell plasticity and injury memory in human pulmonary fibrosis

Acute and repetitive lung epithelial injury can lead to irreversible and even progressive pulmonary fibrosis; Idiopathic pulmonary fibrosis (IPF) is a fatal disease and quintessential example of this phenomenon. The composition of epithelial cells in human pulmonary fibrosis - irrespective of disease etiology - is marked by the presence of Aberrant Basaloid cells: an abnormal cell phenotype with pro-fibrotic and senescent features, localized to the surface of fibrotic lesions. Despite their relevance to human pulmonary fibrosis, the exotic molecular profile of Aberrant Basaloid cells has obscured their etiology, preventing insights into how or why these cells emerge with fibrosis. Here we identify cellular intermediaries between Aberrant Basaloid and normal alveolar epithelial cells in human IPF tissue. We track the emergence of Aberrant Basaloid cells from alveolar epithelial cells ex vivo and uncover a role for similar cells in epithelial regeneration under normal conditions. Lastly, we characterize the epigenetic changes that distinguish Aberrant Basaloid cells from their progenitors and identify hallmarks of AP-1 injury memory retention. This study elucidates the phenomenon of maladaptive epithelial plasticity and regeneration in pulmonary fibrosis and re-contextualizes therapeutic strategies for epithelial dysfunction.

cell biology↗

bioRxiv2025

DTractor enhances cell type deconvolution in spatial transcriptomics by integrating deep neural networks, transfer learning, and matrix factorization

Spatial transcriptomics (ST) captures gene expression with spatial context but lacks single-cell resolution. Single-cell RNA sequencing (scRNA-seq) offers high-resolution profiles without spatial information. Accurate spot-level decomposition requires effective integration of both. We present DTractor, a deep learning-based framework that improves cell-type deconvolution in ST data through spatial constraints and transfer learning. DTractor achieves dual utilization of scRNA-seq reference data by incorporating both a cell-type-specific gene expression matrix and learned latent embeddings into a unified matrix factorization model. This joint modeling enables accurate estimation of cell-type proportions and cell-type-resolved gene expression within each spatial spot, while preserving biological and spatial coherence. DTractor further applies spatial regularization to maintain local tissue structure. Across multiple ST platforms and tissue types, DTractor demonstrates improved decomposition accuracy, robustness, and interpretability compared to existing methods. The results from DTractor support downstream applications such as spatial domain analysis and the study of spatially organized cellular behaviors.

bioinformatics↗

bioRxiv2025

ERK and PKA signaling drive the evolutionary expansion of the cortex

The human cerebral cortex, essential for intelligence, cognition, and language, evolved its uniqueness through mechanisms that drive neuronal expansion. Here, we demonstrate that ERK and PKA signaling cooperatively preserve the neurogenic capacity of cortical radial glia (RGs) by inhibiting gliogenic YAP/TAZ and SHH signaling. YAP/TAZ signaling drives cortical RGs toward ependymal fate, while SHH signaling facilitates the generation of cortical tripotential intermediate progenitor cells that produce cortical astrocytes, oligodendrocytes, and olfactory bulb interneurons. During evolution, cortical outer radial glia (oRGs) acquire dominant ERK/PKA signaling through a self-reinforcing feedback mechanism that suppresses both YAP and SHH pathways, thereby significantly enhancing oRG self-renewal capacity and extending the neurogenic period. These findings reveal that cortical neurogenesis, gliogenesis, and evolutionary expansion are coordinately regulated by an integrated ERK, PKA, YAP/TAZ, and SHH signaling, identifying a unifying principle of mammalian cortical development and evolution.

developmental biology↗

bioRxiv2025

Spatial transcriptomics AI agent charts hPSC-pancreas maturation in vivo

Spatial transcriptomics has revolutionized our understanding of tissue organization by simultaneously capturing gene expression and spatial localization within intact tissues. However, analyzing these increasingly complex datasets requires specialized expertise across computational biology, statistics, and biological context. To address this challenge, we introduce the Spatial Transcriptomics AI Agent (STAgent), an autonomous multimodal agentic AI that integrates multimodal large language models (LLMs) with specialized computational tools to transform weeks-long analysis tasks into minutes of automated processing. Unlike conventional machine learning approaches that are limited to narrow, predefined tasks, STAgent leverages the emergent capabilities of multimodal LLMs - such as flexible reasoning, contextual understanding, and cross-modal integration - which allow it to adapt to novel data, execute multi-step analyses, and generate biologically meaningful insights with minimal human input. STAgent enables autonomous deep research through integrated capabilities, including dynamic code generation for complex analytical workflows, visual reasoning for interpreting spatial patterns, real-time retrieval of relevant peer-reviewd scientific literature, and synthesis of comprehensive, actionable reports. We applied STAgent to investigate the in vivo maturation of human stem cell-derived pancreatic cells (SC-pancreas) transplanted into immunodeficient mice. We generated single-cell spatial transcriptomics data spanning multiple developmental timepoints. STAgent autonomously (1) identified the maturation of initially scattered endocrine cells into well-defined islet-like structures, with predominantly peripheral -cells surrounding {beta}-cell cores supported by an expanding mesenchymal network; (2) revealed strengthening endocrine-endocrine cell interactions over time and, through context-aware gene set analysis, uncovered spatially resolved biological processes driving maturation; (3) unlike traditional analytical approaches, STAgent offers mechanistic explanations of spatial patterns, contextualizing findings with relevant literatures and developing cohesive insights into human pancreatic development. This agentic approach establishes a new paradigm in spatial transcriptomics analysis by substantially lowering the expertise barrier and reducing analysis time, accelerating biological and biomedical discovery.

genomics↗

bioRxiv2025

Longitudinal changes in functional connectivity networks in the first year following stroke

The functional organization of the brain consists of multiple subsystems, or modules, with dense functional communication within modules (i.e., visual, attention) and relatively sparse but vital communications between them. The two hemispheres also have strong functional communications, which presumably supports hemispheric lateralization and specialization. Subsequent to stroke, the functional organization undergoes neuroplastic changes over time. However, empirical longitudinal studies of human subjects are lacking. Here we analyzed three large-scale, whole-brain resting-state functional MRI connectivity measures: modularity, hemispheric symmetry (based on system segregation), and homotopic connectivity in a group of 17 participants at 1-month, 3-months, and 12-months after a single left-hemisphere stroke. These measures were also compared to a group of 13 age-matched healthy controls. The three measures exhibited different trajectories of change: (1) modularity steadily decreased across the 12-month period and became statistically inferior to control values at 12 months, indicating a less modular organization; (2) hemispheric symmetry values were abnormally low at 1-month and then increased significantly in the first 6 months, leveling off at levels not significantly below control levels by 12 months, suggesting that the two hemispheres diverged initially after the unilateral damage, but improved over time; and (3) homotopic connectivity exhibited a U-shaped function with a significant decrease from 1-6 months and then an increase from 6-12 months, to levels that were not significantly different from controls. The results revealed a complex picture of the dynamic changes the brain undergoes as it responds to abrupt onset damage.

neuroscience↗

bioRxiv2025

Distinguish risk genes functioning at presynaptic or postsynaptic regions and key connectomes associated with pathological α-synuclein spreading

Previous studies have suggested that pathological -synuclein (-Syn) mainly transmits along the neuronal network, but several key questions remain unanswered: (1) How many and which connections in the connectome are necessary for predicting the progression of pathological -Syn? (2) How to identify risk gene that affects pathology spreading functioning at presynaptic or postsynaptic regions, and are these genes enriched in different cell types? Here, we addressed these key questions with novel mathematical models. Strikingly, the spreading of pathological -Syn is predominantly determined by the key subnetworks composed of only 2% of the strongest connections in the connectome. We further explored the genes that are responsible for the selective vulnerability of different brain regions to transmission to distinguish the genes that play roles in presynaptic from those in postsynaptic regions. Those risk genes were significantly enriched in microglial cells of presynaptic regions and neurons of postsynaptic regions. Gene regulatory network analyses were then conducted to identify key drivers of genes responsible for selective vulnerability and overlapping with Parkinsons disease risk genes. By identifying and discriminating between key gene mediators of transmission operating at presynaptic and postsynaptic regions, our study has demonstrated for the first time that these are functionally distinct processes.

neuroscience↗

bioRxiv2025

Spike sorting AI agent

Spike sorting is a fundamental process for decoding neural activity, involving preprocessing, spike detection, feature extraction, clustering, and validation. However, conventional spike sorting methods are highly fragmented, labor-intensive, and heavily reliant on expert manual curation, limiting their scalability and reproducibility. This challenge has become more pressing with advances in neural recording technology, such as high-density Neuropixels for large-scale neural recording or flexible electrodes for long-term stable recording over months to years. The volume and complexity of these datasets make manual curation infeasible, requiring an automated and scalable solution. Here, we introduce SpikeAgent, a multimodal large language model (LLM)-based AI agent that automates and standardizes the entire spike sorting pipeline. Unlike traditional approaches, SpikeAgent integrates multiple LLM backends, coding functions, and established algorithms, autonomously performing spike sorting with reasoning-based decision-making and real-time interaction with intermediate results. It generates interpretable reports, providing transparent justifications for each sorting decision, enhancing transparency and reliability. We benchmarked SpikeAgent against human experts across various neural recording technology, demonstrating its versatility and ability to achieve curation consistency that are equal to, or even higher than human experts. It also drastically reduces the expertise barrier and accelerates the curation and validation time by orders of magnitude. Moreover, it enables automated interpretability of the neural spiking data, which cannot be achieved by any conventional methods. SpikeAgent presents a paradigm shift in processing signals for neuroscience and brain-computer interfaces, while laying the ground for AI agent-augmented science across various domains.

neuroscience↗

bioRxiv2025

Spatio-temporal distribution patterns and ecological correlates of new mammal records in China

AbstractSpecies geographic distributions are central to research in biogeography, macroecology, and conservation biology. However, incomplete or inaccurate knowledge about their spatiotemporal ranges--known as the Wallacean shortfall--hampers our understanding of biodiversity patterns and processes. In this study, we systematically reviewed 192 papers that reported new distribution records of Chinas mammals from 2001 to 2023, covering 150 species in 26 families across 7 orders. We examined the taxonomic, spatiotemporal, and biogeographic characteristics of these newly recorded species. We used Bayesian Phylogenetic Generalized Linear Mixed Models (BPGLMM) to assess how intrinsic traits and extrinsic factors influenced the likelihood of discoveries and partial Poisson regressions to evaluate how species-level attributes, environmental and socio-economic factors shaped the number of new occurrence records across provinces. Our results showed that Chiroptera (n=69), Eulipotyphla (n=26) and Rodentia (n=23) had the highest number of new records. Among provinces, Yunnan (n=31), Guangdong (n=22), and Xizang (n=18) yielded the most newly recorded species. The amount of annual discoveries varied and peaked between 2017 and 2021. The Western Mountain and Plateau (n=39) and the East Hilly Plain (n=33) zoogeographic subregions had the greatest number of newly recorded species, including two new records extending into historically unrecognized zoogeographic realms. Notably, 61 (40.7%) species were found to extend towards the north or northeast of their known ranges, and 32 (21.3%) eastward, mainly due to the sampling bias. Smaller-bodied species and those with broader habitat ranges are more likely to yield new records, and the total number of new records was positively correlated with regional species richness and current survey efforts (r > 0.5, p < 0.05). These findings expand the known distributions of Chinas mammals and provide essential data for mitigating the Wallacean shortfall. They underscore the urgent need for intensified surveys in biodiversity-rich, underexplored regions--especially targeting small-bodied, data-deficient taxa--and for timely updates and data-sharing to better inform conservation strategies.

ecology↗

bioRxiv2025

Depicting pathogenesis of osteomyelitis by single cell RNA-sequencing and an involvement of Morrbid in the autoinflammatory disease

Autoinflammatory diseases (AIDs) are defined as abnormal activation of the innate immune system leading to spontaneous and uncontrolled inflammation. The AIDs affect bone tissue and lead to chronic recurrent multifocal osteomyelitis (CRMO). However, the etiology and treatment of CRMO remain elusive. A mouse strain, Pstpip2cmo/cmo (cmo: chronic multifocal osteomyelitis), exhibits phenotypic characteristics similar to human CRMO. Morrbid is a long non-coding RNA gene and has been indicated in leukemogenesis in our previous studies. In this study, we demonstrated that Morrbid and Pstpip2 are co-expressed in mature myeloid cells and hypothesized a role of Morrbid in osteomyelitis. The Pstpip2-/- mice have the same phenotype as Pstpip2cmo/cmo, mimicking CRMO, while loss of Morrbid in Pstpip2-/-mice significantly inhibited the initiation and progression of CRMO symptoms, as well as the dysregulated activation of myeloid cells and the excessive release of inflammatory cytokines. Furthermore, single-cell RNA-sequencing (scRNA-seq) analysis from the Pstpip2-/- mice and the compound mutant mice supports that reduction of osteoclasts and inflammatory cells caused by Morrbid loss. The study systematically profiles the etiology of CRMO by scRNA-seq and warrants that inhibiting the lifespan of inflammatory myeloid cells by targeting Morrbid is an effective therapeutic strategy for osteomyelitis. HighlightsO_LIWe generated a frameshift mutant mouse strain Pstpip2-/-, which have a classic CRMO-like phenotype as same as Pstpip2cmo/cmo and could be used for testing various anti-inflammation perturbations. C_LIO_LILoss of Morrbid significantly inhibited the autoinflammatory symptoms in Pstpip2-/- mice, suggesting that Morrbid is a novel target for mitigating CRMO. C_LIO_LIscRNA-seq analysis of the affected bone marrow cells in Pstpip2-/-revealed the abnormalities of osteoclasts (OC), neutrophils (NE) and granulocyte macrophage progenitors (GMP) in both their fractions and inflammatory activities. C_LIO_LIUpon loss of Morrbid, we observed reduced composition and proliferation of OC and decreased activity of Nfkb2 and Rela in the compound mutants. C_LI

immunology↗

bioRxiv2025

Single-cell and spatial transcriptomics reveal the pathogenesis of chronic granulomatous disease in a natural model

Genetic defects in NOX2 can cause chronic granulomatous disease (CGD), characterized by increased susceptibility to infections and pronounced inflammatory responses that lead to granuloma formation. We developed a CGD model using Ncf2-/- mice through controlled environmental exposure. Unlike in specific pathogen-free environment, these mice spontaneously developed pulmonary granulomas under clean-grade conditions. Employing single-cell RNA sequencing, we observed an expansion of neutrophils and monocyte-derived macrophages (MDMs) within the lung tissue, identifying pro-inflammatory NOS2high-neutrophils and a distinct MDMs subset marked by NOS2 and ARG1. Spatial transcriptomics demonstrated that NOS2high-neutrophils were located at the core area, while the MDMs subset was positioned at the periphery, facilitating extracellular matrix remodeling. Pharmacological targeting of MIF, deletion of the pro-survival gene Morrbid, and elimination of Il1r1 all suppressed granuloma formation by mitigating inflammation. This study underscores how environmental control can establish a natural CGD model, elucidates the mechanisms of granuloma formation, and develops potent therapeutic interventions. HighlightsO_LIThrough installing a mutation in a NOX2 subunit gene Ncf2 and changing the husbandry environment, we generated a natural CGD mouse model with progressive pulmonary granuloma formation. C_LIO_LINOS2high neutrophils located in the core region of CGD lung granulomas, exhibit a pro-inflammatory transcriptional profile. C_LIO_LIMonocyte-derived macrophages marked by Nos2 and Arg1 are located at the periphery of granulomas and exhibit a profibrotic transcriptional profile. C_LIO_LIThe transcriptomic studies assist the development of three effective perturbations for treating the rare disease CGD via targeting Il1r1, Morrbid, and Mif respectively. C_LI

immunology↗

bioRxiv2025

Elevated SUN1 promotes aging-related polarity defects through mechanical coupling microtubules to the nuclear lamina

In migratory fibroblasts, the front-rear polarity required for cell migration is defined by an anterior centrosome relative to the nucleus. To achieve this polarity, actin cables drive the nucleus backward by coupling to nuclear membrane proteins nesprin-2G and SUN2. Aging disrupts this cell polarity by increasing the protein levels of SUN1, a SUN2 homolog. Here, we investigated the molecular mechanisms behind this disruption and found that the dominant negative effect of SUN1 and progerin, an aging-related lamin A variant, required direct SUN1-lamin A interaction. Microtubule interaction and force transmission through a nesprin, identified as nesprin-2, are crucial for SUN1s effect. We further discovered that stable microtubules are both necessary and sufficient to inhibit cell polarity. Using SUN1-SUN2 chimeric proteins, we demonstrated that the SUN domains determine their roles in cell polarization. Our findings reveal how elevated SUN1 disrupts cell polarity through coupling microtubule and nuclear lamina, emphasizing the impact of altered microtubule stability and nuclear mechanotransduction in aging.

cell biology↗

bioRxiv2025

CellUntangler: separating distinct biological signals in single-cell data with deep generative models

Single-cell RNA-seq data have provided new insights into intracellular and intercellular processes. Because multiple processes are active in each cell simultaneously, such as its cell type program, differentiation, the cell cycle, and environmental responses, their respective signals can confound one another, requiring methods that can separate and filter different complex biological signals. Each such signal is based on different gene activities and can define different relationships between cells. However, existing methods often focus on a single process or rely on overly restrictive assumptions, thus removing, rather than disentangling biological signals. Here, we develop CellUntangler, a deep generative model that embeds cells into a flexible latent space composed of multiple subspaces, each designed with an appropriate geometry to capture a distinct signal. We apply CellUntangler to datasets containing only cycling cells and both cycling and non-cycling cells, generating embeddings in which the cell cycle signal is disentangled from non-cell cycle specific signals, such as cell type or differentiation trajectory. We demonstrate CellUntanglers extensibility by using it to capture and separate spatial from non-spatial signals. With CellUntangler, we can obtain latent embeddings that capture various biological signals and perform enhancement or filtering at the gene expression level for downstream analyses.

bioinformatics↗

bioRxiv2025

Generating cognate epitope sequences of T-cell receptors with a generative transformer

Single-cell TCR sequencing enables high-resolution analysis of T Cell Receptor (TCR) diversity and clonality, offering valuable insights into immune responses and disease mechanisms. However, identifying cognate epitopes for individual TCRs requires complex and costly functional assays. We address this challenge with EpitopeGen, a largescale transformer model based on the GPT-2 architecture that generates potential cognate epitope sequences directly from TCR sequences. To overcome the scarcity of TCR-epitope binding pairs ({approx} 100, 000), EpitopeGen uses a semi-supervised learning method, termed BINDSEARCH, which searches over 70 billion potential pairs and incorporates high binding affinity predictions as pseudo-labels. To incorporate CD8+ T cell biology into the model as an inductive bias, EpitopeGen employs a novel data balancing method, termed Antigen Category Filter, that carefully controls antigen category ratios in its training dataset. EpitopeGen significantly outperforms baseline approaches, generating epitopes with high binding affinity, diversity, naturalness, and biophysical stability. Notably, the epitopes generated by EpitopeGen follow biologically plausible antigen category distributions, a crucial feature not achieved by other models. Using EpitopeGen, we directly identify subsets of clonally expanded tumorinfiltrating lymphocytes that recognize tumor-associated antigens, exhibiting elevated cytotoxicity and reduced exhaustion markers. From COVID-19 patients, EpitopeGen detects T cells that recognize COVID-19 spike proteins and non-structural proteins with distinct transcriptomic characteristics. In conclusion, EpitopeGen represents the first computational method that enables direct inference of antigen recognition profiles of CD8+ T cells from plain TCR repertoires.

immunology↗