Biology subjects

Ding, J.

Publications and source records attributed to Ding, J..

At least 37 records · Page 2Linked to original sources

bioRxiv2026

Sex-specific hormone-sensitive regulatory architecture in adolescence as a scaffold for depression vulnerability

Adolescence is a critical developmental period during which sex differences in depression risk emerge, yet the underlying molecular mechanisms remain unclear. Here, we integrated longitudinal transcriptomic, epigenomic, and gene regulatory network analyses to characterize sex-specific regulatory dynamics of the brain across adolescence under normative conditions and elucidate how baseline molecular differences may shape later vulnerability. We observed that immediate early genes exhibited the strongest sex-biased signals, with males and females showing opposing developmental trajectories, suggesting a role in sex-specific stress-responsive programs. Across analyses, a consistent pattern emerged: males showed stronger activity-dependent neuronal signatures, whereas females showed stronger stress- and immune-related signatures. Extending these findings to a disease context using single-nucleus chromatin accessibility data, we identified hormone-responsive regulatory elements enriched in depression-associated regions, with sex- and cell type-specific links to genetic risk. Grounded in established models of sex-specific stress responses, these findings offer a developmental regulatory framework for understanding the differential presentation of, and vulnerability to, depression.

genomics↗

bioRxiv2026

AM-DMF-SCP-pro: An Ultrafast and Streamlined Nanolitre-Scale Lab-on-a-Chip Platform for Single-Cell Proteomics and Tumor Microenvironment Profiling

Single-cell proteomics (SCP) remains constrained by slow sample preparation and background interference in low-input, multistep workflows, limiting its broader practical use in complex biological samples. Here we report AM-DMF-SCP-pro, a streamlined active-matrix digital microfluidic workflow that enables rapid on-chip lysis and digestion in 6 nL droplets within a closed nanolitre-confined microenvironment. AM-DMF-SCP-pro reduces total sample-preparation time to as little as 30 min, including a 15 min digestion step, while maintaining quantitative reproducibility and digestion performance. Systematic optimization of mass spectrometry acquisition and nanolitre-scale digestion conditions enables deeper, more sensitive and lower-background proteome profiling from ultra-low peptide inputs. In murine tumor samples, AM-DMF-SCP-pro reliably resolves immune and tumor compartments, identifies treatment-associated tumor cell states and enables quantitative analysis of neutrophil-associated proteomic signatures. These results establish AM-DMF-SCP-pro as a practical workflow for rapid, low-background single-cell proteome profiling in complex biological samples.

systems biology↗

bioRxiv2026

ASFV early protein p30 suppresses antiviral type I IFN induction by targeting TRIM21 and RIG-I like receptor signaling adaptor MAVS

African swine fever (ASF) is a highly pathogenic disease caused by the African swine fever virus (ASFV) infection, which can affect pigs of all ages and breeds, posing significant threat to the global pig farming industry. The ASFV p30 protein is an early-expressed viral structural protein; however, its function is not fully understood. In this study, the interaction of viral p30 with host TRIM21 was identified. The ectopic TRIM21 inhibited ASFV replication, while knockdown or knockout of TRIM21 promoted ASFV replication. Further, p30 was found to interact with RIG-I-like receptor (RLR) signaling adaptor MAVS, and during ASFV infection, p30-TRIM21-MAVS interacted with each other. Mechanistically, TRIM21 activated the K27 polyubiquitination of MAVS to induce IRF3 mediated type I interferon (IFN) production, whereas p30 counteracted TRIM21 activated MAVS K27 polyubiquitination to evade RLR signaling mediated antiviral IFN induction. In summary, our study revealed a novel function of ASFV p30, and provided new insights into the immune evasion of ASFV.

immunology↗

bioRxiv2026

Profiling Peripheral Blood with an Optimized, Multiplexed, Single-cell Multiome Approach Supports an Insulin-driven Asthma Subtype

RNA sequencing (RNA-seq) and the Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) have become standard techniques for studying gene regulation in human populations. Single-cell (sc) "multiomic" genomic methodologies now enable researchers to dissect cellular heterogeneity while simultaneously measuring gene expression and chromatin accessibility within individual cells. However, single-cell approaches remain experimentally complex and cost-prohibitive, limiting their application in population studies, and motivating the development of new strategies for population-scale single-cell investigations. To this end, we have adapted and optimized a previous multiomic protocol, "Transcriptome, Epitope, and ATAC sequencing" (TEA-seq) through experimentation and simulation to incorporate sample multiplexing, thus resulting in our "multiplexed TEA-seq" (mTEA-seq) protocol. Using mTEA-seq, we sought to determine whether asthma that develops in conjunction with early-life elevated insulin levels might have an identifiable molecular signature. We studied samples from adult individuals (54 subjects, 272,003 cells) from the Tucson Childrens Respiratory Study (TCRS), a birth cohort phenotypically characterized over four decades, to identify unique molecular characteristics of blood cells from asthmatics who had high serum insulin levels at age 6. Using a Bayesian approach, we found striking sex-specific effects. Male asthmatic subjects with high insulin at age 6 displayed widespread immune transcriptional and epigenetic alterations into adulthood compared to male non-asthmatic subjects without elevated insulin at age 6. We also found that male non-asthmatics with early-life high insulin showed epigenetic perturbations in adulthood, but not transcriptional changes. The consistency of epigenetic signals between these two groups that had high insulin at age 6 was highly cell-type-specific. For example, CD14+ monocytes displayed broadly common insulin-associated chromatin remodeling regardless of asthma status, while NK cells exhibited unique patterns of insulin-associated epigenetic reprogramming depending on asthma status. Finally, genotyping performed directly from our single-cell data enabled cell type-specific cis-QTL mapping that suggested HLA-DQB1 and AHI as genes for future study in insulin-associated asthma. Our investigation of childhood insulin-associated asthma demonstrates a metabolically-driven alterations on immune cells persisting into adulthood, thus providing a molecular signature of this asthma subtype, and offering novel insights for disease prevention and therapeutic intervention.

genomics↗

bioRxiv2026

DeepSRFusion: a point cloud deep learning framework for super-resolution particle fusion

Deciphering the spatial organization of macromolecular complexes in their native context is central to structural biology. Particle fusion in single-molecule localization microscopy offers a unique capability for high-resolution structural reconstruction in situ. However, existing methods face significant challenges from large rotational perturbations and sparse labeling, resulting in compromised accuracy and substantial computational cost. We present DeepSRFusion, a self-supervised pretraining framework for three-dimensional super-resolution particle fusion. By representing single-molecule point clouds as Gaussian Mixture Models, DeepSRFusion integrates data-driven feature learning with physical imaging constraints. A two-stage optimization strategy with dynamic template updating enhances robustness, and a novel Clustering Error metric quantifies fusion quality. Nanometer-scale validation on both simulated and experimental datasets demonstrates high reconstruction fidelity and structural consistency with cryo-electron microscopy and AlphaFold3. DeepSRFusion remains effective under challenging imaging conditions, including large 3D rotations, sparse labeling, high localization uncertainty, and limited particle numbers, while achieving over 100-fold speedups compared to current methods. It resolves fine structural features with a measured spatial resolution of 1.60 {+/-} 0.10 nm, sufficient to distinguish ~10 nm spaced protein pairs and visualize tilted internal substructures within macromolecular assemblies. DeepSRFusion provides a powerful tool for high-precision structural analysis in native cellular environments.

biophysics↗

bioRxiv2026

PantheonOS: An Evolvable Multi-Agent Framework for Automatic Genomics Discovery

The convergence of large language model-powered autonomous agent systems and single-cell biology promises a paradigm shift in biomedical discovery. However, existing biological agent systems, building upon single-agent architectures, are narrowly specialized or overly general, limiting applications to routine analyses. We introduce PantheonOS (https://PantheonOS.stanford.edu), an evolvable, privacy-preserving multi-agent framework designed to reconcile generality with domain specificity. Critically, PantheonOS enables agentic code evolution, allowing evolving state-of-the-art batch correction and our reinforcement-learning augmented gene panel selection algorithms to achieve super-human performance. PantheonOS drives biological discoveries across systems: uncovering asymmetric paracrine Cer1-Nodal inhibition in proximal-distal axis formation of novel early mouse embryo 3D data; integrating human fetal heart multi-omics with whole-heart data to reveal molecular programs underpin heart diseases; and adaptively selecting virtual cell models to predict cardiac regulatory and perturbation effects. Together, PantheonOS points towards a future where scientific discoveries are increasingly driven by self-evolving AI systems across biology and beyond. Websitehttps://pantheonos.stanford.edu Ecosystemhttps://github.com/aristoteleo SummaryLarge language model-powered agent systems are driving a paradigm shift in scientific discovery by automating, scaling, and accelerating data analysis. This transformation is particularly profound in biology, where the rapid expansion of single-cell and spatial genomics has effectively reshaped the field into a data-intensive science. However, existing biological agent systems are typically constrained to single-agent designs, are narrowly specialized, or are overly general without sufficient domain expertise, limiting their applicability to routine or shallow analyses. Here, we introduce PantheonOS (https://pantheonos.stanford.edu), an evolvable, privacy-preserving, and general-purpose multi-agent framework designed to reconcile generality with deep domain specificity. PantheonOS provides an abstract, extensible architecture that enables customized agent composition and supports end-to-end single-cell and multi-omics analysis, spanning reinforcement-learning-augmented gene panel design, raw FASTQ processing, multimodal data integration, and three-dimensional spatial genomics reconstruction. Central to this framework, Pantheon-Evolve enables agentic code evolution, allowing the system to autonomously improve state-of-the-art batch-correction algorithms and new reinforcement-learning based gene panel design algorithms, achieving performance beyond manually designed baselines. We demonstrate the power of PantheonOS across multiple biological domains. In early mouse embryogenesis, PantheonOS automatically reconstructs three-dimensional spatial gene expression landscapes and resolves asymmetric Cer1 expression and paracrine Cer1-Nodal inhibition, revealing a robust proximal-distal axis at embryonic day six (E6.0). In human development, PantheonOS integrates fetal heart single-cell multi-omics with whole-heart 3D MERFISH+ data at post-conception week 12, uncovering spatially resolved molecular programs underlying heart disease ontogeny. Finally, an intelligent model-routing mechanism enables PantheonOS to adaptively select optimal virtual cell models across heterogeneous tasks, revealing minimal regulatory networks of cardiogenesis and predicting spatially resolved perturbation effects in the developing heart. Together, PantheonOS establishes a foundation for fully automated, evolvable, and domain-aware agentic analysis in genomics, and points toward a future in which scientific discovery is increasingly driven by self-improving AI systems across biology and beyond.

bioinformatics↗

bioRxiv2026

KRT17 stabilizes EPN1 via inhibiting SMURF1-mediated ubiquitination to modulate Wnt/β-catenin signaling output and stem-like traits in ovarian cancer

BackgroundOvarian cancer (OC) progression and chemoresistance are closely linked to dysregulated oncogenic signaling, including Wnt/{beta}-catenin pathways that contribute to cancer stem-like traits. However, the upstream mechanisms connecting cytoskeletal regulation to Wn/{beta}-catenin signaling in OC remain incompletely understood. Keratin 17 (KRT17), a type I intermediate filament protein, has been implicated in tumor progression, but its mechanistic role in OC requires clarification. MethodsGene Expression Omnibus (GEO) datasets and clinical specimens were analyzed to assess KRT17 and EPN1 expression and prognostic significance. Functional assays, xenograft models, co-immunoprecipitation, ubiquitination analyses, and rescue experiments with wild-type and ubiquitination-resistant EPN1 mutants were performed to investigate molecular mechanisms. ResultsKRT17 expression was elevated in OC tissues and correlated with poor patient survival. KRT17 depletion suppressed proliferation, migration, stem-like properties, tumor growth, and cisplatin resistance. Mechanistically, KRT17 interacted with EPN1 and weakened its association with the E3 ligase SMURF1, reducing SMURF1-mediated ubiquitination at lysine 107 and preventing proteasomal degradation. Stabilized EPN1 was associated with increased {beta}-catenin abundance and stemness-associated markers, and enhanced self-renewal capacity. ConclusionsThese findings identify a KRT17-EPN1 axis that links intermediate filament dynamics to ubiquitin-dependent regulation of EPN1 stability and Wnt/{beta}-catenin signaling outputs in ovarian cancer.

cancer biology↗

bioRxiv2026

Mutant FGFR3 restricts bone yet expands cortex via ERK-mediated self-repression

FGFR3 gain-of-function mutations cause achondroplasia 1,2, the most common form of dwarfism, yet trigger paradoxical cerebral overgrowth and skeletal stunting 3-9. Here, we demonstrate that Fgfr3-K644E mutant drives enhanced ERK activation, which then represses Fgfr3 expression and downregulates ERK signaling. In the developing cortex, this mechanism transiently expands cortical stem cells by activating ERK, then shifts signaling to downregulate Fgfr3/ERK/PKA and enhance YAP/TAZ, promoting cerebral overgrowth and premature ependymal maturation. In growing bones, Fgfr3-K644E mutant first elevates ERK to inhibit chondrocyte pre-hypertrophy, then downregulates Fgfr3/ERK/PKA and upregulates YAP/TAZ to accelerate premature hypertrophy and ossification. Our findings establish that FGFR3 mutations limit bone yet expand cortical size via an ERK negative feedback loop, revealing a unified disease mechanism and novel therapeutic targets.

developmental biology↗

bioRxiv2026

ProtoCloud: a Prototypical Self-explaining Model for Single-cell Analysis

Cell type annotation is a fundamental task in single-cell genomics. Although various methods have been developed for automatic cell type annotation, they often function as black-box models, making predictions without explaining their reasoning and lacking proper uncertainty estimation for their predictions. Furthermore, they often struggle to annotate rare cell types. We introduce ProtoCloud, a self-explaining deep generative model trained end-to-end to embed cells into a structured, low-dimensional space organized around cell type-specific prototypes. Coupled with a specifically-designed data augmentation strategy, it matches or outperforms existing methods in cell type annotation across 11 large-scale datasets, particularly for rare cell types. Moreover, ProtoCloud improves data annotation quality by identifying and re-annotating mis-annotated training cells through a built-in certainty quantification mechanism based on cell-prototype similarity. Finally, ProtoCloud provides interpretable predictions by identifying key genes that drive its classifications, facilitating the discovery of both known and novel cell type marker genes. Applied to a time-course dataset of post-injury retinal neurons, ProtoCloud successfully annotates previously unassigned cells; on the esophageal cell atlas, it identifies rare but potentially important cell populations and their marker genes relevant to esophageal inflammation.

genomics↗

bioRxiv2026

Inflammatory signaling differentially changes chromatin accessibility and gene expression of the PD- associated kinase LRRK2 between human and mice

The genomic locus that encodes the Leucine-rich repeat kinase 2 (LRRK2) gene is highly pleiotropic, being associated with both Parkinsons disease (PD) and Crohns disease (CD). Coding variants associated with risk of PD and CD act as gain of function kinase mutations increasing phosphorylation of RAB substrates, and non-coding variants in the promoter region of LRRK2 increase expression of the gene, notably in immune cells. If regulation of LRRK2 expression is a causal contributor to age-related diseases, it is important to understand the mechanism(s) by which LRRK2 is regulated, particularly in the context of inflammation. Here, we show that interferon-{gamma} exposure induces robust LRRK2 activation in human iPSC-derived microglia through signaling of the Janus-activated Kinase complex to phosphorylate STAT1, which then binds to the LRRK2 promoter and is associated with remodeling of chromatin structure in this genomic locus. Additional regulatory mechanisms include the stress-induced transcription factor and long non-coding RNA encoded at the same locus, resulting in increased LRRK2 mRNA levels. We also show evidence of the same effect in acutely cultured human brain slices. While we were unable to demonstrate any induction of Lrrk2 mRNA in the mouse brain, the introduction of a human bacterial artificial chromosome transgene into the mouse genome recapitulated sensitivity to interferon-{gamma} in microglia. A comparative genomic analysis across mammals suggests that these species differences are driven by regulatory regions upstream of LRRK2 that are specific to anthropoid primates. These results demonstrate that there are differences between species in how genes associated with human diseases are regulated and provide important information that should be incorporated in disease modeling.

neuroscience↗

bioRxiv2026

Dissecting and steering cell dynamics using spatially-informed RNA velocity with veloAgent

RNA velocity enables inference of cell state transitions from single-cell transcriptomics by modeling transcriptional dynamics from spliced and unspliced mRNA. However, existing methods overlook spatial context and struggle to scale to large datasets, limiting insights into tissue organization and dynamic processes. We introduce veloAgent, a deep generative and agent-based framework that estimates gene- and cell-specific transcriptional kinetics while integrating spatial information through agent-based simulations of local microenvironments. By leveraging both molecular and spatial cues, veloAgent improves velocity accuracy and achieves sublinear memory scaling, enabling efficient analysis of large and multi-batch spatial datasets. A distinctive feature of veloAgent is its in silico perturbation module, which allows targeted manipulation of spatial velocity vectors to simulate regulatory interventions and predict their impact on cell fate dynamics. These capabilities position veloAgent as a scalable and versatile framework for dissecting spatially resolved cellular dynamics and guiding cell fate manipulation across diverse biological processes.

bioinformatics↗

bioRxiv2026

A cohesin-centric gene regulatory network resource with regulatory site annotations

Cohesin is a central regulator of transcription and chromatin organization, binding to the majority of cis-regulatory elements (CREs) and mediating enhancer-promoter communication through three-dimensional genome architecture. Although gene regulatory networks (GRNs) provide an interpretable framework for modeling transcriptional regulation, most existing GRN analyses focus on transcription factor (TF)-gene relationships and largely ignore regulatory sites. Consequently, cohesin-associated regulatory contexts are rarely incorporated into GRN reconstruction. Here, we present a cohesin-centric gene regulatory network database that explicitly integrates TF binding, regulatory sites, and gene targets into unified TF-site-gene regulatory paths. Building upon our previously developed multiomics resource CohesinDB, we mapped TF binding to cohesin-associated regulatory sites and linked these sites to their downstream target genes. The resulting Cohesin-GRN module in CohesinDB (http://cohesindb.wangjklab.com/) (http://120.24.147.32/)comprises 61,222,502 TF-cohesin site links and 2,228,634 cohesin site-gene links, collectively forming over 270 million TF-site-gene regulatory paths. By enabling a CRE-informed and site-aware representation of gene regulation, Cohesin-GRN bridges conventional TF-gene GRNs with regulatory site-centric mechanisms. Given the pervasive roles of cohesin in enhancer activity, transcriptional control, and human disease, Cohesin-GRN provides a valuable resource for exploring transcriptional dysregulation and gene regulatory networks.

bioinformatics↗

bioRxiv2025

Developing High-Yield and Safe Therapeutic EVs by Ablating Tissue Factor-Mediated Toxicity

The clinical translation of extracellular vesicles (EVs) as next-generation drug delivery vehicles is currently stalled by two interrelated challenges: the difficulty of manufacturing EVs at an industrial scale ("the yield problem") and the risks of toxicity associated with systemic administration of high doses ("the safety ceiling"). In this comprehensive study, we present a two-pronged genetic engineering strategy to overcome these limitations. First, we discovered that overexpression of glycosylphosphatidylinositol (GPI)-anchored proteins, such as CD55 and CD59, significantly enhances EV production in HEK293T and Expi293F cells. By engineering a truncated CD55 variant (TR3) that retains the GPI-anchor but lacks functional domains, we developed the ExoBoost cell line, which boosts EVs yield by approximately 50-fold without altering vesicle morphology. Second, we observed acute toxicity in mice following intravenous administration of high-dose ExoBoost EVs, with symptoms resembling venous thromboembolism. Through comparative analysis of EVs derived from other cell types, including keratinocyte stem cells (KSCs) and mesenchymal stem cells (MSCs), we identified Tissue Factor (TF/F3) as the critical causative agent of this lethal response. Moreover, we completely abolished this toxicity by generating a F3-knockout cell line (ExoBoost2.0) using CRISPR/Cas9, thereby raising the safe dose to as high as 1.5E12 particles per mouse, as quantified by nano flow-cytometry (nFCM). Finally, using a novel CD46-nanoluciferase reporter system, we demonstrated that high-dose administration of these safe EVs saturates hepatic and splenic clearance pathways, leading to a dramatic increase of systemic EVs accumulation, particularly in hard-to-target tissues such as the brain ([~]10,000-fold) and muscle ([~]1,000-fold). Additionally, repeated high-dose administration of ExoBoost2.0 EVs did not upregulate inflammatory cytokines or production of IgG and IgM. In summary, these findings establish a scalable, safe, and highly biocompatible EVs platform, potentially revolutionizing the drug delivery system in clinical application.

bioengineering↗

bioRxiv2025

Single cell RNA sequencing reveals limited effects of Lrrk2 genotype in a mouse model of acute CNS inflammation

BACKGROUNDPrevious data implicates neuroinflammation in the pathogenesis of Parkinsons disease (PD). Of the genes associated with PD, Leucine rich repeat kinase 2 (LRRK2) has previously been proposed to play a role in neuroinflammation. However, the extent to which LRRK2 is involved in endogenous inflammatory signaling is unclear. OBJECTIVETo examine whether endogenous mutations in LRRK2 affect responses to neuroinflammation in vivo. METHODSWe injected cohorts of mice with knock-in mutations in Lrrk2, homologous to those causing human PD, with a single intrastriatal injection of lipopolysaccharide (LPS) or control. We used single cell RNA-Sequencing to examine cell type specific responses to treatment and genotype and validated key results with orthogonal approaches. RESULTSWe found that our chosen paradigm of acute LPS exposure evokes robust transcriptional changes consistent with a multicellular neuroinflammatory response. We also found evidence of peripheral immune cell recruitment into the brain and interaction with brain-resident microglia. However, the transcriptional effects of Lrrk2 mutations were limited to small numbers of genes, including down regulation of gene ontogeny terms related to lysosomes, in microglia. CONCLUSIONSOur data clearly demonstrate that many cells in the brain respond to a single inflammatory insult with strong transcriptional responses and that, even in a model focused on CNS injection, there is interaction between peripheral and central immune cells. In contrast, the quantitative effects of Lrrk2 mutations are modest at the transcriptional level, demonstrating that additional studies are needed to clarify whether Lrrk2 mutations affect neuroinflammation in an endogenous context.

neuroscience↗

bioRxiv2025

Substrates Coupling and Inhibition of the human Na+- dependent Cl-/HCO3-exchanger

Maintaining pH homeostasis is essential for neuronal excitability and for regulating cerebrospinal fluid (CSF). The Na+-dependent Cl-/HCO3- exchanger, NCBE (SLC4A10), abundant in the choroid plexus, contributes to CSF pH control by mediating bicarbonate flux, yet its molecular mechanism has remained unclear. Here we report cryo-EM structures of human NCBE in multiple functional states, revealing conserved binding sites for Na+, Cl-, and HCO3-, as well as an unexpected K+ site indicating relaxed cation selectivity. The inhibitor DIDS binds above the substrate cavity, locking NCBE in an outward-facing conformation and defining a subfamily-selective inhibitory mechanism. Functional assays and molecular dynamics simulations validate ion coordination and inhibitor effects. Comparison of cryo-EM structures with an AlphaFold3 inward-facing model supports an elevator-like mechanism and completes the conformational cycle of Na+-coupled HCO3- transporters. Together, these findings define the structural basis of NCBE and provide a framework for understanding ion coupling and pharmacological modulation within the SLC4 family.

biophysics↗

bioRxiv2025

Drosophila melanogaster model of RVCL-S demonstrates age dependent disease progression

Retinal vasculopathy with cerebral leukoencephalopathy and systemic manifestations (RVCL-S) is a disease that causes deterioration of small vessels, affecting various organs: eyes, brain, liver, and others. The RVCL-S carriers have lower life expectancy. There is no cure available to date. The disease has been linked to mutations in TREX1 gene disrupting its cytoplasmic localization. To facilitate the disease mechanism investigation, we employed model organism D. melanogaster, identified human TREX1 ortholog cg3165, and confirmed its vital significance to flies. Then, we expressed human TREX1 and its mutant form TREX1 V235Gfs in flies and used optical coherence microscopy (OCM) to monitor the dynamics of flies vascular system. We detected the relapse of fly dorsal vessel, movement impairment, and reduced longevity in TREX1 V235Gfs-expressing transgenic animals. Vascular deterioration and shorter life span recapitulate the RVCL-S manifestations in humans. We have established a robust quantitative Drosophila RVCL-S phenotypic system that can potentially serve as a screening platform for drug discovery and drug targets identification.

genetics↗

bioRxiv2025

scLinguist: A pre-trained hyena-based foundation model for cross-modality translation in single-cell multi-omics

Single-cell multi-omics provides complementary insights into cellular states, heterogeneity, and regulatory programmes. However, paired assays remain costly, low-throughput, and technically challenging, whereas large-scale single-modality data such as scRNA-seq are abundant but do not capture protein-level biology. Here we present scLinguist, a foundation model for cross-modality translation introduces a three-stage framework: self-supervised pretraining on large-scale unimodal datasets to learn modality-specific representations, post-pretraining on limited paired data to capture cross-modality relationships, and inference to predict missing modalities (e.g., protein from RNA) in fine-tuning or zero-shot settings. Systematic benchmarking shows that scLinguist consistently outperforms state-of-the-art methods in predicting protein abundance from RNA across diverse biological contexts. It achieves high predictive performance while preserving cellular heterogeneity and further enables mechanistic and generalizable inference under simulated gene perturbations. Furthermore, scLinguist exhibits strong transferability across health states and datasets. By leveraging abundant unimodal data and minimizing dependence on paired assays, scLinguist establishes a scalable and versatile framework for cross-modality translation in single-cell analysis.

cell biology↗

bioRxiv2025

NQO1 phase condensation promotes stress granule assembly to facilitate pancreatic carcinogenesis

G3BP1 promotes pancreatic ductal adenocarcinoma (PDAC) tumorigenesis driven by oncogenic KRAS mutations through liquid-liquid phase separation (LLPS)-mediated assembly of stress granules (SGs). However, the regulatory mechanisms remain elusive. We identify the antioxidant enzyme NAD(P)H quinone dehydrogenase 1 (NQO1) as a novel SG regulator that enhances SG assembly in pancreatic cancer cells independently of reactive oxygen species (ROS). Mechanistic studies reveal that NQO1 does not regulate the expression of G3BP1 or other SG-associated genes. Instead, NQO1 undergoes LLPS dependent on its RNA-binding K homology (KH) like domain. Further analysis demonstrates that residues 121-131 of NQO1 directly interact with G3BP1s RNA-binding domain (RBD), enhancing the multivalency of G3BP1 complexes to potentiate LLPS-driven SG assembly. This interaction accelerates cell proliferation, KRAS mutation-induced acinar-to-ductal metaplasia (ADM), and pancreatic carcinogenesis. Notably, the interaction between NQO1 residues 121-131 and G3BP1s RBD is essential for NQO1 phase condensation under stress condition. Integrative analysis of human PDAC transcriptomic datasets reveals a weak association between NQO1 and G3BP1 levels. Importantly, both NQO1 and G3BP1 formed biological condensates and co-localized in the lesions of human PDAC tissue sections. Our study uncovers a novel KRAS mutation-driven mechanism of pancreatic carcinogenesis through the lens of phase separation, transcending conventional gene expression regulation and offering new insights into non-canonical KRAS-NQO1-G3BP1-SG regulatory networks in PDAC initiation.

cancer biology↗