Biology subjects

Xia, Y.

Publications and source records attributed to Xia, Y..

At least 55 records · Page 3Linked to original sources

bioRxiv2026

Detection of Somatic Point Mutations Directly from Spatial Transcriptomics Enables in vivo Spatiotemporal Lineage Tracing

Spatial transcriptomics reveals tissue organization but lacks in vivo lineage-tracing methods applicable to humans. We introduce SpaceTracer, a computational framework that accurately detects somatic single-nucleotide variants (SNVs) directly from spatial transcriptomics data. By leveraging naturally occurring somatic SNVs, SpaceTracer reconstructs cellular phylogenies within native tissue architecture, enabling the mapping of lineage spread, migration, lineage-coupled expression changes and lineage-aware local interactions. Applied to human cutaneous squamous cell carcinoma, it traced tumor initiation and progressions, uncovered widespread pre-invasive migration of dedifferentiated epithelial cells and characterized mutant B cells migrating from tertiary lymphoid structures (TLS) into the tumor boundary. The framework also reconstructed developmental lineages across multiple tissues and identified tissue-resident mutant immune cells. SpaceTracer thus provides a perturbation-free platform for high-resolution spatiotemporal lineage tracing, offering a transformative tool for elucidating complex biological systems--especially tumor-immune ecosystems--with direct implications for advancing cancer immunotherapy.

bioinformatics↗

bioRxiv2026

PhyloSOLID: Robust phylogeny reconstruction from single-cell data despite inherent error and sparsity

While lineage tracing based on somatic mutations in single-cell sequencing data offers a powerful approach to reconstructing cellular histories in vivo, its reliability is fundamentally limited by pervasive technical artifacts--specifically, high error rates and data sparsity. These issues introduce false phylogenetic signals that corrupt tree topology and lead to spurious evolutionary conclusions. To overcome these limitations, we present PhyloSOLID, a phylogenetic algorithm designed to be inherently robust to these data imperfections. PhyloSOLID employs a progressive scaffolding strategy that begins with graph-based construction of a low-resolution, high-confidence backbone tree from reliable and uniformly covered mutations. This scaffold is then refined through the iterative integration of remaining data, guided by a Bayesian statistical model that penalizes phylogenetic inconsistencies to effectively separate the true evolutionary signal from technical artifacts. Benchmarking on both simulated datasets and multiple ground-truth datasets demonstrates that PhyloSOLID achieves superior lineage reconstruction accuracy over existing methods, for both single-cell RNA-seq and DNA-seq data. Additionally, a user-friendly web interface enables customized quality assessment, artifact removal, and interpretation of lineage structures. PhyloSOLID provides a powerful solution for decoding cellular evolution in developmental and disease contexts.

bioinformatics↗

bioRxiv2026

Unravelling genome-wide mosaic microsatellite mutations at single-cell resolution

Short tandem repeats (STRs), or microsatellites, are highly mutable genomic elements that modulate gene regulations and are implicated in a range of human diseases. However, detecting mosaic STR mutations at single-cell resolution remains challenging due to both technical and biological complexities. To address this, we developed BayesMonSTR, a robust algorithm that enables accurate detection of mosaic STR mutations. Using this tool in single-cell analysis of human tissues, we reveal an accumulation of longer mosaic STR insertions and deletions (indels) in aging mitotic and post-mitotic cells. Strikingly, prefrontal cortex (PFC) neurons accumulate a higher burden of STR mutations than B cells or lung epithelium, with aged neurons exhibiting a particularly pronounced increase in longer STR deletions. These mutations are enriched at transcription start sites (TSSs) and active enhancers of highly expressed genes. Our work establishes a foundation for genome-wide, hypothesis-free discovery of disease-associated mosaic STR mutations and reveals a previously unexplored landscape of mosaic STR variation in development and aging.

bioinformatics↗

bioRxiv2026

Optimal hearing aid design through restoration of the neural code

Hearing loss introduces complex distortions in the neural coding of sound that current hearing aids fail to address. Here, we combine electrophysiology and deep learning to identify novel sound processing strategies to correct these distortions. We use large-scale intracranial recordings from the gerbil inferior colliculus to train deep neural network models of neural coding (ICNets) to serve as in silico surrogates for brains with normal and impaired hearing. We then use the ICNets to train another network (AidNet) to act as an optimal hearing aid, providing the individualized sound processing required to elicit normal neural activity in impaired brains. We find that AidNet outperforms state-of-the-art hearing aid processing by a wide margin in correcting distortions in neural coding and deficits in simulated phoneme recognition. Much of this advantage was retained when AidNets parameters were swapped across animals with similar hearing thresholds, suggesting that AidNet can compensate for complex effects of hearing loss without direct access to neural recordings. These results demonstrate that closed-loop optimization can identify novel and generalizable strategies for neural restoration, providing a foundation for the development of more effective hearing aids as well as other sensory devices and neurotechnologies.

bioengineering↗

bioRxiv2026

ALK R1275Q mutation drives expansion of SCP-like cells during sympathoadrenal commitment and primes neuroblastoma initiation.

Neuroblastoma (NB) is a pediatric malignancy developing in the sympathoadrenal lineage of the neural crest, characterized by clinical heterogeneity ranging from spontaneous regression to poor outcomes. Activating mutations in the receptor tyrosine kinase anaplastic lymphoma kinase (ALK) are frequently observed in both sporadic and familial NB, yet the functional role of ALK in tumor initiation is not fully understood. Using a patient-derived human induced pluripotent stem cell (iPSC) model of sympathoadrenal development, we show that upon sympathoadrenal lineage commitment, ALK R1275Q, the most common hotspot mutation found in familial NB, sustain a proliferative, immature Schwann cell precursor (SCP)-like cell state with elevated ALK signaling and increased susceptibility to MYCN-driven transformation. While ALK-mutant cells alone did not form tumors in vivo, they cooperated with MYCN to accelerate tumor initiation, suggesting that ALK R1275Q creates a permissive but insufficient state for transformation. These findings define an ALK-driven cell progenitor-like state that facilitates the initiation of NB during embryonal development.

cancer biology↗

bioRxiv2026

Covalent Leader Peptide Probes Enable Proteome-Level Mapping of RiPP Enzymes beyond Biosynthetic Gene Clusters

Genome-based analyses have enabled widespread discovery and biosynthetic annotation of ribosomally synthesized and post-translationally modified peptides (RiPPs) pathways, yet often fail to capture all participating enzymes, particularly those encoded outside canonical biosynthetic gene clusters. In this study, leveraging the central role of leader peptides (LPs) in RiPP enzyme-substrate recognition, we developed a substrate-guided chemical-proteomic strategy using covalent LP probes to directly identify LP-binding enzymes from native proteomes. Using the LctA-LctM and PatE-LynD systems, we demonstrate that LP probes carrying a photocrosslinker engage enzymes at their authentic LP-recognition sites while preserving full catalytic competence. Application of this strategy to a model strain Streptomyces sparsogenes uncovered lanthipeptide synthetases encoded outside canonical biosynthetic gene clusters and revealed graded cross-cluster activities among multiple RiPP pathways. This work establishes a proof-of-concept model platform for proteome-level discovery of LP-binding RiPP enzymes and highlights cross-cluster enzymatic crosstalk as a latent mechanism for structural diversification.

biochemistry↗

bioRxiv2026

Phenotype-Guided In Silico Molecular Generation Using Large Language Models

Complex diseases often emerge from coordinated, system-level cellular state changes that are difficult to address with target centric drug discovery. Phenotypic drug discovery offers a principled alternative but remains constrained by the cost and scalability of pathologically relevant assays. Here we present GEMGen, a large language model-based framework that performs in silico phenotypic drug discovery by generating small molecules directly from transcriptomic representations of cellular states. GEMGen encodes desired phenotypic transitions as text-based representations of up- and down-regulated gene sets, enabling transferable modeling across experimental platforms and data modalities. Trained on large scale chemical perturbation data, GEMGen robustly identifies phenotype-oriented compounds and mechanistically related but structurally distinct candidates across multiple benchmarks. Applied to signatures induced by genetic perturbations, GEMGen produces small molecules that phenocopy gene knockdown effects and identifies chemically novel inhibitors, including previously unreported KEAP1 inhibitors that activate NRF2 signaling. Extending this approach to a disease relevant model of fibrosis, GEMGen generates compounds that reverse profibrotic transcriptional programs and cellular phenotypes. These results establish a scalable framework for translating transcriptomic phenotypes into candidate therapeutic molecules, enabling systems-level exploration of vast chemical space and offering a complementary in silico counterpart to physical phenotypic drug screens.

cell biology↗

bioRxiv2025

Riemannian Metric Learning for Alignment of Spatial Multiomics

Recent spatial technologies measure the transcriptome, epigenome, proteome, metabolome and other modalities from thousands of cells across a tissue. Most assays typically profile only one modality from a tissue slice, raising the question of how to align spatial data from heterogeneous feature spaces. While multiple approaches have been developed for multi-modal integration of single-cell datasets, few existing techniques perform spatial alignment across arbitrary modalities incorporating both spatial and feature information. We introduce Manifold Gromov-Wasserstein (MGW), a metric-learning framework that exploits the product structure of spatial multiomics to infer modality-specific Riemannian pull-back metrics with neural fields. MGW aligns Riemannian distances induced by these metrics via Gromov-Wasserstein optimal transport, yielding a hyperparameter-free cost across arbitrary modalities sharing a spatial base. The formulation enjoys theoretical invariances - including orthogonal transformations of the spatial and feature domains as well as global feature scalings. We demonstrate the advantages of MGW on multiple alignment tasks, including Stereo-Seq spatiotemporal transcriptomics of mouse embryo, Xenium and Visium spatial transcriptomics of colorectal cancer, and spatial metabolomics-transcriptomics from human striatum and kidney cancer. MGW recovers biologically meaningful correspondences and spatially coherent tissue structures, outperforming existing OT- and non-OT-based multi-modal baselines. Code availabilitySoftware is available at https://github.com/raphael-group/MGW

bioinformatics↗

bioRxiv2025

Low quality evidence dominates discussion of carbon benefits of alternative grazing strategies

There is enormous interest in utilizing alternative grazing strategies to improve rangeland condition, increase profitability and decrease the carbon footprint of livestock production via soil organic carbon (SOC) sequestration. Here we present a systematic review and meta-analysis of the literature on alternative grazing strategies and their impact on SOC. Most studies (47 out of 70) failed multiple quality criteria. The 10 studies with 27 observations that met all inclusion criteria showed no change in SOC. Further dividing these observations by study design (controlled small-plot experiment versus observational paired site study) or grazing management style (adaptive or prescribed) or by aridity class or by grassland type failed to reveal any trends. Gains in SOC were only found for a secondary group of 13 studies with 25 observations, primarily composed of paired comparisons where there were unresolved questions about the quality of the site pairings for supporting causal inference. This divergence in results between primary and secondary studies highlights that low-quality evidence dominates the discussion around the climate benefits of alterative grazing strategies, underscoring a critical need for stronger evidence before asserting climate change mitigation benefits from alternative grazing practices. For submission toCommunications Earth & Environment

ecology↗

bioRxiv2025

Super-CUT&Tag: A Sensitive, Spatially Resolved Approach for Epigenomic Profiling in Tissue Sections

Spatial profiling of protein-DNA interactions is essential for understanding gene regulatory processes, but current methods suffer from low sensitivity, limited resolution, and complex workflows. Here we introduce Super-CUT&Tag (Solid-phase interface embedded with an Ultra-dense DNA barcode array to Profile Epigenomic landscapes for spatially Resolved CUT&Tag), a robust spatial epigenomic method that directly transfers in situ protein-DNA interactions from tissue sections onto a pre-fabricated spatially barcoded array. Incorporating 3D polyamidoamine (PAMAM) dendrimers markedly improves chromatin capture efficiency, achieving several-fold higher sensitivity compared to current spatial CUT&Tag approaches. Using Super-CUT&Tag, we profile H3K27ac spatial landscape and uncover previously hidden active-chromatin heterogeneity across mouse embryonic tissues and stages. Integration with spatial transcriptomics further reveals spatiotemporal enhancer regulation dynamics of Neurod2 during corticogenesis. By overcoming key bottlenecks in sensitivity, resolution and usability, Super-CUT&Tag establishes a powerful platform for spatial epigenomic profiling in complex tissues.

genetics↗

bioRxiv2025

Disrupted tRNA modification leads to intestinal mitochondrial dysfunction and microbial dysbiosis

Background and aimsTransfer RNA (tRNA) modifications determine translation fidelity and efficiency. It occurs through the action of specific enzymes that modify the nucleotides within the tRNA molecule. Our previous study demonstrated tRNA modopathies and altered queuine-related metabolites in inflammatory bowel diseases. Queuine tRNA-ribosyltransferase catalytic subunit 1 (QTRT1) and QTRT 2 co-localize in mitochondria and form a heterodimeric TGT participating in tRNA Queuosine (tRNA-Q) modification. Human body acquires Queuine/Vitamin Q from intestinal microbiota or from diet. However, the roles of tRNA-Q modifications in the maintenance of intestinal mitochondrial homeostasis and microbiome are still unclear. MethodsWe used publicly available human IBD datasets, QTRT1 knockout (KO) mice, QTRT1 intestinal epithelial conditional KO (QTRT1{Delta}IEC) mice, cultured cell lines with QTRT1-specific siRNA, and organoids from patients with IBD to investigate the mechanism of tRNA-Q modifications in intestinal mitochondrial homeostasis and therapeutic potential in anti-inflammation. ResultsIn single cell RNA sequencing datasets of human IBD, we identified significant reduced intestinal epithelial QTRT1 expression in the patients with Crohns Disease. Using publicly available datasets, we identified significantly changes of Vitamin Q-associated bacteria in human IBD, compared to the healthy control. Qtrt1-/- mice had significant reduction of Q-associated bacteria, e.g., Bacteroides. Alcian Blue and Mucin-2 staining revealed mucosal barrier damage and disrupted homeostasis, with reduced colonic cell proliferation. Intestinal tight junction integrity was impaired in QTRT1-KO mice, as evidenced by reduced ZO-1 and increased Claudin-10 expression. QTRT1{Delta}IEC mice also showed dysbiosis and disrupted TJs. ATP synthesis was significantly decreased in the colon of QTRT1-KO mice, accompanied by severe mitochondrial dysfunction: reduced mitochondrial quality, Cytochrome-C release, and mitochondrial DNA (mtDNA) leakage. Mitochondrial dysfunction contributed to colonic cell death, as shown by elevated expressions of Cleaved Caspase-3 and Cleaved Caspase-1, increased BAX/Bcl-2 ratio, and positive TUNEL signals. Elevated CDC42, CD14, and CD4 levels in QTRT1-KO colon suggested mucosal immune activation and tissue repair responses. QTRT1-deficient CaCO2-BBE cells showed mitochondrial dysfunction. Cytochrome-C and mito-DNA release leading to cell death characterized by elevated expressions of Cleaved Caspase-3 and Caspase-1, increased BAX/Bcl-2 ratio, and higher apoptosis rate. Organoids isolated from patients with IBD showed reduced levels of QTRT1 and dysfunctional mitochondria. Restoring mitochondrial function leads to enhanced QTRT1. ConclusionsThese findings underscore the critical role of QTRT1/Q-tRNA modification in maintaining intestinal and microbial homeostasis. Mechanistically, QTRT1 loss impacts mitochondrial integrity and mucosal homeostasis. Our study highlights the novel roles of tRNA-Q modification in maintaining mucosal barriers and innate immunity in intestinal health. What is already known about this subject?O_LIEukaryotes acquire queuine (q), also known as Vitamin Q, as a micronutrient factor from intestinal microbiota or from diet. C_LIO_LIVitamin Q is needed for queuosine (Q) modification of tRNAs for the protein translation rate and fidelity. C_LIO_LIQueuine tRNA-ribosyltransferase catalytic subunit 1 (QTRT1) is reduced in human IBD. C_LIO_LIHowever, health consequences of disturbed availability of queuine and altered Q-tRNA modification in digestive diseases remain to be investigated. C_LI What are the new findings?O_LIQTRT1 deficiency leads to altered microbiome and reduced Vitamin Q-associated bacteria in human IBD and a QTRT1 KO animal model. C_LIO_LIQTRT1 protects the host against losing intestinal integrity during inflammation. C_LIO_LIQTRT1 localizes in mitochondria and plays novel functions by maintaining intestinal mitochondrial function. QTRT1 loss impacts tRNA modification in the intestine, linking to mitochondrial integrity and mucosal homeostasis. C_LIO_LIHuman IBD showed reduced levels of QTRT1 and dysfunctional mitochondria. Restoring mitochondrial function leads to enhanced QTRT1. C_LI How might it impact on clinical practice in the foreseeable future?Targeting tRNA-Q modification in enhancing mitochondrial function will be a novel method to maintain intestinal health.

cell biology↗

bioRxiv2025

Zebrafish otic vesicle and mouse epididymis as model systems for studying columnar epithelial cell division

Epithelial cell division maintains tissue architecture through coordinated nuclear migration, cell shape changes, and spindle orientation. In columnar epithelia, interkinetic nuclear migration (INM) involves apical nuclear translocation in G2 phase and basal return post-mitosis, yet its regulation remains incompletely understood in vertebrates, in part due to limited in vivo live-imaging models. In this methodological study, we adapted the zebrafish embryonic otic vesicle as an in vivo model to investigate cell division dynamics in simple columnar epithelium using high-resolution live imaging and genetic tools. We demonstrated that apical INM initiates in mid-to-late G2 and is driven by dynein, not myosin II. Mitotic rounding is achieved via actomyosin-mediated basolateral constriction while maintaining basal attachment. Inhibiting myosin II impairs rounding and planar division, causing apical retention of daughter cells, suggesting planar division ensures proper integration. We additionally analyzed the mouse epididymal epithelium, a simple columnar epithelial tissue, to allow cross-species comparison of nuclear migration dynamics. Together, these optimized in vivo vertebrate models uncover conserved and tissue-specific mechanisms underlying epithelial organization and function, and importantly, provide tools for deeper mechanistic dissection in the future.

cell biology↗

bioRxiv2025

Precision targeting of autoreactive B cells in systemic lupus erythematosus using anti-9G4 idiotope synthetic immune receptor T cells

Chimeric antigen receptor (CAR)-T cell therapies can induce drug-free remission in systemic lupus erythematosus (SLE), but indiscriminate B cell targeting causes immunosuppression, unnecessary infections, and cytokine toxicities that preclude widespread use. Here, we overcome this by targeting the 9G4 idiotope, a shared structural feature of pathogenic B cell receptors encoded by the IGHV4-34 gene. We engineered anti-9G4 CAR-T cells and chimeric TCR-T cells to selectively eliminate autoreactive B cells while preserving protective immunity. Both platforms eradicated autoreactive B cells and autoantibodies in vitro and in vivo, spared normal B cells, and markedly reduced cytokine release compared to conventional CAR-T cells. This precision extended to cold agglutinin disease and lymphoma. These findings establish a framework for IGHV idiotope-directed cellular therapies for treating autoimmune and neoplastic diseases while preserving immune competence.

immunology↗

bioRxiv2025

Finding the perfect promoter: Correlating single-cell transcriptome data with gene drive performance

Gene drive can control pathogen transmission or suppress vector populations by spreading drive alleles with super-Mendelian inheritance. CRISPR homing drive currently represents the most powerful type, and regulating Cas9 expression with specific promoters has been effective for improving drive performance. However, selecting these is often a major challenge. Here, we evaluated 35 Cas9 constructs driven by distinct promoters in different gene drive systems and identified associations between drive performance and single-cell RNA expression patterns of the promoter-associated genes. Our results indicate that higher drive conversion is significantly associated with elevated expression of the promoter-associated gene in the respective reproductive cells, but embryo resistance allele formation correlates with excessive female germline expression. For males, early germline expression produces superior performance. Thus, we find that optimal drive performance requires restricting Cas9 expression to a tight quantitative and spatiotemporal window. In addition, found that in situ integrated rhino-Cas9 constructs significantly reduce somatic expression, underscoring the importance of genomic locus. On the basis of these results, we propose criteria for selecting promoters, providing a theoretical rationale and practical guidance for optimization of promoter elements in homing gene drive systems.

genetics↗

bioRxiv2025

FiGS-MoD: Feature-informed Gibbs Sampling Motif Discovery Algorithm for Mapping Human Signaling Networks

MotivationShort linear motifs (SLiMs) are short sequence patterns that mediate transient protein-protein interactions, often within disordered regions of proteins. SLiMs play central roles in signaling, trafficking, and post-translational regulation, but their short length and low complexity make them difficult to identify both experimentally and computationally. Since the latest release of motif discovery tools like MEME Suite, the availability of protein-protein interaction data (e.g., BioGRID) has increased by more than fivefold, providing richer network contexts where SLiMs can be inferred from recurring patterns of interaction. Combined with recent advances in machine learning, this creates new opportunities for large-scale, high-resolution motif discovery. ResultsWe present FiGS-MoD, a Feature-informed Gibbs Sampling Motif Discovery algorithm with two key innovations: (i) incorporating biased sampling informed by residue-level features, including Protein Language Model (PLM) embeddings, AlphaFold2-derived disorder and solvent accessibility, and evolutionary conservation, and (ii) replacing the traditional position-specific scoring matrix (PSSM) with a Hidden Markov model (HMM) to accommodate insertions and deletions. We applied our algorithm to 12,765 sub-networks from the human interactome and provided 221,840 human SLiM predictions and quality scores as a public resource, along with the tool itself. Our method outperformed MEME in terms of recovering known motifs from the Eukaryotic Linear Motif (ELM) database and phosphosites from PhosphoGRID. Through three case studies, we further highlight the biological relevance of our results and the generalizability of the method to diverse motif classes. Availability and ImplementationSource code and a predicted SLiM dataset using FiGS-MoD are freely available at https://github.com/Eric3939/FiGS-MoD

bioinformatics↗

bioRxiv2025

Redirection of SARS-CoV-2 to phagocytes by intranasal sACE2-Fc as a universal decoy confers complete prophylactic protection

The rapid evolution of SARS-CoV-2 and other respiratory RNA viruses limits the success of current vaccines and antibody-based therapies. Engineered decoy receptors based on soluble angiotensin-converting enzyme 2 (sACE2) offer promising alternatives. Clinical-grade recombinant sACE2 inhibits SARS-CoV-2 replication in vitro but shows limited clinical success. This study reports an optimized sACE2 mutant fused to human IgG1 Fc (B5-D3), which redirects virus-decoy complexes to lysosomal degradation in macrophages. Intranasal prophylactic delivery of B5-D3 confers complete protection in SARS-CoV-2-infected K18-hACE2 mice. Abrogation of Fc effector functions compromises antiviral protection, indicating that Fc-mediated uptake of virus-decoy complexes is critical. Transcriptomic analysis suggests that B5-D3 induces early immune activation in lungs of infected mice. Bio-distribution and flow cytometry reveal selective targeting of airway phagocytes. In vitro assays confirm lysosomal degradation of virus-decoy complexes by macrophages without productive infection. These findings reveal a distinct antiviral mechanism via phagocytic clearance, supporting refined regimens for decoy treatments against SARS-CoV-2 and potentially other respiratory viruses.

immunology↗

bioRxiv2025

Whole genome sequencing reveals the phylogenetic relationship and reticulate evolution of Phrynocephalus lizards in Eurasia

Toad-headed agama (genus Phrynocephalus) is a widely distributed group of lizards across Central Eurasia, inhabiting diverse arid and semi-arid environments from the Caspian Basin to the Qinghai-Tibetan Plateau (QTP). Despite considerable research efforts over the past decades, a robust phylogenetic framework for Phrynocephalus has remained elusive. In this study, we collected 26 samples representing a substantial portion of the distribution of this genus and established a well-supported phylogeny based on whole-genome sequencing and phylogenomic approaches. We found that the Phrynocephalus lizards comprised of three major clades, including the Western Asia, Central Asia and viviparous clades. Furthermore, we provided compelling evidence for an ancient hybridization event that led to the speciation of P. mystaceus, which likely contributed to the observed phylogenetic discordance across the genome. Divergence time estimation and ancestral area reconstruction indicated that Phrynocephalus lizards originated approximately 31.57 million years ago in the Central Asia and Iranian Plateau region, followed by Miocene dispersals into the Arabian Peninsula, the Gobi Desert, and the QTP. Our results illuminate how hybridization and historical range dynamics have jointly driven the origin and diversification of Phrynocephalus, providing new insights into the formation of biodiversity in Eurasia.

bioinformatics↗

bioRxiv2025

STARNet enables spatially resolved inference of gene regulatory networks from spatial multi-omics data

Biological tissues are composed of distinct microenvironments that spatially orchestrate gene expression and cell identity. However, the regulatory principles governing domain-specific cellular functions remain poorly understood due to the lack of effective methods for mapping gene regulatory networks (GRNs) in situ. To address this gap, we introduce STARNet, a representation learning approach that leverages heterogeneous hypergraph modeling of spatial transcriptomic and epigenomic data to resolve tissue-domain-specific regulatory interactions. By integrating graph neural networks with contrastive learning in a self-supervised framework, STARNet learns unified embeddings that preserve both multi-modal molecular features and anatomical spatial context, enabling accurate and domain-resolved GRN reconstruction within complex tissues. Benchmarking on both simulated and real datasets demonstrates that STARNet achieves state-of-the-art performance. We further demonstrate its broad applicability across diverse biological contexts, including neural development, genetic disease risk, and drug-induced developmental toxicity. In the mouse brain, it delineates region-specific regulatory networks and reconstructs spatiotemporal programs underlying neural stem cell differentiation. In human genetics, it provides a mechanistic link between genotypes and phenotypes by showing how genome-wide association study (GWAS) variants for complex diseases perturb hippocampus-specific GRNs. In developmental toxicology, STARNet reveals that drug-induced disruptions of GRNs in defined embryonic regions underlie tissue-specific vulnerability. Collectively, STARNet offers a powerful and versatile framework for resolving the spatial regulatory logic of complex tissues, providing multi-angle insights into tissue patterning, development, and disease mechanisms.

bioinformatics↗