Biology subjects

Gong, X.

Publications and source records attributed to Gong, X..

At least 19 recordsLinked to original sources

bioRxiv2026

Keratin 17 Defines an Epithelial-Mesenchymal Transition-Associated Migratory Cell State in Pancreatic Ductal Adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is characterized by early invasion, metastasis, and therapeutic resistance that contribute substantially to its poor clinical outcome. Keratin 17 (K17) is a recognized marker of the aggressive basal-like subtype of PDAC and an independent predictor of adverse prognosis; however, the biological basis for this association remains incompletely understood. To define K17-associated cellular programs, we integrated single-cell RNA sequencing (scRNA-seq) data from 42 patients, Clinical Proteomic Tumor Analysis Consortium (CPTAC) proteomic data from 140 tumors, MACSima multiplex spatial imaging of 63 patients, dual-plex immunohistochemistry of 60 cases, and three-dimensional spheroid migration assays. Single-cell transcriptomic analysis demonstrated significant enrichment of epithelial-mesenchymal transition (EMT)-associated gene programs among K17-positive malignant epithelial cells. Consistent with this finding, K17-positive tumors in the CPTAC cohort exhibited enrichment of the Hallmark EMT pathway at the protein level. Among six candidate markers identified through transcriptomic analyses, laminin subunit gamma-2 (LAMC2) was the only candidate associated with K17 at every validation step. Tumors positive for both K17 and LAMC2 had significantly worse overall survival than double-negative tumors (adjusted hazard ratio = 2.53). Multiplex spatial imaging localized K17-positive cells predominantly to poorly differentiated tumor regions and demonstrated frequent single-cell co-expression with LAMC2. Whole-slide dual-plex immunohistochemistry further confirmed the association between K17 and LAMC2 and showed progressive enrichment of K17/LAMC2-positive tumor cells with increasing intratumoral histologic grade. Functionally, gain- and loss-of-function studies in three-dimensional PDAC spheroids demonstrated that K17 promotes migratory outgrowth, and differences in proliferation did not account for this effect. Collectively, these findings identify K17 as a marker of a poorly differentiated, EMT-associated migratory cell state in PDAC and provide functional evidence that K17 contributes to tumor-cell migration.

cancer biology↗

bioRxiv2026

A J-protein-licensed CHD3 chromatin remodeling complex promotes H3K27me3 spreading to maintain cell identity in plants

Stable inheritance of epigenetic states is essential for cellular identity and developmental robustness and requires ATP-dependent chromatin remodeling, yet the molecular composition and regulatory mechanisms of plant CHD3 chromatin remodelers have remained unknown. Here we show that the Arabidopsis CHD3 remodeler PICKLE (PKL) assembles with a defined set of J-domain proteins (JDPs) to form stable CHD3 chromatin remodeling complexes in plants that are compositionally distinct from their animal counterparts. Loss of JDPs phenocopies pkl mutants, and JDPs are required for nucleosome densification and H3K27me3 spreading at thousands of Polycomb target genes. Mechanistically, JDPs perform dual functions by promoting PKL protein stability while directly stimulating PKL chromatin remodeling activity through their plant-specific DUF3444 domain that binds the N-terminal tail of histone H3. Deletion of the DUF3444 domain severely impairs PKL remodeling activity without disrupting complex assembly, leading to defective H3K27me3 spreading, impaired facultative heterochromatin maintenance, and loss of cell identity. Together, our findings define plant CHD3 chromatin remodeling complexes and reveal a mechanism that regulates CHD3-mediated chromatin remodeling to support the propagation of repressive chromatin states in plants.

plant biology↗

bioRxiv2026

Blood-Mediated Opto-Acoustic Stimulation of Brain Cortex at Sub-millimeter Precision

Direct modulation of neural activity with high spatiotemporal precision is a cornerstone in experimental neuroscience. Here, we present a blood-mediated optoacoustic stimulation (BOAS) approach that utilizes blood as an endogenous transducer for brain stimulation. By delivering 532-nm nanosecond pulsed laser to the cortex, we demonstrate that the absorption of hemoglobin generates sufficient acoustic pressure to trigger neuronal activity. By integrating BOAS with calcium imaging in GCaMP6f-expressing mice, localized neuronal responses were observed. Quantitative analysis reveals that BOAS produces responses comparable to natural visual stimulation and is significantly more efficient than the photothermal stimulation. Furthermore, we show that the response is dose-dependent. At high energy doses, BOAS induces cortical spreading depression. Histological evaluation confirmed that the brain maintains tissue integrity even under these stimulation parameters. Together, this work establishes a versatile method for precise brain stimulation as an alternative method for stimulating neuron at cortex.

bioengineering↗

bioRxiv2026

The Capillary endothelial cell-governed niche sustains angiogenesis in proliferating infantile hemangioma via multilayer stromal support

Infantile hemangioma (IH), the most common benign vascular tumor in infants, is characterized by the aberrant proliferation of hemangioma endothelial cells (HemECs) and excessive angiogenesis. However, it is unclear how the tumor microenvironment sustains these active angiogenic signals. We performed single-cell RNA sequencing and spatial transcriptomic to profile 30 samples from proliferating IH, involuting IH, and normal skin. The results indicated that HemECs comprised distinct subpopulations. Among them, capillary endothelial cells (Cap.ECs) had an immature phenotype and functioned as the core drivers of IH. The endothelial-to-mesenchymal transition endothelial cells (endoMT-ECs) secreted collagen and engaged in CD44-mediated crosstalk with Cap.ECs to construct the angiogenesis-active niche in the tumor microenvironment (TME) of proliferating IH. Blockade of CD44-mediated intercellular communication effectively suppresses the angiogenesis of HemECs and promotes tumor regression. Beyond the endothelial layer, pericytes tightly enveloped Cap.ECs, with SERPINE2+ subtypes sustaining their angiogenic activity via ANGPT signaling. Pseudotime trajectory analysis revealed that the mesenchymal stem cells (MSCs)-enriched niche serves as a progenitor pool for Cap.ECs in proliferating IH. Our findings thus revealed that IH progression relies on CD44-mediated endoMT-EC-Cap.EC crosstalk and on both pericyte and MSC niches to sustain angiogenesis. HIGHLIGHTSO_LICapillary endothelial cells drive infantile hemangioma progression, with pericytes tightly ensheathing them. C_LIO_LICD44-mediated capillary and endoMT-derived endothelial cells crosstalk construct the angiogenic-active niche in the tumor microenvironment of proliferating infantile hemangioma. C_LIO_LIThe MSC-enriched niche provides a progenitor pool for the capillary endothelial cells in proliferating infantile hemangioma. C_LI

cancer biology↗

bioRxiv2026

Internalized Components of Membrane Attack Complexes Disrupt Proteostasis and Acquire Alarmin-Like Properties

Immune effects of membrane attack complexes (MAC) have been widely attributed to their abilities to cause cell death. Here, we show that the MAC component, C9, forms non-cytolytic aggregates with pro-inflammatory effects. Intracellular aggregates of C9 are detected within inflamed tissues of patients in association with endothelial cell (EC) activation but not increased cell death. We identify NUMBL as a Rab35 effector that directly binds surface-bound C9 to promote C9 internalization and entry into the endolysosomal pathway. Within acidified endolysosomes, C9 forms insoluble aggregates that are targeted for degradative aggrephagy in a process that activates NF-{kappa}B. For C9 aggrephagy to occur, ZFYVE21, a Rab5 effector, complexes with RNF34 to bridge C9 aggregates to LC3B+ aggresome membranes. We detect C9 aggregates in vivo, and we show that a ZFYVE21-RNF34 signaling axis is required for C9 aggrephagy and NF-{kappa}B -dependent EC activation in three separate mouse models. Mice with conditional loss of ZFYVE21 in ECs show reduced aggregraphy, resulting in attenuated systemic inflammation and reduced tissue injury following skin transplantation. Our data show that the C9 component of MACs forms intracellular aggregates with alarmin-like properties.

biochemistry↗

bioRxiv2026

Distinguishing Photoacoustic and Photothermal Neuron Stimulation Through Quantitative Mapping Spatiotemporal Field Evolution

Neuromodulation is a rapidly advancing strategy for modulating brain function and treating neurological disorders, yet the mechanisms of optical neural stimulation remain incompletely understood. A central challenge is that photoacoustic (PA) and photothermal (PT) effects are typically generated simultaneously and evolve on overlapping spatial and temporal scales, making their contributions to neuronal activation difficult to distinguish. Here, we overcome this limitation by integrating two carbon-based emitters with a spatial-offset pump-probe imaging platform to distinguish PA and PT neuromodulation through direct mapping of spatiotemporal field evolution. The two emitters produce nearly identical thermal fields while exhibiting a 26-fold difference in acoustic output, enabling decoupled comparison of thermal and acoustic contributions. To directly characterize the physical fields, we establish a spatial-offset pump-probe imaging system capable of visualizing both temperature and pressure evolution, confirming matched thermal profiles between the two emitters. Neuronal stimulation experiments further show that neurons are markedly more responsive to PA stimulation than to PT stimulation. Beyond resolving a longstanding mechanistic question, this work demonstrates the enhanced efficiency of PA stimulation and provides general guidance for designing high-performance optical neural interfaces.

neuroscience↗

bioRxiv2026

CatIF-RL: Activity-Oriented Enzyme Sequence Design by Steered Inverse Protein Folding

Protein inverse folding models are designed to generate amino acid sequences compatible with a given backbone structure, but they are not explicitly optimized for specific biological functions. Here, we present CatIF-RL, a framework that steers a graph-based denoising diffusion inverse folding model toward designing enzyme variants with enhanced catalytic activity. CatIF-RL first adapts the inverse folding model to enzyme structural data, then introduces activity-oriented preference signals using predicted catalytic constant (kcat) as the optimization objective, enabling specialization through generative dataset curation and group-relative policy optimization (GRPO). This process iteratively shifts the sequence distribution toward higher predicted kcat while constraining sequence divergence to sequences that remain compatible with the input structure. On the independent benchmark, CatIF-RL achieves an approximately four-fold increase in predicted kcat relative to native enzymes, substantially outperforming representative inverse folding methods, while maintaining sequence recovery (0.55) and structural fidelity, and supporting motif-preserving partial sequence design. CatIF-RL establishes a practical framework for activity-oriented enzyme design and provides a generalizable strategy for steering structure-conditioned protein generation toward functional optimization.

bioinformatics↗

bioRxiv2026

Foxp4+ Mandibular Skeletal Stem Cells Orchestrate Bone/Tooth Development and Regeneration

The identity of stem/progenitor cells that drive mandible/tooth development and regeneration remains poorly understood. Here, we show that cranial neural crest-derived Foxp4+ mesenchymal cells condense to form Meckels cartilage, and simultaneously promote mandible development and tooth germ formation. In postnatal mandibles, Foxp4+ cells localize in the endosteum, periosteum and dental pulp, which undergo osteogenesis and odontogenesis under steady state, as well as chondrogenesis during fracture repair. Importantly, transplantation of embryonic or adult Foxp4+CD200+CD105- mandibular skeletal stem cells (mdSSCs) generate ectopic bone and tooth. In contrast, genetic ablation of Foxp4+ cells severely impairs mandible/tooth development and delays fracture healing. Mechanistically, we find that primary cilia critically regulate mdSSC maintenance and differentiation into bone and tooth. FOXP4+PDPN+CADM1+ mdSSCs also exist in adult human mandible, which generate ectopic bone and intact tooth upon transplantation into immunodeficient mice. Taken together, we identify Foxp4+ mdSSCs with potential therapeutic applications in regenerating mandibular bone and tooth.

cell biology↗

bioRxiv2026

RNABag: A Generalizable Transcriptome Foundation Model for Precision Oncology across Biopsy Modalities

Transcriptomic data is highly sensitive to cancer state and progression, making transcriptome-based foundation models a great promise for diverse clinical ontological inference. However, analyses of transcriptome are conventionally hindered by technical batch effects and limited generalization across platforms. Here, we introduce RNABag, a foundation model designed to generalize well to external datasets. In particular, the model only focuses on highly variable genes to reduce noise; and extensive data augmentation was utilized to pretrain RNABag to learn robust representations, invariant to batch variations. We demonstrate that RNABag achieves superior performance in pan-cancer tissue-of-origin classification and cancer detection in internal validation sets, as well as in zero-shot generalization to external cohorts and in-house clinical samples. Furthermore, RNABag, after specialized finetuning, exhibits strong capabilities in a wide range of clinical applications. The model effectively stratifies patient survival and predicts relapse risks, highlighting key molecular pathways driving tumor progression. Crucially, we extend RNABags utility to liquid biopsies, achieving high diagnostic accuracy in plasma cfRNA and tumor-educated platelets (TEPs), thereby supporting its application in non-invasive cancer monitoring. Interpretability analysis revealed pivotal role of tumor immune escape in the cancer induced plasma cfRNA signals. In summary, our study indicates that cancer states and progression may be monitored in details and precision via comprehensive modeling of transcriptome across biopsy modalities.

bioinformatics↗

bioRxiv2026

Cardiac oxidative stress monitoring enabled by hierarchical mechanical adaptation

Soft bioelectronics have advanced cardiac monitoring through electrophysiological tracking, yet this alone cannot resolve the metabolic pathology essential to surgical decision-making. However, real-time molecular sensing on beating hearts remains unresolved due to deformation-induced sensor failure and stress-induced metabolite artifacts. This challenge is exemplified by ischemia-reperfusion injury (IRI), a major cardiac surgery complication characterized by reactive oxidative species (ROS) bursts, where true pathological ROS signals being confounded by mechanotransduction-induced ROS artifacts. Herein, we propose an enzymatic cardiac oxidative stress biosensor (E-cardiac) with hierarchical mechanical adaptation: macro-scale biofluid-mediated contact, micro-scale fiber reorganization, and nano-scale enzymatic confinement within gold nanoarches dissipate interfacial stress. This produces ultrathin ([~]460 nm), soft (0.79 kPa) E-cardiac with robust electrochemical stability (100% strain), low detection limit (380 nM), rapid adhesion (<3 s), stable biosensing on beating heart, as well as minimal invasive deployment capability. Mechanical analysis and cellular studies confirm mitigated stress-induced ROS and absent PIEZO channel activation. Validated across cardiomyocytes, ex vivo tissues, multi-species ischemia models (mouse, rat, rabbit, pig), rat ischemia-reperfusion injury, and Langendorff hearts simulating graded perfusion deficits, E-cardiac quantitatively differentiates IRI severity (sham < ischemia < reperfusion) as well as detecting the "ECG blind window". The E-cardiac platform provides real-time metabolic feedback for surgical guidance during cardiac procedures, enabling timely intervention before irreversible damage.

bioengineering↗

bioRxiv2026

DeSCENT: Deconvolutional Single-Cell RNA-seq Enhances Transcriptome-based Cancer Survival Analysis

MotivationAccurate cancer survival prediction requires modeling tumor heterogeneity across both population and cell levels. Most cancer survival analyses use tumor transcriptomes only, since cohorts are usually measured with bulk RNA-seq but are rarely recorded with single-cell RNA-seq. This prevents the direct use of cell-level transcriptomes in cancer survival analysis. ResultsTo bridge this gap, we propose using bulk RNA-seq deconvolution algorithms to reconstruct each subjects scRNA-seq profile from their bulk data. Then, by combining both scRNA-seq and bulk RNA-seq together with their survival labels (paired to bulk), we perform multimodal transcriptome-based survival analysis. We built this framework as DeSCENT and evaluated it with common survival models on eight TCGA cancer cohorts. Results showed notable and consistent improvements in C-index over bulk-only models or models using cellular information alone. AvailabilityOur code is available at GitHub: https://github.com/YonghaoZhao722/DeSCENT. Contactzpluo@swjtu.edu.cn; y.he@imperial.ac.uk Supplementary informationSupplementary data are available at Bioinformatics online.

bioinformatics↗

bioRxiv2026

HMA proteins produce 2',3'-cNMP signaling molecules and activate CNL-mediated immunity in rice

Plants deploy a sensor repertoire, usually NLRs, to perceive pathogen effectors and activate executor NLR-triggered immunity. Here we reveal that rice HMA proteins function as sensors detecting the Magnaporthe oryzae MAX effector AvrPigm to trigger broad-spectrum blast resistance mediated by the CNL PigmR. AvrPigm is conserved with multiple copies in blast genomes and targets HMA proteins, facilitating HMA translocation into the cytoplasm. Cryo-EM structure of the HPP04HMA reveals a filament-like oligomer with ssDNA/RNA bound inside the filament, which display cyclic nucleotide synthase activity to generate 2',3'-cNMP in a CNL-dependent manner. We further discovered that the LRR domains of the executor CNLs perceive 2',3'-cNMP to mount immunity. Our study thus establishes an unrecognized immunity mode with a distinct repertoire of sensor receptors that produce signaling molecules that are perceived by the LRR domains of executor NLRs to mediate immunity, shedding new light on the sensor-executor conception and immune activation in plants. HighlightsO_LIHMA proteins and CNL receptors form a new sensor-executor mode in plant immunity C_LIO_LICryo-EM reveals the filament-like structure of HMA-ssRNA/ssDNA C_LIO_LIHMA proteins are cyclic nucleotide synthetases to generate 2',3'-cNMP C_LIO_LILRR domains of CNLs perceive 2',3'-cNMP to mediate immunity C_LI

plant biology↗

bioRxiv2026

Exploration of the screening and regulatory mechanisms of biomarkers related to ac4C modification in laryngeal squamous cell carcinoma patients based on single-cell analysis and machine learning

BackgroundN4-acetylcytidine (ac4C) modification plays a critical role in cancer development. Exploring ac4C modification in laryngeal squamous cell carcinoma (LSCC) may help elucidate its pathogenesis. MethodsLSCC-related datasets were obtained from GEO. After preprocessing and annotating single-cell data, malignant cells were identified by CNV scoring and further divided into subpopulations. Malignant epithelial cells (MECs) were identified and subclustered based on ac4C-related gene activity. Prognostic genes were screened using Cox regression and machine-learning approaches, followed by validation in clinical samples using qPCR. The biological and immunological relevance of these genes was further explored through immune infiltration, immunotherapy response, and mutation analyses. ResultsThe 14,465 identified MECs were classified into five subgroups (MEC1-5), among which MEC3 showed the strongest association with the ac4C gene set. Machine-learning analysis of MEC3-derived genes yielded seven prognostic markers, including BARX1, FHL2, NXPH4, PKMYT1, TNFAIP8L1, CRLF1, and CENPP. qPCR confirmed their differential expression between tumor and adjacent normal tissues. These genes were significantly associated with alterations in the tumor immune microenvironment, with high-risk patients showing increased immune infiltration and immune activity. ConclusionSeven ac4C-related prognostic genes were identified that may contribute to LSCC progression by modulating the tumor immune microenvironment, providing potential therapeutic insights.

developmental biology↗

bioRxiv2026

A human gut-BBB-brain microphysiological model for studying neurodegenerative diseases

The gut-brain axis has emerged as a crucial factor in neurodegeneration, with growing evidence linking gut dysbiosis and metabolic dysfunction to Alzheimers disease (AD) progression. Unfortunately, the lack of human-relevant in vitro models limits our ability to effectively explore the mechanism of this axis. To address this gap, we have developed a human induced pluripotent stem cell (iPSC)-derived gut-blood-brain barrier (BBB)-brain microphysiological system that enables systematic investigation of gut-brain interaction in the context of AD under controlled conditions. Our findings reveal that the interaction between gut and brain organoids can promote the maturation of brain organoids, making them more similar to their physiological characteristics in vivo. Additionally, co-culture gut and brain organoids better recapitulates the pathological features of AD. We also discovered that gut organoids of AD can trigger neurodegenerative disease manifestations in healthy brain organoids. In summary, our microphysiological system provides a novel and versatile in vitro platform for studying the interaction between the gut and brain in neurodegenerative diseases.

neuroscience↗

bioRxiv2026

Volumetric mechanoplasticity couples melanoma drug tolerance to susceptibility to CD8+ T cell killing

Cell swelling has been reported in physiological tumor contexts, but whether sustained cell volume expansion leaves a durable state that reshapes drug and immune responses in melanoma is unknown. Using controlled hypotonic dilution as an experimental handle, we compare two priming regimens with comparable hypotonic exposure but different magnitude and persistence of volume expansion, enabling us to test which post-recovery phenotypes track with sustained enlargement beyond hypotonic exposure alone. Sustained enlargement produces a measurable size imprint after return to isotonic conditions, accompanied by nuclear lamina and chromatin remodeling with p53 pathway engagement and suppression of replication and growth programs. After recovery, primed cells reinforce F-actin organization, migrate faster in 3D collagen, and induce antioxidant and anti-ferroptosis defenses consistent with improved stress survival. A gene expression signature derived from this response is associated with poorer outcome in TCGA skin cutaneous melanoma. Unexpectedly, the same volume history also increases IFN-{gamma} response, elevates MHC-I, reduces sialylation, and increases susceptibility to CD8+ T cell killing. These findings indicate that persistent volume history can couple drug tolerance programs to an exploitable increase in immune visibility.

biophysics↗

bioRxiv2026

CCycDB: an integrative knowledgebase to fingerprint microbially mediated carbon cycling processes

Microorganisms play essential roles in mediating biogeochemical cycling of carbon across Earths ecosystems. Understanding the processes and underlying mechanisms for microbially mediated carbon cycling is therefore critical for advancing global ecology and climate change research. To comprehensively depict these complex biogeochemical processes, we developed CCycDB, a knowledge-based functional gene database, to accurately fingerprint microbially-mediated carbon cycling pathways and gene families, particularly from shotgun metagenomes. The CCycDB database comprises 4,676 gene families classified into six major functional categories, further structured into 45 level-1 and 188 level-2 sub-categories, encompassing a total of 10,991,724 high-quality reference sequences. Validation using both synthetic and real-world datasets demonstrated that CCycDB outperforms existing orthology databases in terms of accuracy, coverage and specificity. By directly targeting carbon-cycling functional gene families, CCycDB provided promising routines to reconstruct both functional gene and taxonomic profiles associated with microbially mediated carbon cycling. Application of CCycDB to shotgun metagenomes from diverse and complex ecosystems revealed pronounced habitat-specific differences in carbon cycling processes and their associated microbial taxa. Collectively, CCycDB provides a powerful and reliable tool for profiling carbon cycling processes from both functional and taxonomic perspectives in complex ecosystems. CCycDB is accessible at https://ccycdb.github.io/. Impact StatementThe microbially mediated carbon cycling processes are the most complex biogeochemical processes in the Earths biosphere, playing profound regulatory roles on global climate changes. A key bottleneck in linking microbial communities to global change is the lack of integrated tools for comprehensive carbon cycle profiling. Here, we present CCycDB, a tool that serves a dual purpose--first being a reference database that obtains functional gene and taxonomic profiles and functioning as a customized routine for efficiently aligning sequences and querying associated functional information. CCycDB enables researchers to accurately link microbial community dynamics to carbon cycling and transforming pathways, thereby advancing integrated global change studies with microbes and ecological research via complex metagenomic datasets.

microbiology↗

bioRxiv2026

Targeted therapy-induced chromosomal instability dictates mitotic dependency on Aurora Kinase A

Targeted therapies eliminate cancer cells by inhibiting oncogenic signaling; however, tumor cells often evade cytotoxicity through proteomic and epigenetic reprogramming that enables survival. These adaptive responses may create collateral cellular stresses, such as DNA damage, that can be therapeutically exploited. When unresolved, DNA damage leads to chromosomal instability (CIN), a potential source of vulnerability. Whether KRAS inhibition induces DNA damage or CIN in KRAS-mutant non-small cell lung cancer (NSCLC) has not been established. Here, we show that the KRASG12C inhibitor LY3499446 induces CIN in KRAS-mutant NSCLC cell lines. A targeted compound screen revealed that the extent of CIN induction by KRASG12C inhibition strongly correlates with therapeutic synergy with the selective Aurora kinase A inhibitor LSN3321213. Mechanistically, KRASG12C inhibition stabilizes cyclin B1 during mitosis through activation of mitotic ATR/ATM signaling. In the presence of Aurora Kinase A inhibition, cyclin B1 stabilization delays mitotic exit and diverts cell fate from mitotic slippage or division toward mitotic catastrophe. Together, our findings identify CIN as a predictive marker of response to combined KRASG12C and Aurora Kinase A inhibition, providing mechanistic rationale to enhance the therapeutic window of AURKA inhibitors when used with targeted therapies.

cancer biology↗

bioRxiv2025

scHG: a supercell framework with high-order graph learning enables hyper-fast multi-omics analysis

Multi-omics profiling--spanning proteomics, transcriptomics, and additional omics data types--is rapidly advancing, providing increasingly detailed maps of cellular identity and function. In parallel, computational methods have evolved to integrate diverse omics with increasing precision in resolving cellular heterogeneity and delineating subtle subpopulation structures. Yet, capturing rare cell population while maintaining computational tractability remains a major challenge. Here, we introduce the supercell paradigm, in which expression-coherent cells are compressed into candidate units for rare cell population. Supercells are constructed using angle-aware similarity metrics and second-order co-occurrence neighbors, with impurity cells pruned by degree centrality. To address scalability, we implement sparse matrix optimization and iterative high-order graph updates, enabling efficient integration of large-scale multi-omics datasets. Building on this framework, we develop scHG, a high-order graph learning approach guided by an omics-weighted optimizer that adaptively balances contributions from gene expression, surface proteins, and chromatin accessibility. Across six benchmark datasets (up to 30672 cells), scHG consistently outperforms state-of-the-art methods, improving mean ARI and NMI by 3.97% and 3.54%, respectively, while reducing runtime by 26.40%. Beyond performance gains, scHG was able to resolve fine-grained cellular heterogeneity within conventionally defined T cell populations, distinguishing subpopulations such as memory-like or progenitor-exhausted T cells and innate-like cytotoxic T cells. At the same time, the supercell framework uncovered rare populations, including dendritic-cell populations and NK-like B cells, that remained masked at the cluster level under standard pipelines. These results underscore both the rare-cell detection capability and the computational efficiency of scHG. Our code and data are available at http://mialab.ruc.edu.cn/scHG_code/zip.

bioinformatics↗