Biology subjects

Kim, Y.

Publications and source records attributed to Kim, Y..

At least 73 records · Page 4Linked to original sources

bioRxiv2026

Targeting Noncanonical TEAD Axis to Overcome DNA Repair-Driven Chemoresistance

How transcription factors contribute to DNA damage repair (DDR) independent of canonical gene regulation remains poorly understood. Here, we show that DNA double-strand break (DSB) triggers a functional state transition in TEAD, disengaging it from YAP-dependent transcription and translocating it to DNA lesions as a chromatin-associated DDR factor. Upon genotoxic stress, TEAD is rapidly recruited to damaged chromatin through a biphasic mechanism involving early poly(ADP-ribose)-dependent recruitment followed by ATM-driven {gamma}H2AX-mediated retention. At DNA lesions, TEAD constrains chromatin over-relaxation, suppresses excessive end resection, and promotes non-homologous end joining (NHEJ) independent of its canonical function. Conserved residues within the TEA domain mediate this noncanonical chromatin engagement of TEAD and define a druggable N-terminal interface. Pharmacological targeting of this interface impaired TEAD-dependent DSB repair and sensitized tumors to chemotherapy. Together, these findings establish noncanonical TEAD as a critical DNA repair factor and a clinically actionable driver of chemoresistance, while providing a conceptual framework for understanding how broader families of transcription factors may be repurposed as targetable DNA damage regulators in cancer.

cancer biology↗

bioRxiv2026

Optical Spectral Fingerprinting Enables Sensitive Detection of Anthracycline Chemotherapeutics in Synthetic Clinical Biofluids

Anthracycline chemotherapeutics are commonly used as frontline treatments for a wide array of cancers. However, their administration to patients results in substantial side effects, primarily cardiotoxicity, as well as myelosuppression and gastrointestinal toxicity. Current clinical management of such side effects is solely based on a lifetime dosage limit, which inhibits their anti-tumor efficacy. Many individualized factors, including age, family history of cardiovascular disease, treatment regimen, and other co-morbidities influence drug pharmacology. Despite this heterogeneity, there is no method for determining actual organ or tumor exposure to the treatment in an individual. Here, we developed an optical nanosensor array for four anthracyclines--doxorubicin, daunorubicin, epirubicin, and idarubicin. We used single-walled carbon nanotubes as the signal transducer due to their tunable near-infrared fluorescence. We screened twelve distinct ssDNA sequences paired with seven SWCNT (n,m) species at increasing concentrations of each of the four anthracyclines. The spectral responses were then used to develop machine learning-based classification models to identify different anthracycline types and concentrations. The optimized extreme gradient boosting model was able to classify high levels of each anthracycline with 100% accuracy. Concentration-based classification by PCA was performed for each anthracycline, distinguishing low ([≤] 5 {micro}M) and high (> 5 {micro}M) concentrations. Finally, we validated the sensor performance using synthetic urine and sweat. Our findings demonstrate the potential of carbon nanotube-based sensor array to measure the pharmacokinetics of anthracyclines in patients with the goal of enhancing anti-tumor efficacy and monitoring off-target toxicities.

bioengineering↗

bioRxiv2026

Endogenous mutational mechanisms and metabolic context shape endometrial cancer

Uterine corpus endometrial carcinoma (UCEC) is a common gynecologic malignancy with rising mortality, yet its genome-wide mutational architecture remains incompletely understood. Here we analyze deep whole-genome sequencing data from 440 UCEC tumors from The Cancer Genome Atlas, integrated with transcriptomic and clinical data, to define subtype-specific mutational processes and genomic architectures. We uncover pronounced molecular subtype-specific differences in endogenous mutational mechanisms, retrotransposition activity, structural variation, and chromosomal instability. LINE-1 retrotransposition emerges as a key contributor to genome instability in UCEC, acting as a prominent source of structural variation in copy-number stable tumors and associating with chromothripsis and ecDNA-mediated oncogene amplification in copy-number high tumors. Mutational signature SBS28 contributes significantly to mutations in POLE-deficient tumors and is strongly correlated with SBS10a and SBS10b. Mismatch repair-deficient tumors exhibit a previously unrecognized doublet base substitution signature and a defining imbalance between ID2 and ID1 indel processes, reflecting pervasive DNA template-strand replication slippage and associated with increased tumor immunogenicity. Copy-number low tumors follow a distinct low-mutagenesis evolutionary trajectory characterized by reduced replication stress, low proliferative activity, genomic stability and enrichment of SBS18 associated with oxidative damage. Notably, we identify a consistent inverse association between body mass index and tumor mutational burden in UCEC, suggesting that host metabolic state may influence fundamental cellular processes governing mutation accumulation, promoting tumor development through non-mutagenic mechanisms rather than elevated genomic instability. Together, these findings establish a mutagenesis-centric framework for UCEC that links endogenous mutational mechanisms to tumor architecture and host metabolic context, uncovering previously unrecognized subtype-specific genomic features with important implications for refined risk stratification and therapeutic strategies.

genomics↗

bioRxiv2026

Bulk RNA sequencing deconvolution of pancreatic ductal adenocarcinoma identifies cancer-associated fibroblast subsets associated with survival and tumor microenvironment composition

Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal cancer characterized by a high abundance of cancer-associated fibroblasts (CAFs), which influence therapy response, tumor biology and tumor aggressiveness. CAFs are a heterogeneous cell type and previous single-cell RNA sequencing (scRNAseq) of PDAC tumors identified three main CAF subtypes: myofibroblastic, inflammatory and antigen-presenting CAFs (myCAF, iCAF, apCAF, respectively). However, scRNAseq on large patient cohorts is often not feasible due to costs and technical constraints. Therefore, bulk RNAseq deconvolution can be used to identify cell types within the heterogeneous tumor microenvironment. Here, Statescope deconvolution was used to identify different cell types of the tumor microenvironment within an early onset PDAC cohort, comprising 74 patients aged under 60. Three CAF populations were identified (iCAFs, myCAFs and desmoplastic CAFs), and their correlations with tumor microenvironment components, mutational signatures and survival were examined. iCAFs were associated with classical-like tumor cells, whereas myCAFs and desmoplastic CAFs correlated with basal-like tumor cells. Desmoplastic CAFs are associated with inflammatory granulocytes/neutrophils, while negatively associating with monocyte-derived macrophages and immature/transitional B cells. No associations were observed between mutational signatures and the abundance of CAF and epithelial tumor subtypes. Interestingly, a high abundance of CAFs, and specifically increased iCAF abundance, was associated with improved survival. This iCAF-mediated survival effect was predominantly apparent in female patients. All in all, deconvolution of bulk RNA sequencing data, followed by its integration with clinical and biological parameters, reveals the heterogeneity and prognostic implications of CAF subpopulations in the tumor microenvironment of early onset PDAC patients.

cancer biology↗

bioRxiv2026

PHGDH is a targetable driver of PDAC progression

Pancreatic ductal adenocarcinoma (PDAC) arises in a nutrient-deprived microenvironment through progressive stages from pancreatic intraepithelial neoplasia (PanIN) to invasive carcinoma. While serine metabolism supports tumor growth across multiple cancer types, the stage-specific role of de novo serine synthesis in PDAC evolution remains undefined. Here, we show that expression of phosphoglycerate dehydrogenase (PHGDH), the rate-limiting enzyme of serine biosynthesis, increases progressively from PanIN to invasive PDAC in human and mouse specimens. Using genetically engineered mouse models with inducible PHGDH knockdown, we found that PHGDH loss delayed PDAC development. Unexpectedly, PHGDH-deficient tumors did not increase reliance on exogenous serine, and dietary serine/glycine manipulation had no effect on tumor development. Instead, stable isotope tracing and metabolomic profiling revealed that PHGDH loss suppressed mTOR signaling, reduced expression of the glutamine transporter ASCT2, and impaired glutamine uptake and utilization. Leveraging this metabolic liability, we demonstrated that PHGDH-deficient tumors exhibited selective sensitivity to the glutamine antagonist DRP-104, whereas PHGDH-intact tumors were resistant. These findings reveal an unanticipated connection between serine biosynthesis and glutamine metabolism in PDAC and identify a therapeutic vulnerability that may be exploited through combined metabolic targeting. Statement of significancePHGDH supports PDAC progression not primarily through serine provision, but by maintaining glutamine metabolism and mTOR signaling. This unanticipated metabolic crosstalk creates a synthetic lethal vulnerability to glutamine antagonism in PHGDH-deficient tumors, providing a rationale for combining serine synthesis pathway inhibitors with glutamine-targeting therapies in pancreatic cancer.

cancer biology↗

bioRxiv2026

Energy-precision trade-off in mitotic oscillators revealed by ATP modulation in artificial cells

The temporal precision of biochemical oscillators is fundamentally constrained by the energy dissipated to suppress molecular fluctuations, a widely predicted trade-off governing information processing across biology and physics, from molecular motors to kinetic proofreading to computing. Yet, experimental validation in complex biological oscillators remains elusive due to challenges of systematically modulating energy while quantifying stochastic dynamics across large ensembles. Here, we establish a high-throughput droplet-microfluidics platform to reconstitute mitotic oscillations from Xenopus laevis egg extracts within thousands of sub-nanoliter compartments. By precisely tuning ATP across a broad free-energy landscape and developing an analytical framework that decouples intrinsic phase diffusion from quenched period heterogeneity, we uncover a hidden trade-off linking metabolic budget, oscillation speed, and precision. While speed peaks non-monotonically near physiological ATP levels and declines toward both high and low bifurcation limits, precision increases monotonically with energy. These findings provide direct experimental evidence that mitotic timing is actively shaped by energy budgets. Intriguingly, embryonic cell cycles are not optimized for maximum fidelity, but for a metabolic compromise maintaining just enough coherence for synchronous yet rapid divisions, placing the endogenous ATP budget near an energetic optimum balancing speed and accuracy. Our integrated artificial-cell and analytical strategy provides a generalizable framework for mapping thermodynamic limits in non-equilibrium biological dynamics.

biophysics↗

bioRxiv2026

EvoStructCLIP: A Mutation-Centered Multimodal Embedding Model for CAGI7 Variant Effect Prediction

AO_SCPLOWBSTRACTC_SCPLOWWe present EvoStructCLIP, a mutation-centered multimodal embedding model that integrates local 3D structural windows and evolutionary constraints to predict missense variant effects. EvoStructCLIP combines two encoders: a structure voxel encoder derived from AlphaFold residue neighborhoods and an MSA-based evolutionary encoder. It aligns the modalities through CLIP-style contrastive learning, with FuseMix regularization and an auxiliary pathogenicity loss trained on 153,787 ClinVar variants. Evaluations using lightweight regressors demonstrate that EvoStructCLIP embeddings capture highly transferable predictive signals across diverse phenotypes, including gene-specific functional readouts of BRCA1, KCNQ4, and PTEN/TPMT. This transferability is further supported in the CAGI7 blind competition setting, where models generalized to predicting different gene-specific readouts for BARD1, FGFR, and TSC2 without target-specific retraining and achieved competitive performance across heterogeneous biological tasks.

bioinformatics↗

bioRxiv2026

Stress-Encoded Mitochondrial Plasticity: ATF4 Control of Mega-Mitochondria and Nanotunnel Communication

Mitochondrial structural plasticity is a critical adaptive response to cellular stress, yet the transcriptional networks governing the formation of specialized mitochondrial architectures remain poorly defined. Here, we identified and demonstrated that activating transcription factor 4 (ATF4), the master regulator of the integrated stress response, directly regulates mitochondrial morphological remodeling through a novel ATF4-NRF1/Nrf2-MFN2 signaling axis. Using serial block-face scanning electron microscopy and three-dimensional reconstruction in Drosophila flight muscle, primary myotubes, and human skeletal muscle, we show that overexpression of ATF4 promotes significant mitochondrial elongation, increased cristae concentration, enhanced mitochondrial-endoplasmic reticulum contact site (MERC) formation, and the initiation of Mitochondrial Nanotunnels. In contrast, loss of ATF4 results in mitochondrial fragmentation and impaired aerobic capacity. Chromatin immunoprecipitation sequencing reveals direct ATF4 binding at the promoters of the genes encoding NRF1 and Nrf2, which in turn regulate MFN2 expression. Small-molecule inhibition studies further establish that activation of this hierarchical pathway is both necessary and sufficient for stress-induced mitochondrial structural adaptation. Together, these findings position ATF4 as a master regulator of mitochondrial architectural plasticity, providing a direct mechanistic link between cellular stress signaling and organelle remodeling.

cell biology↗

bioRxiv2026

Selective loss of Primary Cilia and Neurotrophic Signaling in G51D alpha-Synuclein Mice Highlights a Common Pathway to Parkinsons Disease

Parkinsons disease is characterized by dopaminergic neuron loss and accumulation of -synuclein aggregates in the brain. G51D -synuclein knock-in mice provide a genetically and clinically relevant model of disease, exhibiting early olfactory deficits, age-dependent motor impairment, and progressive phospho--synuclein accumulation. In multiple Parkinsons disease models, striatal cholinergic and parvalbumin interneurons, as well as astrocytes, lose primary cilia and the neurotrophic signaling needed to sustain dopaminergic neurons. We show here that G51D--synuclein mice share these phenotypes. Phospho-Ser129 -synuclein accumulation correlates with cilia loss in cholinergic interneurons but not in medium spiny neurons that accumulate higher phospho--synuclein levels. In the piriform cortex, parvalbumin neurons lose primary cilia and downregulate Neurturin, potentially contributing to olfactory dysfunction. Within the peripheral olfactory epithelium, horizontal basal cells lose cilia, whereas multi-ciliated olfactory sensory neuron cilia remain intact. These findings reveal convergent cellular vulnerabilities across Parkinsons disease models and highlight a pathogenic role for impaired ciliary signaling. TeaserLoss of primary cilia may contribute to dopamine neuron loss in both inherited and common Parkinsons disease.

neuroscience↗

bioRxiv2026

Microfluidic Agarose Microdroplets for DNA-Encoded Chemical Library Screening

DNA-encoded library (DEL) technology enables high-throughput small-molecule discovery but is typically performed using purified proteins under in vitro conditions that do not reflect native intracellular environments. Here, we present a microfluidic agarose -droplet platform for cellular-context DEL screening. The porous hydrogel droplets provide mechanically stable yet permeable microenvironments that protect weak protein-ligand interactions while enabling efficient buffer exchange and ligand diffusion. Importantly, mild cell permeabilization within droplets selectively retained chromatin-associated proteins, allowing screening directly in a cellular context. Using BRD4 as a model target, we validated intracellular ligand engagement by fluorescence imaging and super-resolution microscopy. Small-scale DEL screening selectively enriched JQ1 in both bead-based and cell-based formats, and large-scale DEL screening across millions of encoded compounds successfully identified hit molecules by sequencing. This agarose -droplet-based strategy expands DEL technology toward biologically relevant and chromatin-associated targets under near-native conditions.

bioengineering↗

bioRxiv2026

Metformin promotes broad neuroprotection and proteostatic resilience via rngo/DDI2 stabilisation

Alzheimers disease (AD) is one of the most common age-related causes of death, with limited effective disease-modifying treatments. Although metformin shows promise as a disease-modifying agent for AD, its molecular mechanism--specifically how it confers neuroprotection despite potentially increasing amyloid-beta (A{beta}) load--remains obscure. Here, we demonstrate that metformin functions as a pharmacological activator of the ubiquitin-binding protease rngo/DDI2. Through a genetic screen in Drosophila, we identified rngo/DDI2 as a potent suppressor of A{beta} toxicity. We provide in silico and genetic evidence that metformin interacts with the conserved D257 residue of the rngo/DDI2 RVP domain, inducing homodimerisation and subsequent protein stabilisation. This activation boosts proteasome activity in the presence of A{beta} preferentially clearing highly abundant proteins to preserve proteostasis. Crucially, this intervention is broadly effective; rngo/DDI2 upregulation robustly suppresses toxicity in models of TDP-43 and C9orf72-repeat expansion pathology, indicating a generalised mechanism of neuroprotection. Supported by human iPSC data and evidence of DDI2 depletion in AD patient brains, our results identify rngo/DDI2 as a conserved regulator of neuronal resilience. We propose that directly targeting DDI2 stabilisation represents a novel, broadly applicable therapeutic strategy to counteract proteotoxic stress across a broad spectrum of neurodegenerative diseases.

neuroscience↗

bioRxiv2026

A plant virus, TYLCV, up-regulates an endocrine oxylipin signal in its insect vector, Bemisia tabaci, via the viral C2 virulence factor for viral transmission

To survive and efficiently transit between plant and insect hosts, circulative plant viruses have evolved sophisticated strategies to exploit insect vector factors. Tomato yellow leaf curl virus (TYLCV) is transmitted by Bemisia tabaci through a circulative and replicative pathway. In insects, C20 oxylipin (eicosanoid) and C18 oxylipin (EpOME) antagonistically regulate antiviral responses. Upon TYLCV infection, the intestinal apoptosis of B. tabaci facilitated the viral multiplication. The apoptosis was suppressed by eicosanoid but induced by EpOME. EpOME treatment also upregulated other proviral factors, including defensin, PGRP, and cathepsins, while eicosanoid signaling exerted opposite effects. TYLCV infection suppressed eicosanoid biosynthetic enzymes but induced a cytochrome P450 gene involved in EpOME biosynthesis, consistent with elevated EpOME levels in the viruliferous B. tabaci detected by LC-MS/MS. Individual RNA interference treatments specific to each of the TYLCV genes in the viruliferous insects revealed that only silencing of the viral C2 gene abolished EpOME-mediated proviral effects. These findings uncover a lipid-mediated mechanism by which TYLCV enhances vector competence to promote transmission. IMPORTANCEVarious plant viruses depend on insect vectors for their horizontal transfer. Some of them exhibit a circulative and propagative transmission by multiplying the viral titers within the insects using the host machinery. Here is a fascinating manipulation of the host immunity by a plant virus, tomato yellow leaf curl virus (TYLCV), which uses the insect endocrine signals associated with immunity of its vector, Bemisia tabaci. Two types of oxylipins, eicosanoid and EpOME, antagonistically act to insect immunity, in which eicosanoid induces immune responses while EpOME, as an insect immune resolvin, suppresses them. TYLCV uses its C2 gene component as a virulent factor to induce the EpOME biosynthesis of B. tabaci. The elevated EpOME levels in the vector insect led to proviral responses by inducing intestinal apoptosis and selectively suppressing the immune-associated genes. These findings demonstrate the viral manipulation of the host endocrine signal for inducing proviral responses.

plant biology↗

bioRxiv2026

C-terminus CD28 phosphorylation (Y218) modulates IL-2 secretion and antitumor effect of CAR-T cells

CD28 is a co-stimulatory component of several second-generation chimeric antigen receptor (CAR)-T cells, providing signals essential for T cell proliferation, survival, and cytokine secretion. However, the specific contribution of individual CD28 intracellular motifs to CAR-T cell function remains incompletely understood. Here, we identify tyrosine 218 (Y218) in the CD28 cytoplasmic domain as a critical regulatory site, and demonstrate that its phosphorylation is essential for optimal CAR-T cell activity. Using a 218F mutant, we show that loss of Y218 phosphorylation leads to impaired IL-2 production and abrogates antitumor efficacy. Transcriptomic profiling of 218F CAR-T cells revealed increased expression of IL-17A, IL-17F, and related cytokines, suggesting a shift toward a pro-inflammatory Th17-like phenotype that may contribute to dysfunction. Mechanistically, we demonstrate that the interleukin-2-inducible T-cell kinase, ITK, mediates Y218 phosphorylation. To further understand the role of this kinase, we engineered a novel CAR incorporating an ITK-binding motif (PYRP), which enhances ITK recruitment, increases Y218 phosphorylation, and boosts IL-2 secretion, and improves anti-tumor efficacy in vivo. Our findings underscore the functional relevance of Y218 phosphorylation in modulating CAR-T cell fate and reveal a strategy to fine-tune CAR signaling through targeted kinase recruitment to enhance therapeutic efficacy.

immunology↗

bioRxiv2026

A curvilinear coordinate flatmap for visualizing hippocampal structure and development

The hippocampal formation is a highly curved and topographically complex forebrain structure. This complex geometry presents persistent challenges for analyzing subregional, laminar, and connectivity patterns. Here, we present a computational workflow that generates curvilinear-coordinate flatmaps from Common Coordinate Framework (CCF) registered hippocampal and retrohippocampal regions by solving the Laplacian equation to derive geodesic streamlines. This transformation unfolds the hippocampus into a planar slab, bounded by the meningeal and ventricular surfaces, with the depth defined along the radial axis. We apply this transform to image volumes, single neuron reconstructions, and point data, including spatial transcriptomic and rabies tracing datasets, revealing topographic variations in the dorsoventral and radial axes that are obscured in the CCF coordinate space. As proof of principle, we use flatmaps to show connectivity loss in a mouse model of Alzheimers disease and track postnatal development of microglial distribution in the hippocampus. This work provides an efficient and accessible resource for visualizing hippocampal organization across development and disease, offering new opportunities to interrogate the structure and function of this important brain region.

neuroscience↗

bioRxiv2026

Programmable artificial RNA condensates in mammalian cells

Artificial biomolecular condensates have emerged as powerful tools to control cellular behaviors. Here we introduce a method to build artificial condensates within living mammalian cells through the design of modular RNA motifs formed by a single, short strand of RNA. These condensates emerge spontaneously, creating RNA-rich compartments that remain separated from their surrounding environment. The RNA sequences include stem-loop domains that fold as the RNA is transcribed, and then condense in the nucleus and cytoplasm through loop-loop interactions. These sequences can be optimized and diversified, enabling the generation of distinct, non-mixing condensate populations and the programmable control of their subcellular localization. The RNA motifs can also be modified to recruit small molecules, proteins, and RNA molecules in a sequence-specific manner to the RNA-rich phase. By introducing additional RNAs that link two distinct types of condensates, we can create droplets with multiple subcompartments, whose organization can be controlled by tuning the stoichiometry of different RNA sequences. These artificial condensates provide a versatile platform for studying and manipulating molecular functions inside living cells.

synthetic biology↗

bioRxiv2026

Knowledge-Guided Learning with Curated Prior Genetic Biomarkers for Robust Model Interpretation

MotivationKnowledge-guided learning offers effective and robust model training strategies in data-scarce settings by incorporating established domain knowledge, thereby enhancing generalization, robustness, and interpretability. By contrast, conventional deep learning approaches rely purely on data-driven learning, which can limit robust model interpretability, particularly in high-dimensional settings with limited size samples. In computational biology, knowledge-guided learning has primarily leveraged network- and structural-based knowledge, leading to biologically interpretable representations and enhanced predictive performance compared to conventional approaches. However, curated biomarkers, one of the most accessible forms of biological knowledge, remain largely unexplored within knowledge-guided paradigms. ResultsIn this study, we propose a model-agnostic training paradigm, Biomarker-driven Explainable Prior-guided Learning (BioExPL), that can be applied to any neural networks that incorporates curated prior knowledge. BioExPL enforces neural networks to reflect curated biomarker priors in their latent representations through a novel knowledge-alignment loss. BioExPL consistently demonstrated significantly improved predictive performance and enhanced model interpretability with minimized computational overhead in simulation studies and intensive experiments on multiple cancer datasts. BioExPL not only integrates prior curated knowledge into the model but also accurately identifies unknown associated signals additionally. BioExPL is model-agnostic and domain-independent, enabling its integration into diverse neural network architectures. Availability and implementationThe open-source is publicly available at: https://github.com/datax-lab/BioExPL.

bioinformatics↗

bioRxiv2026

MANTIS: Analytics toolkit for spatial metabolomics with matching spatial transcriptomics data

MotivationJoint Spatial Metabolomics (SM) and Spatial Transcriptomics (ST) profiling is a powerful approach to fine-mapping of metabolic states associated with tissue function. Current computational tools for analysis of "SM+ST" data focus primarily on alignment and integration of the two modalities, with limited support for probing biological relationships between the two molecular layers. ResultsWe present MANTIS, a statistical framework for analyzing co-registered SM+ST profiles at single cell or spot resolution, along with spatial domain or cell type information, to discover metabolite spatial patterns and gene-metabolite relationships. It employs an autocorrelation-preserving permutation strategy to assess statistical significance, yielding calibrated inference under spatial dependence. It disentangles different sources of spatial patterns and correlations, viz., those arising from regional preferences, cell type associations, or other unknown factors. It introduces the use of spatial cross-correlation and spatial partial correlation statistics for quantifying gene-metabolite associations. Across data sets spanning different spatial technologies, tissues and species, MANTIS provides more specific and interpretable discoveries than existing methods through rigorous statistical testing and explicitly modeling confounding structure. To our knowledge, MANTIS is the first toolkit to unify spatial metabolomics, spatial transcriptomics, cell type information and spatial domains within a single framework that emphasizes spatial statistics, hypothesis testing and confounder correction. Availability and ImplementationFreely available on the web at https://github.com/yuhaotuo/MANTIS.

bioinformatics↗

bioRxiv2026

Spatial Decoding of Tertiary Lymphoid Structure Maturation in Non-Small Cell Lung Cancer Using Deep Neural Networks

Understanding the role of tertiary lymphoid structures (TLS) is crucial in non-small cell lung cancer (NSCLC), as they are associated with patient prognosis and treatment outcomes. Specific cellular ecosystems that originate anti-tumor activity or predict immunotherapy response remain poorly characterized. To this end, we developed a high-resolution, multimodal spatial atlas jointly profiling transcriptomics, proteomics, and histology to characterize TLS maturation in NSCLC alongside secondary lymph organs as a baseline. Using this atlas, we proposed a pathologist-in-the-loop framework that combines a variational graph autoencoder (VGAE) with diffusion pseudotime to refine human expert annotations and characterize TLS maturation. These spatial molecular representations were extended to H&E whole-slide images via a vision transformer-based foundation model. Next, we resolved cellular composition, spatial organization, and cell-cell interactions within these data and defined two divergent spatial ecosystems. Clinical evidence suggests that these ecosystems are associated with distinct patient outcomes: a mature germinal center niche with favorable prognoses, and a tumor-macrophage-fibroblast niche with unfavorable prognoses. In summary, our work decodes key components of TLS heterogeneity, identifies hallmark spatial patterns involved in NSCLC adaptive immunity, and provides a framework for translating spatial omics insights into clinical applications.

genomics↗