Biology subjects

Xu, Y.

Publications and source records attributed to Xu, Y..

At least 91 records · Page 5Linked to original sources

bioRxiv2026

ANCHOR: haplotype-aware allelic and isoform inference from single-cell long-read RNA sequencing with de novo variant calling

Long-read RNA sequencing enables haplotype- and isoform-resolved allelic analysis of transcriptomes, yet extending this capability to single cells and distinct cell types remains computationally challenging due to sparse coverage, sequencing errors, incomplete variant information, and reference-biased transcript assignment. Here we present ANCHOR, a haplotype-aware framework for single-cell long-read RNA sequencing that performs de novo expressed-variant discovery, molecule-level haplotype assignment and isoform-resolved allelic quantification. ANCHOR combines a signed-graph variant caller, pair hidden Markov modelling and beta-binomial UMI aggregation to infer parental allele counts for genes and splice-resolved isoforms, without requiring a pre-existing phased genotype or deep learning. In human single-cell long-read RNA benchmarks, ANCHOR improved variant-calling performance over tested long-read RNA callers at single-cell and low-to-moderate coverage, and its beta-binomial model reduced depth-driven false positives in allele-specific expression testing. Applied to newly generated single-cell long-read RNA-seq data from reciprocal mouse crosses during gastrulation, ANCHOR resolved cell-type- and isoform-specific parent-of-origin imprinting and identified an antagonistic maternally biased Sgce isoform. ANCHOR provides a general framework for allele-and isoform-resolved analysis of diploid single-cell long-read transcriptomes.

bioinformatics↗

bioRxiv2026

SPOTTER: Automated Tissue-Barcoding Platform for Spatial Proteomics and Phosphoproteomics

Spatial proteomics aims to characterize proteome in situ with regional resolution, linking molecular states to tissue architecture and microenvironments. While recent advances have expanded our access to spatially resolved proteomics, major barriers in scalability and throughput remain. Building on our prior SPOT (Spatial Proteomics through On-site Tissue-protein-labeling) framework1, we introduce SPOTTER, a customizable robotic platform that enables automated, micron-scale spatial barcoding directly on intact tissue sections. SPOTTER barcodes proteins by labeling tissue proteins in situ to encode spatial origin across the whole tissue section. This strategy enables unbiased whole-tissue proteome mapping and, for the first time, spatial phosphoproteome profiling from intact sections. Coupled with high-resolution LC-MS/MS, SPOTTER achieves deep proteome and phosphoproteome coverage while preserving histological integrity. By replacing labor-intensive laser capture microdissection workflows and low-plex antibody arrays, SPOTTER provides a scalable route to high-plex spatial proteomics, resolving spatially distinct regions within a single experiment.

bioengineering↗

bioRxiv2026

Oxidized protein states define metastatic fitness in lung cancer

AbstractReactive oxygen species (ROS) are a pervasive feature of human cancers, yet the protein targets through which ROS-regulated cell states shape tumor biology remains poorly understood. Here, using cysteine chemical proteomics, we define signatures of protein states under distinct cellular ROS environments that capture protein oxidation and conformational changes. Quantifying these signatures in primary lung tumors and brain metastases revealed a surprising enrichment of oxidative states in metastasis. To determine how these states support fitness, we performed genome-wide CRISPR screens and identified the mitochondrial Complex I subunit NDUFA10 as a key oxidation-dependent vulnerability. Oxidation of NDUFA10*Cys253 supports Complex I function through a previously unrecognized nucleotide kinase activity that maintains mitochondrial DNA levels. Enforcing a reduced conformation in NDUFA10 disrupts brain metastatic colonization in vivo. These findings establish ROS regulated protein states as a functional layer of tumor fitness, providing a framework for identifying redox-dependent mechanisms that support cancer progression.

Cancer Biology↗

bioRxiv2026

DyMoTree decodes early cell state transitions and drivers from single-cell transcriptomes using a tree-structured neural network

Inferring early cell fate from single-cell RNA-sequencing data is essential for identifying cellular origins and fate plasticity in development and disease. However, existing methods often fail to exploit tree-structured lineage trajectories, limiting the accuracy and interpretability of fate mapping. Here we present DyMoTree, a computational framework that models cell fate decisions as nonlinear mappings between progenitor and terminal cell states under explicit lineage constraints. By integrating lineage graphs with a tree-structured neural architecture, DyMoTree learns lineage-resolved cell-state transition maps from single-cell transcriptomes, enabling robust inference of early fate bias and identification of fate-specific progenitor substates and driver genes. Across simulations, lineage-tracing experiments, and in vivo systems, DyMoTree outperformed existing methods in resolving early fate biases. Applications to mouse embryogenesis, lung adenocarcinoma progression, and CAR-T immunotherapy revealed regulatory programs underlying developmental and disease-associated transitions. DyMoTree provides a general framework for modeling lineage-resolved cell-state dynamics underlying development and disease progression. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=156 SRC="FIGDIR/small/731114v2_ufig1.gif" ALT="Figure 1"> View larger version (49K): org.highwire.dtl.DTLVardef@11e715dorg.highwire.dtl.DTLVardef@1a4c695org.highwire.dtl.DTLVardef@e98805org.highwire.dtl.DTLVardef@1e14054_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2026

HSSM: A Widely Applicable Toolbox for Hierarchical Bayesian Neuro-cognitive Modeling

Computational models are central to cognitive neuroscience, but their rigorous application to experimental datasets is often constrained to a narrow set of canonical models that afford tractable analytical computations. We introduce the HSSM (Hierarchical Sequential Sampling Model) ecosystem, a Python toolbox that democratizes access to a broad, extensible array of neurocognitive process models through hierarchical Bayesian inference. Naturally leveraging simulation-based inference via likelihood surrogates, HSSM enables fast parameter estimation for models lacking closed-form likelihoods. Built atop PyMC and Bambi, HSSM provides a user-friendly formula syntax for specifying hierarchical mixed-effects regressions on model parameters, incorporating trial-by-trial neural or physiological covariates. The ecosystem allows fast model simulation and training data generation, as well as the neural network training utilities to deploy surrogate likelihood networks via HuggingFace. Contributions are designed to benefit not only the single researcher working on a problem, but organically, the entire research community. Together, the tools in the HSSM ecosystem bridge the interests of computational theorists as well as experimentalists, accelerating the cycle from model development to rigorous empirical testing.

neuroscience↗

bioRxiv2026

Decoding Hierarchical Cell-Cell Communication in Spatial Multi-Omics with CellSTIC

Cell-cell communication coordinates tissue development, homeostasis, and immunity, yet defining signaling interactions within intact tissues remains challenging. Single-cell transcriptomics enables systematic ligand- receptor inference, but tissue dissociation removes spatial context and obscures local and region-specific signaling. Spatial transcriptomics and spatial multi-omics can recover communication in situ, although existing methods often incompletely integrate heterogeneous data or produce poorly interpretable ligand-receptor lists. Here we present CellSTIC, a framework that resolves cell-cell communication in spatial multi-omics as structured programs grounded in tissue architecture. CellSTIC integrates multimodal evidence from local neighborhoods and organizes interactions into a hierarchical semantic representation that remains traceable to underlying molecules. This enables analysis from individual ligand-receptor pairs to functional modules comparable across tissues, regions, and states. In simulations and diverse tissue datasets, CellSTIC recovered spatially coherent communication structures and domains, revealing immune, brain, developmental, and regenerative programs, and providing a general approach for generating mechanistic hypotheses in situ. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=81 SRC="FIGDIR/small/728114v2_ufig1.gif" ALT="Figure 1"> View larger version (37K): org.highwire.dtl.DTLVardef@a04e82org.highwire.dtl.DTLVardef@826435org.highwire.dtl.DTLVardef@81092dorg.highwire.dtl.DTLVardef@1818e8b_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2026

Allele-specific expression reveals complex cis- and trans-regulatory divergence underlying fruit phenotypic differences between cultivated and wild tomato species

Tomato (Solanum lycopersicum) is one of the most important agricultural crops and serves as a model system for fleshy fruit biology. However, domestication bottlenecks have led to limited genetic diversity for further crop improvement. Wild relatives represent a rich reservoir of phenotypic diversity, yet the molecular basis underlying such phenotypic diversity remains largely unknown. As gene expression variation is a major driver of phenotypic diversity, we performed genome-wide allele-specific expression analyses in F1 hybrids between cultivated tomato and three wild relatives spanning a range of evolutionary distances, across three fruit tissues and multiple developmental stages. Our analysis generated a multi-species map of cis - and trans-regulatory variation, revealing a predominant role of cis-regulatory effects in expression divergence across species, tissues, and stages. The majority of cis effects were tissue- and/or stage-specific, underscoring the importance of tissue context in regulatory variation. Regulatory mechanisms and inheritance patterns shifted with evolutionary distance, with more distantly related species showing increased cis-regulatory contributions. Finally, we identified extensive regulatory divergence in biosynthetic pathways related to fruit nutritional value and flavor quality, including carotenoids, flavonoids, alkaloids, sugars, and volatiles. This study presents a high-resolution map of regulatory variation underlying tomato fruit development and provides evolutionary insights into the regulation of fruit nutrition and flavor traits.

genomics↗

bioRxiv2026

Disentangling RNA evolution and thermodynamics in genomic language models

Genomic language models (gLMs) trained only on large-scale nucleic acid sequence data seem to capture signals of RNA structure, yet the specifics of how remain unclear. Using the categorical Jacobian (CJ) operation, a model-agnostic operation for querying pairwise dependencies, we systematically compared three flagship gLMs: RNA-FM, Evo 2, and gLM2. We found that CJ signals recover base pairs supported by evolutionary covariation analyses, consistent with findings in protein language models. Surprisingly, CJ also recovers base pairs lacking evolutionary support but predicted by biophysical nearest-neighbor models. Is it possible gLMs have "learned" RNA thermodynamics? We noticed nearest-neighbor RNA folding models often predict reflected structures when given reversed sequences, consistent with these models modular and grammar-like nature. We leveraged this observation to create a simple "mirror test" that we found gLMs routinely fail, indicating they have not learned generalizable biophysics-based rules for RNA structure. Nevertheless, their apparent thermodynamic signal potentially confounds interpreting gLM pairwise dependencies as evidence of evolutionary conservation. We therefore introduce a method using synthetic sequences as a control for detecting significant learned signal. Our results demonstrate that gLMs can mimic thermodynamics through learned sequence context rather than general physical principles, but solutions exist for disentangling patterns in language models.

biophysics↗

bioRxiv2026

Graph-based pangenome provides insights into the adaptive evolution of Cucurbita pepo

Understanding how crops respond to environmental variation is crucial for biodiversity conservation and food security. Cucurbita pepo (pumpkin, squash, gourd) is one of the first domesticated crop species and exhibits remarkable phenotypic and ecological diversity, making it a powerful system for investigating the genomic basis of adaptation. Here, we constructed a graph-based C. pepo pangenome using nine chromosome-level assemblies and identified 229,431 high-confidence structural variants (SVs) that were genotyped across 206 wild and cultivated accessions. Our results demonstrate that C. pepo underwent parallel domestication, with two deeply diverged gene pools independently giving rise to the pepo and ovifera cultivated lineages, followed by expansion into diverse environments that produced strong signatures of differentiation largely mediated by young adaptive alleles. Single-nucleotide polymorphisms (SNPs) and SVs contribute complementary dimensions of environmental responsiveness. Biogeographical modeling predicts continued range contraction of wild relatives and elevated genetic offset in Eastern North American populations under projected future climates. Populations with higher genetic load harbor fewer adaptive variants and exhibit greater predicted maladaptation, indicating that deleterious mutations constrain adaptive potential. These findings highlight how genomic variants, adaptive diversity, and genetic load together shape environmental adaptation and inform conservation and crop improvement.

genomics↗

bioRxiv2026

An NO-binding Cache domain receptor interacts with a Ser/Thr kinase through a conserved HAMP domain interaction

Cache_heme domains are a family of bacterial heme c proteins that combine a conserved Cache fold with an -helical insertion containing a heme-binding CXXCH motif. We characterize an unusual Cache_heme-HAMP (CHH) receptor from Pseudomonas azotoformans with a periplasmic NO-binding Cache_heme domain, transmembrane region, and a cytoplasmic HAMP domain uncoupled from intrinsic enzymatic output. Genomic analysis reveals that CHH receptors are frequently co-localized with genes encoding Ser/Thr kinases (STPKs) that possess a catalytic domain similar to PknB, but lack peptidoglycan-binding PASTA domains. Using biochemical assays, biolayer interferometry, SEC-MALS, and cryo-electron microscopy, we show that the P. azotoformans CHH receptor binds its operon-associated kinase with nanomolar affinity via conserved C-terminal repeat modules in the kinase composed of two pseudo-symmetric -helical bundles. Cryo-EM structures demonstrate that this interaction orders the otherwise flexible HAMP domain, with the repeated helical domains from the kinase contacting each HAMP subunit symmetrically. Kinase-receptor binding is ATP-independent, but kinase phosphorylation of the receptor at a specific HAMP threonine residue is substantially enhanced when NO binds to the heme c sensor domain. Thus, NO-induced conformational changes that transverse the membrane either modulate substrate accessibility in the HAMP or alter kinase activity directly. These findings potentially define a mode of bacterial NO signaling in which a periplasmic heme c receptor couples ligand sensing to cytoplasmic phosphorylation via a physically associated Ser/Thr kinase, a mechanism distinct from that of cytoplasmic NO sensors. The conserved nature of the kinase recognition motif suggests broader relevance of CHH-STPK interactions across proteobacteria.

biochemistry↗

bioRxiv2026

Signaling mechanism of the transmembrane energy receptor Aer

The E. coli aerotaxis receptor Aer is a bacterial chemoreceptor that senses intracellular redox changes via an N terminal PAS domain bound to a flavin adenine dinucleotide (FAD) cofactor. Distinct from canonical methyl-accepting chemotaxis proteins (MCPs) such as Tar/Tsr, Aer lacks a periplasmic ligand-binding domain and adaptive methylation, transmitting conformational signals laterally from the PAS domain to the HAMP domain and the methylation helix-like cap (MHL cap) of the kinase control domain (KCD). To elucidate the Aer signalling mechanism, we determined cryo electron microscopy (cryo EM) structures of full length Aer in oxidized flavin quinone (kinase on) and reduced semiquinone (kinase off) states. Structural comparison reveals redox linked rearrangements of the FAD binding pocket, reorientation of PAS-HAMP interactions, and strikingly altered MHL cap stability. PAS-MHL-cap contact in the oxidized state compresses the receptor and stabilized proximal KCD helices, whereas reduction disrupts these contacts, increasing KCD flexibility. To probe distal effects on KCD architecture, we performed nanodisc reconstitution and pulse dipolar ESR spectroscopy on spin labelled positions along the four helix bundle. Distance distributions indicate redox dependent changes in helix separation, particularly at the C terminal MHL2 region, consistent with PAS driven loosening of KCD packing in kinase off states. These data support a model in which FAD redox chemistry reorganizes flavin pocket residues that in turn subtly alter PAS conformation to influence PAS-HAMP and PAS-MHL-cap packing and hence KCD conformational stability. The findings reveal an Aer specific signaling axis distinct from periplasmic ligand binding MCPs that has adapted MCP architecture for lateral PAS input and cytoplasmic redox sensing.

biochemistry↗

bioRxiv2026

Daikenchuto modulates gut microbial metabolism to mitigate irinotecan-induced enterotoxicity

Gut microbiota metabolic remodeling is a pivotal determinant in irinotecan-induced enterotoxicity and epithelial damage, although the underlying mechanisms remain unclear. Herein, we discovered that Daikenchuto (DKT), a traditional Chinese prescription for intestinal disorders, alleviated irinotecan-induced enterotoxicity without compromising its anti-tumor efficacy by improving weight loss, diarrhea, intestinal inflammation, and barrier damage, and these effects were partially dependent on gut microbiota. DKT significantly restored microbial tryptophan metabolism in irinotecan-treated rats, which was characterized by the enrichment of Limosilactobacillus reuteri, and elevated levels of indole-3-ethanol (IE) and indole-3-propionic acid (IPA). Multi-omics analysis further revealed a positive correlation between L. reuteri and IE and IPA. Consistent with this, DKT promoted L. reuteri proliferation, leading to the conversion of tryptophan to IE and IPA, which improved epithelial barrier damage in the irinotecan-treated Caco-2 cells. In addition, DKT suppressed the growth of Loop 1 {beta}-glucuronidase ({beta}-GUS)-producing bacteria, such as Escherichia coli. Furthermore, the main constituents of DKT selectively inhibited Loop 1 {beta}-GUS activity independent of the gut microbiota, which reduced the intra-luminal level of 7-ethyl-10-hydroxycamptothecin, the toxic metabolite of irinotecan. Taken together, this study reveals a dual gut microbiota-driven mechanism by which DKT mitigates irinotecan-induced enterotoxicity, which provides a promising strategy for managing chemotherapy-related enterotoxicity.

pharmacology and toxicology↗

bioRxiv2026

The in vivo inhibitory function of the MHC-I alpha3 domain/CD8alpha interaction

The interaction between the Major Histocompatibility Complex Class I (MHC-I) 3 domain and CD8 has classically been viewed as a positive coreceptor interaction that stabilizes TCR signaling during antigen recognition. However, its physiological function in mature peripheral CD8+ T cells in vivo remains incompletely understood. Here, we identify the MHC-I 3 domain-CD8 interaction as a previously unrecognized inhibitory pathway that tonically restrains peripheral CD8+ T-cell activation and maintains T-cell tolerance in vivo. Antibody-mediated disruption of the MHC-I 3 domain-CD8 interaction induced spontaneous activation of peripheral CD8+ T cells without impairing their survival, lowered the threshold for antigen-induced activation, and enhanced responsiveness to cognate peptide stimulation. In peptide-induced OT-I T-cell anergy models, blockade of either H-2Db or H-2Kb 3 domain interactions with CD8 prevented the induction of anergy and restored responsiveness of previously anergic T cells. Notably, blockade of the H-2Kb 3 domain enhanced OT-I responses despite simultaneously disrupting the classical positive coreceptor interaction within the TCR-peptide-MHC complex, indicating that tonic inhibitory signaling mediated by the MHC-I 3 domain predominates under these conditions. Together, these findings redefine the classical MHC-I-CD8 interaction as a bidirectional pathway that not only supports antigen recognition but also imposes tonic inhibitory control over peripheral CD8+ T cells. These results identify the MHC-I 3 domain-CD8 axis as a potential target for reversing T-cell tolerance and enhancing antitumor or antiviral immunity.

immunology↗

bioRxiv2026

Longitudinal Awake Mouse Brain Imaging Using Functional Ultrasound and Functional Ultrasound Localization Microscopy

Ultrafast ultrasound offers a unique route to cross-scale neurovascular phenotyping by integrating functional ultrasound (fUS), ultrasound localization microscopy (ULM), and functional ULM (fULM). Yet the baseline variability, longitudinal stability, and biological safety of such multimodal imaging in awake animals remain insufficiently defined, limiting its use for detecting subtle disease-associated neurovascular changes. Here, an awake longitudinal fUS-ULM-fULM framework is established and validated in mice over five months. Structural vascularity, microvascular flow velocity, mesoscale hemodynamic responses, and microvascular functional responses are repeatedly quantified in the same animals during monthly imaging sessions. Across all metrics, no significant longitudinal drift is detected (p > 0.60). Structural and flow-derived measures are markedly more reproducible than functional readouts, with within-subject coefficients of variation of 5.1% for mean flow velocity and 7.3% for vascularity, compared with 25.0% for fUS-derived cerebral blood volume responses and 53.2% for fULM-derived microvascular functional responses. Mean flow velocity shows the strongest longitudinal consistency (ICC = 0.70) and the lowest detection threshold. Behavioral testing and GFAP/Iba1 staining further reveal no memory impairment or chronic neuroinflammation. This study defines quantitative baselines, reproducibility limits, and safety evidence for awake cross-scale ultrasound imaging, providing a reference framework for longitudinal neurovascular phenotyping in preclinical disease models.

neuroscience↗

bioRxiv2026

Proteomic Identification of Pig Xenoantigens for Clinical Xenotransplantation

Xenotransplantation using genetically engineered pig organs offers a promising solution to the shortage of donor organs for life-saving transplants. However, human preformed antibodies against unknown pig xenoantigens remain a significant barrier to successful xenotransplantation. Current methods for characterizing these antibodies or xenoantigens are limited to cellular-level crossmatch assays. In this study, we developed a novel approach to identify pig xenoantigens, including peptide and glycopeptide epitopes that react with human preformed antibodies. First, human preformed antibodies against xenoantigens were enriched from plasma using immobilized pig kidney proteins. The enriched antibodies were then immobilized and used to isolate pig kidney proteins, peptides, and intact glycopeptides, followed by liquid chromatography-tandem mass spectrometry analysis. This dual-level approach identified 221 peptides corresponding to 153 proteins, with a significant enrichment of plasma membrane and extracellular proteins. Notably, 11 peptides were unique to pig sequences, suggesting their potential role in driving xenogeneic immune responses. Glycoproteomic analysis identified 122 intact glycopeptides, predominantly complex/hybrid glycoforms and Neu5Gc-containing glycans. Our method effectively identifies peptides and intact glycopeptides reactive to human preformed antibodies, providing critical insights for discovering xenoantigens. These findings could guide genetic engineering strategies and enhance recipient candidate screening for xenotransplantation, ultimately increasing the feasibility and success of xenogeneic organ transplantation.

biochemistry↗

bioRxiv2026

Hierarchical Immune Suppressive Functions of Regulatory T Cells Built on Mechanical Force

Regulatory T (Treg) cells maintain immune tolerance via contact-dependent inhibition and soluble mediators. While canonical suppressive mechanisms are well characterized, their hierarchical spatiotemporal organization remains unclear. We previously showed Foxp3 represses ER Ca2+ channel RyR2, reducing cytoplasmic Ca2+ to inactivate m-Calpain (Calpain-2) and stabilize high-affinity LFA-1-mediated Treg-dendritic cell (DC) adhesion--an early step blocking DC antigen presentation. Using m-Calpain as a molecular switch, we generated mechanically deficient Tregs (TregFD) with constitutively active m-Calpain that abrogated high-force Treg-DC adhesion. TregFD retained a near-wild-type transcriptome (differential expression limited to adhesion pathways), yet failed to suppress autoimmunity or DC-driven T cell proliferation. Stable adhesion was required for localized TGF-{beta}/IL-10 delivery to DCs. Single-cell RNA sequencing of rescued Scurfy mouse lymphoid tissues revealed organ-specific division of labor: non-mechanical modules support homeostasis/tissue repair, while mechanical modules dominate epithelial barrier/antigen presentation suppression. Combined TregFD and RyR2-deficient Tconv transfer restored wild-type Treg activity, demonstrating module synergism. These findings establish a mechanical force-centered, two-tiered hierarchical model of Treg suppression, providing a framework for targeting Treg mechanics in autoimmunity and inflammation.

immunology↗

bioRxiv2026

Adrenal Gland Macrophage-derived TGF-β Governs Vascular Permeability to Drive Monocyte Recruitment during Stress

The adrenal glands are central regulators of systemic stress responses through tightly controlled glucocorticoid production. Yet, the contribution of local immune-vascular interactions to adrenal stress adaptation remains poorly understood. Here, we investigated the role of adrenal gland macrophages in coordinating stress-induced immune remodeling and vascular function. By integrating single-cell RNA sequencing datasets across four distinct stress models, including acute cold exposure, chronic social defeat, chronic inflammation, and systemic Candida albicans infection, we identified a conserved increase in monocyte recruitment to the adrenal gland, accompanied by dynamic macrophage transitions. Comparative transcriptomic and ligand-receptor analyses identified transforming growth factor-{beta} (TGF{beta}) as a dominant macrophage-derived signal targeting adrenal endothelial cells across all stress conditions. Pharmacological blockade of TGF{beta} receptor signaling reduced endothelial activation, vascular permeability, and monocyte infiltration into the adrenal gland following stress, without directly altering resident macrophage numbers. Using genetic fate-mapping and conditional knockout models, we demonstrate that macrophage-derived, but not endothelial-derived, TGF{beta} is required to promote enhanced endothelial adhesion molecule expression, vascular fenestration, permeability, and efficient monocyte recruitment. Loss of macrophage TGF{beta} production also led to exacerbated systemic stress hormone levels. Together, these findings uncover a previously unrecognized macrophage-endothelial axis in the adrenal gland, whereby macrophage-derived TGF{beta} regulates vascular properties to support immune cell recruitment and stress adaptation. This immune-vascular crosstalk provides new mechanistic insights into adrenal homeostasis and suggests potential therapeutic avenues for disorders associated with dysregulated chronic stress.

immunology↗

bioRxiv2026

Decoding heterogeneous aging clocks and disease risk stratification using a metabolomic foundation model

Metabolomic aging clocks estimate biological age by modeling metabolite concentrations, thereby capturing aging signals from healthspan and adverse outcomes. However, existing clocks generally assume homogeneous aging trajectories and yield only a single age acceleration metric, limiting their capacity to capture inter-individual metabolic heterogeneity and characterize nuanced individual-level representations. To address these limitations, we proposed MetAgeFormer, a transformer-based metabolomic model pre-trained on nuclear magnetic resonance (NMR) metabolomic profiles from over 430,000 participants in UK Biobank via self-supervised learning. This large-scale pre-training enables MetAgeFormer to learn a metabolomic representation space that captures the complex, nonlinear structure of systemic metabolism as reflected in NMR data. Building on MetAgeFormer, we developed a mortality-informed metabolomic aging clock by fine-tuning an attached survival module, deriving age acceleration that demonstrates significant associations with multiple age-related diseases and factors. We further validated zero-shot transfer in the independent Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort. More importantly, we utilized embeddings generated by MetAgeFormer to identify 13 distinct metabolic subtypes and consolidated them into four meta-subtypes with markedly divergent susceptibility profiles for major age-related diseases, particularly type 2 diabetes and neurodegenerative disorders. This finding empirically demonstrated substantial metabolic heterogeneity across populations, persisting even at comparable levels of age acceleration. To enhance clinical applicability, we further employed contrastive learning to distill a lightweight model that approximates the learned metabolomic representation space using only 14 routine clinical blood test measurements as inputs. Both hold-out testing within UK Biobank and external validation in the China Health and Retirement Longitudinal Study replicated similar disease onset patterns across the identified subtypes, underscoring the robust generalizability of MetAgeFormer and supporting its translational potential as a scalable framework for metabolomic aging assessment and early disease risk stratification.

bioinformatics↗