Biology subjects

Zhao, Y.

Publications and source records attributed to Zhao, Y..

At least 55 records · Page 3Linked to original sources

bioRxiv2026

Age-related differences in neurometabolite concentrations during early infancy: A cross-sectional analysis of the HBCD Release 1.0

Background: Developmental trajectories of low-concentration neurometabolites such as the neurotransmitter {gamma}-aminobutyric acid (GABA), and the antioxidants glutathione (GSH) and ascorbate (Asc) across early infancy remain unexplored. Advances in spectral editing enabled the measurement of these key molecules together with high-concentration metabolites like N-acetylaspartate (NAA) and glutamate (Glu) in the HEALthy Brain and Child Development (HBCD) study, the largest longitudinal study of early brain development in the United States. Purpose: To determine the age-associated trajectories of 14 key neurometabolites during early infancy from a cross-sectional 1H-MRS dataset. Materials and Methods: HBCD utilizes ISTHMUS, an integrated MRS sequence that includes both an unedited short-echo-time PRESS acquisition and an advanced 4-step Hadamard-encoded sequence, HERCULES to enable measurement of both high- and low-concentration metabolites. Metabolite quantification was carried out by the HBCD Data Coordinating Center (HDCC) using an automated Osprey pipeline, with data from 201 infants ages 0 - 10 weeks adjusted age included in the tabular imaging results in HBCD data release 1.0. After excluding preterm born infants and data with poor linewidth or model quality metric, we tested for linear associations with adjusted age for each metabolite. Results: Concentration estimates of total NAA (tNAA), total creatine (tCr) and glutamate (Glu) as well as the combined sum of glutamate and glutamine (Glx) significantly increased across ages 0 - 10 weeks, while myo-inositol (mI) decreased. GABA and GSH showed age-related trends, but did not reach significance. Levels of the lipid precursor phosphorylethanolamine (PE) and Asc are higher in the first months than established adult values. Conclusion: Multiple metabolites showed significant age-related changes during early infancy. While GABA and GSH did not, future work will establish whether the trends suggested here contribute to linear or non-linear patterns across the first years of life.

neuroscience↗

bioRxiv2026

OmniSyn unifies target-aware molecular generation and optimization within a synthesis-native LLM framework across the human proteome

Designing target-specific bioactive molecules with actionable synthesis routes for the human proteome holds enormous potential for expanding therapeutic discovery, but remains a challenge. Existing target-aware generative models often depend on protein structures and generate molecules before assessing synthetic feasibility. Here we present OmniSyn, a protein-sequence-conditioned Mixture-of-Experts (MoE) language model that couples a task-conditioned interaction module with a synthesis-action decoder to generate molecules with explicit synthesis traces, thereby unifying de novo ligand generation, synthesizability projection and hit-to-lead (H2L) optimization within synthesis-traceable chemical space. OmniSyn is pre-trained with self-distillation and post-trained with task-specific reinforcement learning (RL) to adapt expert routing and optimize molecular properties across design modes. On unseen protein targets from MolGenBench, a real-world drug-discovery benchmark, OmniSyn achieves state-of-the-art performance across de novo design and H2L optimization, including target-awareness and hit-rediscovery metrics, despite relying only on protein sequences rather than three-dimensional (3D) pocket structures. By embedding synthesis planning into the design process, OmniSyn shifts molecular generation from a generate-then-filter paradigm toward design-with-synthesis paradigm, transforming virtual predictions into experimentally actionable candidates. Independent AiZynthFinder evaluation yielded retrosynthetic success rates of 68.47% for de novo generation and 71.92% for H2L optimization, improving over the strongest baselines by 61.3% and 184.2%, respectively, and supporting the synthetic feasibility of OmniSyn-generated molecules. In synthesizability projection, OmniSyn further converts outputs from external generative models into close analogues with improved retrosynthetic feasibility while preserving molecular similarity. Having established strong benchmark performance and external retrosynthetic feasibility, we next applied OmniSyn at human-proteome scale, spanning more than 21,000 targets. Rapid sequence-conditioned sampling enabled the construction of, to our knowledge, the largest human-proteome-scale generative virtual library, comprising 2.7 billion target-specific molecules, each accompanied by model-derived synthesis traces and target-specific prioritization scores. By enabling scalable target-specific molecular design with synthesis-aware generation across the human proteome, OmniSyn opens new opportunities for exploring previously inaccessible therapeutic targets, including those lacking experimentally resolved structures.

bioinformatics↗

bioRxiv2026

Cross-domain confidence reliability and remappability of frozen single-cell representations

Single-cell foundation models are increasingly adopted for downstream applications such as cell-type prediction. However, these predictions are often utilized without assessing their reliability, or by relying on a simple cutoff applied to the maximum softmax probability (MSP) derived from the classifier. This raises a critical question: can raw MSP be trusted as a reliable measure of confidence on unseen data (target) exhibiting diverse technical and biological variations? To address this, we introduce an audit framework to systematically evaluate the confidence estimates derived from the training set (source) against the realities of the target test set. We demonstrate that raw MSP fundamentally fails to represent true confidence when applied to target datasets. By decomposing the confidence gap between source and target data, we reveal that these discrepancies stem from a combination of rank-ordering errors and systemic probability drift. We further show that the drift can be successfully recalibrated by leveraging a small subset of labeled target data. While this recalibration improves the reliability of automated acceptance, it inherently introduces a trade-off by increasing the volume of cells requiring manual review. Ultimately, our framework establishes that the safe deployment of these models necessitates target-specific confidence adjustment using representative local labels.

bioinformatics↗

bioRxiv2026

A Cytokine Receptor Signaling Atlas Reveals How STAT Mosaics Fine-Tune T Cell Function

The extent to which JAK/STAT cytokine signaling is functionally redundant or selective remains debated. Here we engineered a double orthogonal IL-2/IL-2R{beta}/{gamma}c ternary system enabling programmable, interference-free activation of each of the 36 mammalian cytokine receptors, and their downstream six STATs, in T cells. At the membrane-proximal level, comprehensive phospho-signaling profiling revealed that while each receptor activates a dominant STAT, unique STAT activation fingerprints derived from combinatorial biases fine-tune nuanced T cell fates. At the membrane-distal level, single-cell transcriptomic atlas of all cytokine receptors confirmed that these STAT mosaics sensitively specify non-redundant transcriptional programs. STAT5-dominant receptors drove proliferative expansion at the expense of stemness; STAT3-driven programs instructed a continuum from stem cell memory to terminal effector states with preserved cytotoxic capacity and mediated superior curative antitumor responses; while other STATs specified highly restricted phenotypes. These findings decode a STAT signaling vocabulary that defines the intrinsic functional bandwidth of natural cytokines.

immunology↗

bioRxiv2026

All-by-All Cytokine Receptor Pairing Network Unlocks Coding of Non-Natural T Cell States

Cytokine receptor pairing rules, set by evolution, confine JAK-STAT signaling to a narrow region of a far larger combinatorial space. Of more than 1,200 pairings theoretically possible among the [~]36 JAK-associated human cytokine receptors, only [~]30-40 exist in nature. Using a double-orthogonal platform, we enforced pairings across the full all-by-all receptor matrix and resolved a fine-grained STAT atlas richer than the natural repertoire. Selected non-natural pairings generated emergent T cell states unpredictable from either parental receptor, with pairing orientation encoding signaling specificity. A synthetic IL-21R>IL-2R{beta} pairing, but not its reciprocal, drove a cytotoxic Tc17-like state, whereas natural IL-9R/{gamma}c drove a Tc1 fate despite similar STAT activation, showing that rebalancing quantitative STAT combinatorics can modulate T cell fate. Recombining IL-31R, not expressed in T cells, with STAT-biased receptor partners generated diverse states, several with superior antitumor efficacy. These findings define a non-natural pairing code for engineering synthetic T cell fates.

immunology↗

bioRxiv2026

Loss of calbindin and rise in pT217-tau in aging monkey prefrontal cortical dendrites

INTRODUCTION: Tau pathology in Alzheimers disease preferentially afflicts excitatory neurons in the limbic and association cortices that utilize high levels of calcium signaling to perform cognitive operations. This includes the layer III pyramidal cells in the dorsolateral prefrontal cortex (dlPFC) that subserve higher cognition, which express the calcium-binding protein, calbindin, when young and healthy, but lose calbindin and develop tangles and degenerate in Alzheimers disease (AD). These data suggest that loss of calbindin may be associated with the emergence of tau pathology. However, the relationship between calbindin and early-stage, soluble tau pathology is challenging to study in human brains, as soluble pTau dephosphorylates within 15min postmortem. In contrast, the relationship between calbindin and soluble pT217-tau expression can be studied in aging macaques with naturally-occurring tau pathology, where perfusion fixation is possible to capture phosphorylation state in situ. METHODS: The current study used multiple-label-immunofluorescence to label MAP2-positive dlPFC layer III pyramidal cells for calbindin and pT217-tau in macaque brains across the adult age span (8-34.5yrs). The study employed a semi-automated CellProfiler workflow to identify labeled pyramidal cell dendrites the cellular compartment where tau pathology begins in AD. RESULTS: Calbindin expression decreased with age, while pT217Tau increased with age. Specifically, the ratio of calbindin/pT217-tau within a dendrite decreased with age, and was especially prominent in the aged macaques with long-term inflammatory disorders. DISCUSSION: These data suggest that the loss of calbindin in dendrites with advancing age, and especially with inflammation, contributes to the rise of tau pathology and the risk of AD.

neuroscience↗

bioRxiv2026

Nanoscale 3D profiling of the T cell membrane reveals CD2 enrichment at microvilli tips, positioning adhesion near TCR zones in the immunological synapse

The T cell membrane features a specialized molecular and topological organization critical for signaling and immune function. During antigen detection, T cells utilize finger-like protrusions called microvilli to dynamically scan the antigen-presenting cell (APC) surface. Enriched with T cell receptors (TCRs) and key signaling molecules, microvilli serve as primary signaling hubs, yet their nanoscale architecture remains poorly defined. Upon antigen engagement, TCR activation drives the formation of the immunological synapse (IS), a highly organized membrane contact with the APC critical for T cell function. However, profiling IS architecture at the nanoscale remains technically challenging. Here, we introduce NanoMAP (Nanoscale Membrane Architecture Profiling), an expansion microscopy-based platform that resolves receptor organization on T cell microvilli and within the IS at 35-60 nm resolution. Using NanoMAP, we show that effector CD8+ T cells possess a 5.1-fold higher microvillar density than naive CD8+ T cells. Mapping the adhesion receptors CD2 and LFA-1 on the T cell surface reveals that CD2 exhibits a stronger preference for localizing to microvilli tips compared to TCR and LFA-1. Within microvilli, CD2 displays a strong spatial association with TCR clusters, contrasting with a markedly weaker TCR-LFA-1 association. At the IS, TCR and CD2 co-occupy close membrane contacts, while LFA-1 is excluded to more distal regions. Upon termination of activation, TCR clusters selectively disengage from microvilli, whereas CD2 and LFA-1 persist. These results suggest a coordinated, activation-dependent organization of adhesion receptors that drives microvillar adhesion during scanning and stabilizes IS membrane contacts. Together, NanoMAP establishes a powerful framework for dissecting nanoscale membrane architecture and spatial signaling in T cells.

immunology↗

bioRxiv2026

In situ Discovery of Immune Repertoire Reveals Antitumor Immunity and Therapeutic Antibodies

Spatial transcriptomics offers a glimpse into the immunology of tissues. However, limitations in spatial transcriptomics preclude the detection of highly diverse, low-abundance, and previously unknown sequences, including immune repertoires and microbiota. Here, we introduce Archimap, a spatial transcriptomic platform that simultaneously profiles spatial transcriptomes, immune repertoires, and microbiota from formalin-fixed paraffin-embedded (FFPE) tissues. Using Archimap, we profile the spatial localization of TCRs, BCRs, and the microbiota landscape in archived clinical tissues at single-cell resolution. Through comprehensive benchmarking, we validate Archimaps performance and fidelity. Archimap in situ assembles the immune complex and reconstructs the clonal evolution of antibodies. Together, Archimap shows the power of in situ discovery of functional immune repertoires for their antitumor immunity.

bioinformatics↗

bioRxiv2026

The live attenuated DGAT1-knockout whole-cell Toxoplasma vaccine confers protective immunity against acute and chronic toxoplasmosis

The intravacuolar parasite Toxoplasma gondii scavenges fatty acids from host mammalian cells and stores excess in lipid droplets. To investigate the physiological relevance of neutral lipid storage in Toxoplasma, we generated a mutant lacking DGAT1, an ER-localized enzyme that synthesizes triacylglycerols, from the virulent type I RH strain of T. gondii. Compared to WT, RH {Delta}DGAT1 parasites grow poorly in mammalian cells, form few LD, suffer from lipotoxicity, and do not cause disease or lethality in immunocompetent or immunodeficient mice. Importantly, mice immunized with RH {Delta}DGAT1 parasites mount strong, long-term immune responses involving both cellular and humoral components, with higher levels of T. gondii-specific IgG antibodies, effector memory T cells, and both pro-inflammatory and anti-inflammatory cytokines, indicating a mixed Th1/Th2 response with Th1 predominance. This immunity provides complete, long-lasting protection (up to 6 months) against rechallenge from homologous type I (acute infection) and heterologous cyst-forming type II (chronic infection) T. gondii strains. Additional analyses reveal that IFN-{gamma}, CD8+ T cells, as well as B cells are crucial for defending against type I T. gondii in immunized mice. Overall, our live-attenuated RH {Delta}DGAT1 strain is a promising vaccine candidate and a model for studying immune responses that control T. gondii infections.

microbiology↗

bioRxiv2026

SpatialMOC: Accurate reconstruction of spatial multi-omics landscapes through cross-modality prediction

Spatial multi-omics technologies provide unprecedented opportunities to characterize tissue organization by measuring complementary molecular layers within their native spatial context. However, simultaneous profiling of multiple molecular modalities remains technically challenging, limiting the widespread application of spatial multi-omics and leaving most studies reliant on single-modality measurements. Here, we present Spatial Multi-Omics Cross-prediction (Spatial-MOC), a computational framework that reconstructs missing spatial molecular modalities by integrating spatial context with cross-modality representation learning. Across multiple tissues, molecular modalities, developmental stages, sequencing platforms, and degraded datasets, SpatialMOC consistently outper-formed existing computational approaches in bidirectional molecular prediction. Beyond accurate prediction, SpatialMOC faithfully reconstructed tissue architecture, preserved dynamic molecular and regulatory heterogeneity, and recovered biologically meaningful spatial landscapes from technically compromised measurements. Together, these results establish SpatialMOC as a general framework for spatial multi-omics reconstruction, extending the analytical value of existing spatial omics datasets and facilitating comprehensive investigations of tissue organization, development, and disease.

bioinformatics↗

bioRxiv2026

Targeting Tumor-derived Sphingosine Kinase 2 Unleashes Antitumor Immunity and Improves Survival of Mice with Group 3 Medulloblastoma

Group 3 medulloblastomas (G3MB) carry the worst prognosis among medulloblastoma subtypes, yet molecularly targeted therapies remain elusive. Standard treatments cause severe long-term morbidity in survivors. Here, we identify tumor-derived sphingosine kinase 2 (SPHK2) as an essential driver of G3MB initiation and progression. SPHK2 exacerbates local immunosuppression by suppressing cytotoxic T-cell and NK-cell activity while promoting regulatory T-cell infiltration. Genetic or pharmacologic SPHK2 inhibition using Opaganib attenuates pro-survival tumor signaling and restores anti-tumor immunity, significantly improving survival in syngeneic G3MB mouse models. Combining Opaganib with fractionated low-dose radiation (f-LDRT) further enhances antigen presentation and reprograms tumor-associated myeloid cells toward an anti-tumor phenotype. This combination therapy markedly prolongs survival without inducing significant toxicity. Overall, our study establishes SPHK2 as a previously unrecognized therapeutic target and presents a safe, effective, microenvironment-reprogramming regimen for G3MB. One Sentence SummaryDirect inhibition of tumor-derived SPHK2 overcomes local immunosuppression and downregulates pro-survival signaling in Group 3 medulloblastoma, while combination with fractionated low-dose radiation further enhances anti-tumor immunity and significantly improves survival.

cancer biology↗

bioRxiv2026

TBpop: an open-access genomic portal integrating genomic variation, population genetic statistics, phylogeny, pangenome composition, and strain metadata of epidemic Mycobacterium tuberculosis strains from China

Tuberculosis remains a major global public health threat. While whole-genome sequencing has transformed our understanding of the causative agent, Mycobacterium tuberculosis (MTB), existing genomic databases are highly fragmented and often underrepresent structural variations (SVs). Furthermore, critical population-genetic statistics are rarely integrated with phylogenetic and geographic context, forcing researchers to reconcile separate datasets manually. To address this gap, we developed TBpop (https://tbpop.chinacdc.cn), an open-access, integrated population genomics portal. TBpop is built from 420 clinical MTB isolates selected from the first national drug resistance baseline survey in China. The portal integrates isolate metadata, pangenome categories, SNPs, SVs, IS6110 insertion sites, strain phylogeny, and gene-level statistics, and provides three interactive explorer modules: the Population Explorer, the Statistics Explorer, and the Variation Explorer. Additionally, a User Analysis module allows researchers to run population genetic workflows on their own alignments. TBpop provides an integrated platform for exploring genome plasticity, signatures of positive selection, and conservation patterns of functionally important genes in MTB.

bioinformatics↗

bioRxiv2026

Neurofeedback Training on Motor Cancellation Enhances Peripheral but not Cortical Beta Band Oscillations

Beta oscillations (13-30 Hz) are a prominent sensorimotor neural rhythm and an important biomarker in neurorehabilitation. These oscillations propagate along the corticospinal pathway and are expressed in the discharge patterns of spinal motor neurons, enabling the assessment of corticomuscular coupling. Moreover, peripheral beta band oscillations have recently emerged as a potential control signal for motor augmentation interfaces. However, it remains unclear whether peripheral beta activity simply reflects cortical oscillations or is partly shaped by peripheral mechanisms, and to what extent it can be voluntarily controlled. To address these questions, we developed a 10-day neurofeedback protocol in which participants learned to up-regulate peripheral beta band activity. Subjects were trained to exploit movement cancellation, a behaviour naturally associated with increased cortical and muscle beta band activity, as a two-state strategy to voluntarily modulate peripheral beta band power. Each session included a guided familiarization phase based on a GO/NO-GO task, in which participants familiarized with movement cancellation through guided visual cues, followed by an asynchronous control phase in which they self-initiated the same strategy without external guidance to increase peripheral beta band activity in a target window. Participants progressively improved their ability to voluntarily modulate peripheral beta band activity. Peripheral beta band power during movement cancellation increased significantly across training days in both the familiarization and asynchronous control phases. Intramuscular coherence in the beta band also increased, indicating enhanced common synaptic input to the motor neuron pool in this band. In contrast, cortical beta power and corticomuscular coherence remained unchanged. Together, these findings demonstrate that peripheral beta band activity is a dynamic neural feature that can be voluntarily shaped through training, supporting its potential as a non-invasive control signal for future neurorehabilitation and motor augmentation technologies.

bioengineering↗

bioRxiv2026

Defining the genetic landscape of acid and oxidative stress tolerance in Streptococcus mutans by pooled CRISPR interference screening

Bacterial stress tolerance is a fundamental ecological adaptation that enables survival, competitive fitness, and persistent colonization under fluctuating environmental conditions. Dental caries remains a major global health burden. As an etiologic agent in caries, Streptococcus mutans (S. mutans) has the ability to adapt to sudden and substantial acid and oxidative stress. Nevertheless, the genome-wide genetic programs that support these stress tolerances remain incompletely defined. Herein, we established a xylose-inducible, genome-wide pooled CRISPR interference (CRISPRi) platform in S. mutans and performed parallel functional screens under acidic conditions (pH 5.0) and low-dose hydrogen peroxide (H{square}O{square}) stress. Using predefined screening thresholds, we identified 422 genes whose repression reduced fitness during acid challenge and 337 genes whose repression reduced fitness during H{square}O{square} exposure. Functional and network analyses revealed that stress tolerance is strongly constrained by core physiological processes, including RNA (particularly transfer RNA) metabolism, macromolecule maintenance, and damage repair. Comparative analyses further indicated that growth-associated pathways displayed opposite trends between the two stresses, consistent with a stress-dependent allocation trade-off. These two complementary findings redefine the dual-stress adaptation paradigm of S. mutans: stress resistance is not determined merely by canonical stress signaling, but requires intact core physiological modules, and bacteria tune resource allocation dynamically to adapt to divergent stress microenvironments. Collectively, the present study provides a comparative, genome-scale functional map of acid- and oxidative-stress tolerance in a key oral pathogen and identifies genetic determinants for mechanistic studies and anti-caries interventions.

microbiology↗

bioRxiv2026

Identifying the causal allele in the CD40 autoimmune locus enables discovery of context-specific trans-effects in B cells

Thousands of genetic variants are associated with autoimmune diseases, but causal variants, their mechanisms, and the pathogenic context in which they act are elusive. Knowledge of pathogenic contexts may enable effective targeted therapies, instead of broad immunosuppressive approaches. First, to focus on the genetics of immune response, we used surface marker CITE-seq data from 1,055,857 peripheral blood mononuclear cells from 356 individuals. We defined genetic associations to 148 surface proteins across eight cell types. We observed a signal in the CD40 locus, implicated in rheumatoid arthritis (RA) and other autoimmune conditions. RA risk variants increased CD40 protein expression by [~]20% on B cells, but with minimal mRNA effects. Second, we deployed base-resolution genome editing, with CRAFT-seq, capturing genomic DNA sequence at the edited site and multimodal phenotypes at single-cell resolution. We defined a single causal allele, rs1883832, within the Kozak motif. Third, we edited this allele, in primary B cells and conducted CRAFTseq to demonstrate trans-effects in >200 genes. These effects were only in the light zone germinal center-like state. Importantly, these trans-effects were not seen in population-scale cohorts of unstimulated B cells. This represents a framework to define disease causal alleles, their cis- and trans-effects. It demonstrates the power of defining causal genetic variation to find trans-effects through editing, which cannot easily be found in population studies.

genetics↗

bioRxiv2026

CD133+ progenitor cells promote pulmonary hypertension through CXCR4 signaling

BackgroundPulmonary hypertension (PH) is characterized by pulmonary vascular remodeling and smooth muscle cell accumulation, but the progenitor-like cells that contribute to this process remain incompletely defined. MethodsWe combined analyses of human pulmonary arterial hypertension lungs and experimental PH models with bulk and single-cell RNA sequencing, lineage tracing, inducible ablation of CD133+ cells, and conditional deletion of Cxcr4 in CD133+ cells. ResultsCD133 expression was markedly increased in human and experimental PH lungs. Transcriptomic analyses identified inflammatory, metabolic, chemokine-associated, and smooth muscle cell-like programs in CD133+ cells from PH lungs. Lineage tracing showed that CD133+ cells contributed to endothelial and smooth muscle cell compartments during experimental PH. Genetic ablation of CD133+ cells attenuated hypoxia-induced PH and pulmonary vascular remodeling, whereas Cxcr4 deletion in CD133+ cells reduced PH severity. ConclusionsCD133+ progenitor cells are functional contributors to pulmonary vascular remodeling, and CXCR4 signaling mediates their pathogenic activity. Targeting pathogenic CD133+ cell states or CXCL12/CXCR4 signaling may provide a strategy to limit vascular remodeling in PH.

pathology↗

bioRxiv2026

Genomics and biogeography of novel Trichodesmium isolates from the Sargasso Sea

The cyanobacterium Trichodesmium is important in ocean nitrogen (N) and carbon (C) biogeochemistry through fixation of N2 gas and CO2 into NH3, amino acids, and carbohydrates. These essential compounds then fuel primary productivity, bacterial and grazing communities, and upper trophic levels. In the oligotrophic surface oceans, they are one of few taxa capable of utilizing both atmospheric C and N. Trichodesmium fall into 2 diazotrophic phylogenomic clades named "Thieb" and "Tery" after the most well characterized botanical species, Trichodesmium thiebautii and Trichodesmium erythraeum. Thieb is commonly the most abundant clade in all major ocean regions and paradoxically, is also the least characterized in terms of morphology, genomics, and ecophysiology due to underrepresentation in culture collections. In this study, 25 novel strains of Thieb were enriched in culture from the Sargasso Sea, phenotypically characterized, and whole genome sequenced. Cultures were found to belong to four well-supported subclades (i.e., ThiebA, ThiebB, ThiebC, and ThiebD). Trichodesmium subclades were found to have environmentally relevant differences in colony shape, gene content, cell dimensions, and inferred temperature range. Specifically, ThiebD and ThiebC had the largest biovolumes whereas ThiebA and ThiebB had the smallest. Importantly ThiebB also became relatively more abundant only in bulk seawater metagenomes collected at high temperature (>28{degrees}C). This finding indicates that as ocean warming continues, there may be shifts in Trichodesmium community structure towards smaller cells, which would affect overall C+N standing stocks.

microbiology↗

bioRxiv2026

CENO: A Genome-Scale World Model for Evolutionary Sequence Interpretation and Programmable Regulatory Design

DNA encodes biological function across a continuum of sequence scales, from single-nucleotide and motif-level grammar to regulatory neighborhoods, chromatin-scale organization and evolutionary constraint. A useful model of genomes should therefore do more than classify short sequence windows: it should maintain nucleotide-resolution state over long contexts, score counterfactual mutations, condition on homologous sequence evidence and generate candidates that can be evaluated against structural or functional objectives. We define such a system operationally as a genomic world model: a general-purpose generative model of genome sequence space that unifies sequence understanding and sequence design through a shared state and likelihood interface. Here we introduce CENO, a family of long-context generative genomic world models designed to preserve local DNA grammar while extending usable context to regulatory and chromatin scales. CENO combines Mamba sequence-mixing layers, sparse attention layers and mixture-of-experts capacity in a single autoregressive backbone, and is trained at 300M, 600M and 1B parameter scales with a staged curriculum that progresses from 8k-token cross-domain genomic pretraining to 131k- and 1M-token whole-genome long-context continuation. We evaluate CENO under a unified world-model benchmark paradigm spanning retrieval, representation, counterfactual perturbation, reconstruction, evolutionary conditioning and design. CENO retains practical long-context inference and retrieves distal sequence in synthetic assays. In zero-shot long-context analyses, without task-specific fine-tuning, long-context continuation yields annotation- and chromatin-boundary-associated attention patterns and frozen-state representations that generalize across human cell types and mouse cell or tissue settings. To incorporate evolutionary information, we further post-train CENO on packed real multiple-sequence-alignment contexts and score variants by reference-mutant likelihood deltas, improving matched variant-effect prediction and producing evolutionary enrichment signals across species. Complementing these perturbation-based variant tests, we evaluate zero-shot long-sequence generation by partial-gene continuation, asking whether the model can recover withheld gene-scale sequence structure across eukaryotic, bacterial and archaeal species; recovery improves with model scale and later whole-genome long-context training. Finally, we use CENO as the backbone for a cell-type-specific enhancer design workflow in mouse cortex, coupling a CENO-based accessibility oracle with conditional supervised fine-tuning and oracle-guided reinforcement learning. Together, CENO provides a genome-scale sequence world-model framework for sequence interpretation, evolutionary reasoning, gene-scale reconstruction and programmable regulatory sequence generation.

genomics↗