Biology subjects

Xu, T.

Publications and source records attributed to Xu, T..

At least 19 recordsLinked to original sources

bioRxiv2026

Aberrant cortical reorganization and astrogliosis underlie memory deficit post-traumatic brain injury: a longitudinal multiscale multimodal study

Integrating high-dimensional neuroimaging data across spatial scales remains a fundamental bottleneck in quantitative neuroscience. Here, we apply emerging analytical techniques including functional connectome gradient mapping and 3D segmentation of whole-brain cellular resolution microscopy data in a mouse model of traumatic brain injury. At the macroscale, longitudinal functional MRI revealed a transition from frontal hyperconnectivity to hippocampal hypoconnectivity and unimodal-transmodal gradient contraction indicating network dedifferentiation. At the cellular scale, deep-learning based segmentation of whole-brain light-sheet fluorescence microscopy volumes quantified regional astrogliosis, neuronal hyperactivity, and single-axon loss undetected by standard diffusion MRI. Multiscale multimodal data integration not only identified quantitative biomarkers predictive of memory performance post-injury but also enabled validation of in vivo modalities against more sensitive ex vivo histology. This study demonstrates a framework for probing complex multi-network cognitive functions and neurological disorders of network dysfunction.

neuroscience↗

bioRxiv2026

Brain-thought states during movie-watching reflect shared sensory anchors and film- and person- specific processes in association cortex

Naturalistic paradigms such as movie-watching have transformed functional neuroimaging by capturing cognitive dynamics during rich, real-world experiences. A fundamental challenge, however, is determining whether the neural signatures identified during these experiences reflect general principles of cognition or the specific contexts in which they are measured. To address this question, we combined fMRI data from participants watching eight diverse films with continuous reports of ongoing thought. We found that the generalizability of brain-thought relationships depends on the cognitive process being recruited. Neural signatures linked to sensory processing remained robust across films, whereas those involving memory-related association networks depended on how we modelled between-film variability, indicating that cognition becomes increasingly context-dependent along the cortical hierarchy from sensory to association cortex. Within default and limbic networks, we identified two contrasting memory-related states with opposite consequences for comprehension. One, reflecting the integration of prior knowledge with ongoing experience, predicted better narrative comprehension and was revealed only after accounting for between-film variability. The other, associated with stimulus-independent thought, predicted poorer comprehension and emerged only when between-film variability was not modelled. Together, these findings establish when brain-experience relationships measured during naturalistic paradigms can generalize across contexts and when they instead reflect interactions between cognition and the specific environments in which cognition unfolds. More broadly, the consistency of a given brain-experiential association may depend less on where it occurs in the brain and more on the viewer's contextual experience: perceiving the film, drawing on memory to understand it, or drifting away.

neuroscience↗

bioRxiv2026

mtDNA Activates AIM2-Mediated PANoptosis in Interstitial Macrophages to Promote Immune Checkpoint Inhibitor Pneumonitis

Immune checkpoint inhibitor pneumonitis (CIP) is a serious immune-related adverse event, yet its mechanisms remain unclear. Using a mouse model of anti-PD-1-induced CIP, we show that PD-1 blockade induces pulmonary inflammation and dysfunction, accompanied by accumulation of lung interstitial macrophages (IMs) and PANoptosis, preferentially in IMs rather than alveolar macrophages. Anti-PD-1 also causes mitochondrial stress and cytosolic mtDNA accumulation. AIM2 deficiency attenuates pulmonary inflammation, IM accumulation, and PANoptosis-associated signaling; NLRP3 or GSDMD deficiency similarly reduces inflammation and PANoptotic responses. In macrophages, TNF- plus IFN-{gamma} induces mitochondrial dysfunction and cytosolic mtDNA accumulation, and mtDNA amplifies cytokine-induced cell death and caspase-1/GSDMD activation, which is abolished by AIM2 deficiency. Thus, AIM2-mediated PANoptosis in lung IMs drives CIP, and mtDNA-associated innate immune signaling amplifies inflammatory macrophage death during CIP.

immunology↗

bioRxiv2026

Stimulus overlap confounds generalization in EEG foundation-model benchmarks

Neural decoding benchmarks commonly evaluate generalization across participants while reusing the same experimental stimuli in training and test sets. This design cannot distinguish transfer to new stimuli from recognition of familiar material, a distinction that is increasingly important for electroencephalography (EEG) foundation models, pretrained models designed to learn reusable neural representations. Here we introduce a training-matched audit that varies stimulus overlap while holding test trials, training participants and stimulus counts fixed. Across ten tasks from seven EEG emotion datasets, stimulus overlap increased differential-entropy decoding accuracy in every task and accounted for 77% of above-chance seen-stimulus performance on FACED-9. Across representations, overlap-dependent gains tracked stimulus identifiability; for CBraMod and LaBraM, advantages over input-matched handcrafted features were concentrated on familiar stimuli. Architecture-matched random controls nevertheless showed transferable benefits of pretraining on unseen stimuli. These results separate participant generalization from stimulus generalization and provide a framework for evaluating both in neural decoding benchmarks.

neuroscience↗

bioRxiv2026

Ultrafast and Accurate Selection of High-Quality Protein Complex Models from Large-Scale Prediction

In the post-AlphaFold era, advances in protein complex structure prediction have enabled the generation of thousands of candidate models per target, making the accurate and efficient selection of high-quality models from large candidate pools a critical challenge. Existing model-selection approaches commonly employ estimated model accuracy (EMA) methods, which include single-model and consensus-based methods. The former lack cross-model comparative information, whereas the latter often incur high computational costs from pairwise structural alignments and are influenced by candidate-pool quality. Here, we present DeepUMQA-Selection, a two-stage framework combining single-model accuracy estimators for preliminary selection with Ultrafast Shape Recognition (USR)-based structural alignment for consensus selection. Preliminary selection enriches the candidate pool with high-quality models and reduces subsequent structural comparisons, while USR-based structural alignment enables efficient consensus selection without explicit structural superposition. On the retrospective CASP16 QMODE3 benchmark dataset, DeepUMQA-Selection outperformed all participating methods and predictor-derived confidence metrics, as measured by mean Top-5 weighted penalty. The advantage was particularly pronounced for heteromeric targets, with a 13.9% improvement in model-selection performance relative to the best-performing CASP16 participating method. Furthermore, a confidence-assisted strategy enabled the framework to accommodate larger protein complexes while maintaining superior model-selection performance over competing methods, supporting its scalability. Runtime benchmarking showed that the proposed USR-based structural alignment achieved a substantial 317-fold speedup over US-align for candidate pools containing 1,000 models, demonstrating its computational efficiency for large-scale model selection. Our results demonstrate that DeepUMQA-Selection enables accurate and computationally efficient selection of high-quality protein complex models from large-scale candidate pools.

bioinformatics↗

bioRxiv2026

Ongoing brain state dynamics shape spontaneous awake reactivation and its relationship with memory

Memory reactivation is thought to support memory consolidation, but how ongoing brain states shape awake reactivation and its relationship to memory remain poorly understood. Here, we asked whether spontaneous hippocampal-cortical reactivation during awake rest is embedded in ongoing brain state dynamics by examining its coupling with whole-brain activity patterns and infra-slow global signal (GS) fluctuations, and whether these dynamics shape its relationship to memory retention. Reactivation preferentially occurred during an internally oriented brain state characterized by default mode network activation and dorsal attention network suppression, as well as states that reinstated encoding-related visual activity. Reactivation also clustered around the peak of infra-slow GS fluctuations and in high GS amplitude time windows, features previously associated with lower arousal. Importantly, the relationship between reactivation and memory depended on the ongoing brain state: distinct whole-brain activity patterns expressed during reactivation were associated with opposing memory outcomes, while reactivation during high GS amplitude windows was positively associated with retention. Together, these findings show that awake memory reactivation is embedded in dynamic whole-brain states that may influence its memory outcome, suggesting that recurring brain states may provide favorable windows for awake memory consolidation.

neuroscience↗

bioRxiv2026

Population Intelligence: Smallest-Model Numerical Experiment

We introduce a minimal individual based model for population intelligence that retains three structures absent from standard single agent descriptions: a manifold valued intelligence state, endogenous population turnover, and heritable lineage dependent reproductive fitness. The validated S^1 intelligence diffusion was extended with neutral logistic birth death dynamics. Each individual occupies an angle on the circle S^1; noisy mean field attraction produces collective synchronization; logistic birth death dynamics regulate population size; and reproductive rates can depend on lineage and position relative to designated safe or unsafe directions. This nested construction is intentionally small enough to admit an analytic equilibrium benchmark while remaining capable of selection, invasion, and safety transitions. Population size returned toward carrying capacity from (N(0)=200,800,1600), with tail means (775.7,800.3,805.1) for K=800. Across K=100 to 1600, the intelligence equilibrium error scaled as K^{-0.513} , consistent with the canonical K^{-1/2} fluctuation rate, while the extinction rate was zero in 120 runs over the simulated horizon. The mean order parameter remained close to the original theoretical (r*=0.7242), as neutral demography predicts. Directional reproductive selection shifts a safety score to +0.7447 or -0.7448 while synchronization remains approximately 0.75, demonstrating that coordination is not equivalent to safety. These are controlled numerical results, not asymptotic proofs.

bioinformatics↗

bioRxiv2026

DNA Sequence-Programmed Protein Coronas Determine Intracellular Fate and Proteostatic Stress of Carbon Nanotubes

Single-walled carbon nanotubes (SWCNTs) show promise for optical biosensing, imaging, and drug delivery, but turning them into safe, precision nanomedicine tools requires understanding how nanotube surface chemistry dictates recognition and processing by cells. Like other nanomaterials, carbon nanotubes acquire a biomolecular corona on contact with biological fluids, and corona identity is increasingly recognized as central to sensor performance and drug delivery efficacy. However, whether corona identity also governs the intracellular fate of carbon nanotubes remains largely unknown. Here, we show that the single-stranded DNA wrapping of (6,5)-enriched single-walled carbon nanotubes reprograms their protein corona, intracellular trafficking, and macrophage response. By profiling (AT)15, (GT)15, and (CT)15 wrapped SWCNTs, we show that the wrapping sequence programs both the protein corona and the resulting proteostatic stress on macrophages. Photoluminescence imaging and confocal Raman measurements reported that (AT)15 is internalized the most yet leaves the proteome and nanotube structure largely undisturbed, whereas (CT)15, taken up the least, undergoes the most aggressive intracellular degradation and drives the highest oxidative and proteostatic stress. Corona proteomics indicated that all three tested nanotubes form coronas with distinct functional identities that are responsible for divergent intracellular routes. Time-resolved intracellular proteomics combined with functional assays resolved how the host cell reorganizes its biomolecular complexity over time, including oxidative outputs, aside from a sequence-independent core response involving particle engagement, phagosomal sorting, and lysosomal processing. These findings provide mechanistic insight into nanomaterial-cell interactions and the wrapping sequence as a tunable, nucleotide-level design handle for controlling the intracellular fate of carbon nanomaterials, with potential implications for safe and effective nanomedicine platforms.

bioengineering↗

bioRxiv2026

Trex-QTL: A mixture-model for identification of genetic effects with global effects on molecular phenotypes

While thousands of cis expression quantitative trait loci (cis-eQTLs) have been reliably identified, detecting trans-eQTL effects has proven to be challenging due to insufficient statistical power, lack of comparable tissues and cohorts, and low reproducibility across studies. Here, we present Trex-QTL, a novel trans-eQTL detection method that models eQTL summary statistics as a mixture consisting of both target gene and null associations. Compared to other recently developed methods, Trex-QTL has improved power for trans-eQTL detection and employs a simplified framework, requiring only eQTL association summary statistics as input. We performed extensive simulations to characterize the conditions under which trans-eQTLs are detectable by Trex-QTL across a range of effect sizes and numbers of target genes. We applied Trex-QTL to the Depression Genes and Networks (DGN) dataset and replicated two well-established trans-eQTLs at ARHGEF3 and IKZF1. We then applied Trex-QTL to the deeply characterized heterogeneous stock (HS) rat cohort with matched brain transcriptomic and genomic data, identifying 7 top-scoring, linkage disequilibrium-independent trans-eQTLs. One previously unreported trans-eQTL is at the locus harboring Jag2, a critical ligand for the Notch signaling pathway, which is associated with decreased Jag2 expression and decreased expression of multiple downstream genes including known Notch targets. A second example is a strong trans-eQTL overlapping a cluster of interferon genes associated with interferon-response genes including C4a and Parp14. We show evidence that this signal is mediated by a cis-eQTL for a cluster of interferon ligand genes that operate upstream of interferon receptor signaling. Both of these signals co-localize with association signals for a range of other phenotypes measured in this cohort. Overall, we demonstrate that Trex-QTL represents a powerful method to identify trans-eQTLs with global effects on molecular phenotypes and identify novel biologically compelling examples of such loci.

genomics↗

bioRxiv2026

IL-1β/IL-6 signaling circuit in the tumor microenvironment drives prostate cancer development

Despite considerable progress in elucidating mechanisms leading to castration-resistant prostate cancer (CRPC), insight into the early stages of prostate cancer initiation and progression remains limited. Genomic drivers of prostate cancer initiation have been defined through patient tumor sequencing, but the subsequent events responsible for local tissue invasion are poorly understood. Here we leverage a well-studied genetically engineered mouse prostate cancer model (Hi-Myc) that, based on robust and reproducible kinetics for transitioning from pre-invasive prostatic intraepithelial neoplasia (PIN) to invasive prostate adenocarcinoma (PCa), provides an ideal system to systematically address this question using single-cell analysis. Surprisingly, the transcriptomic profiles of early PIN lesions are indistinguishable from those of late-stage, highly invasive tumor cells, suggesting that MYC activation at the PIN stage establishes a transcriptional program that is fully capable of driving invasion but is restrained by the local tumor microenvironment (TME). Indeed, we find that progression to PCa is associated with progressive infiltration of IL-1{beta}+ tumor-infiltrating macrophages at the PIN stage that, based on immunodepletion and cytokine neutralization experiments, are required for the PIN-to-PCa transition. Mechanistically, IL-1{beta} from macrophages acts directly on prostate fibroblasts, leading to the release of IL-6, which drives invasion by activating IL-6R in tumor cells. Collectively, these findings identify a pro-tumorigenic IL-1{beta}/IL-6 signaling circuit mediated through local macrophages and fibroblasts that unleashes the full oncogenic potential of a cancer driver (MYC) activated at the PIN stage. We also find evidence of this circuit in other (non-MYC-driven) prostate cancer models as well as human prostate and lung adenocarcinoma, with implications for TME-specific targeted therapeutics in early-stage disease.

cancer biology↗

bioRxiv2026

Manifold Agentic Reasoning: Extending Agentic POMDPs and Post-Training Reasoning to Riemannian State and Reasoning Spaces

Agentic reasoning systems increasingly interact with environments whose states are only partially observed, dynamically evolving, and constrained by physical, biological, or logical structure. Existing agentic reasoning frameworks often model internal reasoning, tool use, and post-training adaptation using flat latent representations and struggle in curved manifold space environments. However, many scientific and embodied domains naturally lie on curved state spaces, including tissue geometry, developmental trajectories, protein conformations, robotic configuration spaces, and constrained physical systems. We introduce Manifold Agentic Reasoning, a geometric framework that extends agentic reasoning from Euclidean latent spaces to Riemannian manifolds. In the proposed framework, observations are encoded as manifold-valued states, memory is retrieved by geodesic similarity, candidate hypotheses are generated in tangent spaces, predicted transitions are projected by exponential maps, and decisions are admitted through verification-gated commitment or repaired by manifold self-correction. We further extend the framework to graph-agentic manifold reasoning, where node states live on manifolds and neighbor information is transported by logarithmic maps before attention-based aggregation. Manifold agent reasoning moves AI past brittle, prompt-chained templates to solve four critical production flaws: silent hallucinations and reasoning drift, brittle tool and context misuse, the black-box evaluation problem and stiff behavior profiles. To evaluate the framework, we introduce a Curved Tissue Manipulation and Recovery benchmark in which an agent must repair damaged tissue on a curved manifold. Simulated results show that the full manifold-agent substantially outperforms both a baseline reasoning agent and a full flat-agent reasoning system, achieving higher recovery success, lower geodesic shape error, lower pattern error, and fewer invalid transitions. Curvature and ablation studies indicate that the performance gain is driven by geometry-aware reasoning, verification, memory, and self-repair. These results suggest that manifold-aware agentic reasoning provides a principled foundation for reliable scientific and embodied AI systems operating in curved, constrained, or mechanistic domains.

bioinformatics↗

bioRxiv2026

METTL1 regulates glioma proliferation through internal m7G methylation of EPHA2

BackgroundMethyltransferase-like 1 (METTL1) is highly expressed in organs like the pancreas but less so in the brain. The METTL1-WDR4 complex catalyzes N7-methylguanosine (m7G) methylation in tRNA, miRNA, mRNA, and rRNA, which impacts RNA stability and function. These modifications affect mRNA translation and tRNA functionality, influencing protein production and cellular activities. Such modifications can regulate tumor growth, invasion, and metabolism by selectively controlling protein expression. MethodGene expression data from public databases were analyzed to compare METTL1 expression in normal and tumor tissues. Western blot (WB) and immunohistochemistry (IHC) were used to quantify METTL1 levels in glioma samples and assess their prognostic significance. Cell viability, migration, invasion, and proliferation were evaluated using Cell Counting Kit-8 (CCK-8), wound healing, Transwell, cell cycle analysis, and colony formation assays. RNA immunoprecipitation PCR (RIP-PCR) identified m7G methylation sites on EPHA2 mRNA, and RNA stability was assessed with actinomycin D. ResultsBioinformatics analysis revealed that METTL1 is overexpressed in gliomas, correlating with poor prognosis. Knockdown of METTL1 significantly affected cell proliferation, migration, and invasion. RNA sequencing (RNA-seq) and m7G analysis identified EPHA2 as a downstream target, influencing the cell cycle via the AKT pathway. RIP and methylated RNA immunoprecipitation (MeRIP) confirmed two m7G sites on EPHA2 mRNA regulated by METTL1. Small interfering RNA (siRNA)-mediated METTL1 knockdown in EPHA2 mutants affected mRNA stability. Rescue experiments restored cell proliferation and AKT pathway gene expression. ConclusionMETTL1 methylates EPHA2 mRNA, enhancing its stability and expression, which activates the AKT signaling pathway and influences glioma cell proliferation. METTL1 could be a potential therapeutic target in glioma treatment.

cell biology↗

bioRxiv2026

GCBM-DCT-HV-Bio-NL-Grow-CHG-CSM-RHEC: A Unified Geometric, Biological, Causal, and Regenerative Framework for Mechanism-Aware Tissue and Connectome Modeling

Modern biological prediction problems increasingly require models that go beyond Euclidean feature regression and local graph smoothing. Tissue, cellular, and connectome systems are nonlinear, geometry-dependent, intervention-sensitive, history-dependent, and subject to regenerative or homeostatic constraints. We propose GCBM-DCT-HV-Bio-NL-Grow-CHG-CSM-RHEC, a unified model for mechanism-aware biological prediction. The model integrates geometric connectome dynamics, differentiable charted tissue geometry, Hamiltonian latent transport, nonlinear biological kinetics, nested latent memory, continual growth without overwriting, causal hypergraph structure, causal structure modeling, and regenerative homeostatic error correction. Unlike Euclidean baselines, which treat observations as flat vectors, and local graph baselines, which use neighborhood smoothing without mechanistic structure, the proposed model represents biological states (Trapnell 2015) as coupled geometric, dynamical, causal, and regenerative objects. We evaluate the model on four synthetic toy studies, Toy A-D, designed to reflect increasing biological complexity: local Euclidean structure, nonlinear mechano-chemical interaction, causal intervention response, and out-of-distribution regenerative shift. Compared with Euclidean and local graph baselines, the full model achieves the lowest mean squared error across all four toy studies. Relative to the Euclidean baseline, the full model reduces MSE by approximately 63.0%, 89.1%, 89.0%, and 90.9% on Toy A, Toy B, Toy C, and Toy D, respectively. These results support the value of integrating geometry, mechanism, causal structure, adaptive growth, and regenerative correction into a single predictive architecture (Figure 1). O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=142 SRC="FIGDIR/small/734320v1_fig1.gif" ALT="Figure 1"> View larger version (71K): org.highwire.dtl.DTLVardef@9e9eorg.highwire.dtl.DTLVardef@add398org.highwire.dtl.DTLVardef@1eb1f1org.highwire.dtl.DTLVardef@1346720_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFigure 1.C_FLOATNO GCBM-DCT-HV-Bio-NL-Grow-CHG-CSM-RHEC, a unified model for mechanism-aware biological prediction. C_FIG

bioinformatics↗

bioRxiv2026

CellTosg2Sequence: A Unified Text-Omics-Signaling-Graph Large Language Model for Single-Cell Analysis

In single-cell (sc)-based scientific discovery, text-formatted biomedical prior knowledge and signaling graphs are essential for annotating and interpreting numeric sc-omics data and for generating novel testable hypotheses. A major limitation of existing single-cell large language models (scLLMs) is that they rely on numeric expression data with gene names as the only textual signal, while comprehensive biomedical priors -- cellular localization, gene function, disease associations, and signaling interaction patterns -- remain absent from the model input. We introduce CellTosg2Sequence, a textual-prior- and signaling-graph-augmented cell-omics-sentence language model. A lightweight heterogeneous graph encoder maps a curated 62,507-node biomedical knowledge graph (KG) into compact virtual tokens that are prepended to each cell sentence, allowing the language model to condition on biological structure with minimal sequence-length overhead. We train CellTosg2Sequence with a three-stage objective: Stage I anchors the KG channel under autoregressive language-model pretraining, leveraging Qwen2.5-32Bs own language reasoning for rapid KG alignment; Stage II aligns labels via supervised fine-tuning with KG-anchored InfoNCE; Stage III applies Group Relative Policy Optimization (GRPO) with an ontology-hierarchy reward, enabling free-generation cell-type prediction that generalizes beyond the closed training vocabulary. Across multiple benchmarks and ablation experiments, CellTosg2Sequence outperforms strong baselines. All results are achieved with lightweight LoRA training and a single unified checkpoint. Data ethicsThis work uses publicly available single-cell datasets from the Human Cell Atlas (https://www.humancellatlas.org) and the Tahoe-100M consortium. All HCA constituent studies were collected under appropriate donor consent and institutional oversight as described in their original publications; we perform computational re-analysis only and introduce no new human subjects data. HCA data access follows the HCA Data Portal terms of use. No new patient data are collected in this study; no additional IRB approval is required for this secondary computational analysis.

bioinformatics↗

bioRxiv2026

A Tripartite Map of the Ventromedial Prefrontal Cortex

The ventromedial prefrontal cortex (VMPFC) has been repeatedly implicated in affect, valuation, and social cognition, yet how these diverse functions are organized within a single cortical territory has remained unresolved. Here, we integrate large-scale meta-analysis, individual-level task fMRI, artificial neural-network encoding models, and multimodal connectivity analyses to reveal the internal functional architecture of the human VMPFC. Across four complementary studies, we identify a robust tripartite organization along the anterior-posterior axis, comprising posterior affective, middle valuation, and anterior social functional motifs. Connectivity fingerprinting demonstrates that each motif is preferentially embedded within distinct large-scale brain networks, providing a mechanistic account of VMPFC functional specialization. This organization is reproducible at the level of individual subjects, generalizes to naturalistic stimuli, extends across development, and shows cross-species correspondence with non-human primates and multiple neurobiological markers. Together, these findings resolve a long-standing organizational question and establish a biologically grounded framework for interpreting VMPFC function.

neuroscience↗

bioRxiv2026

MacaSurfer: unified surface-volume mapping of the macaque brain across the lifespan

Macaque brain MRI is central to translational and comparative neuroscience, yet multi-site, longitudinal, and cross-species analyses are hindered by a lack of unified, automated structural processing tools. Existing pipelines, mostly adapted from human neuroimaging or restricted to fragmented steps, fail to provide robust surface-volume representations across heterogeneous acquisitions and developmental stages. Here we introduce MacaSurfer, a fully automated, containerized framework for unified surface-volume mapping of the macaque brain across the lifespan. MacaSurfer features components tailored for macaque anatomy: a tissue segmentation model, a tissue-guided bias-field correction method optimizing structural mapping from T1-weighted images alone, topology-aware surface reconstruction, and surface-aware volumetric registration. Validated on 1,346 imaging sessions from 965 macaques across 39 international sites (spanning 2 weeks to 23 years of age), MacaSurfer demonstrated exceptional anatomical consistency, test-retest precision, and robustness against image degradation. Leveraging MacaSurfer-derived morphometry, we established normative trajectories from 835 macaques, providing a standardized reference for downstream individualized deviation analysis. MacaSurfer is openly available with source code, containers, and pretrained models, offering a reproducible ecosystem to accelerate developmental, translational, and comparative neuroimaging.

neuroscience↗

bioRxiv2026

The emergent mental operation of the learning brain

How does the learning brain give rise to emergent mental operations that enable skill internalization and generalization? Using five-year longitudinal tracking of children acquiring abacus-based mental calculation, we reveal how sustained practice progressively reconfigures whole-brain states. Learning induces nonlinear transitions from stable baseline configurations through a transient exploratory phase toward restabilized associative brain networks. This hierarchical remodeling cascades from perceptual systems to action and default networks, mediated by adaptive neural variability and selective connectivity reweighting, with the salience-parietal memory network acting as a dynamic relay. These brain-state transitions accompany an ordered cognitive transfer cascade, emerging within arithmetic and extending to visuospatial and executive functions. Our findings are reproducible and provide a systems-level account of how ecological learning sculpts large-scale networks to enable complex skill acquisition.

neuroscience↗

bioRxiv2026

Integrating Histology with Spatial Molecular Programs Using a Multimodal Foundation Model

Histopathological assessment remains central to cancer diagnosis and stratification, yet its mechanistic interpretation remains limited without molecular context. To address this, we developed SQUALL, a multimodal foundation model integrating histology with spatial molecular programs. For pretraining, we assembled histMol, a large-scale corpus of 1.76 billion paired histology-spatial transcriptomics spots/bins across 33 tissues and 12 platforms from 3,446 tissue sections. Following pretraining, SQUALL enables transcriptome-wide virtual biomarker profiling, prognostically relevant spatial niches discovery, and integrative disease progression modeling. Leveraging its multimodal embeddings, SQUALL identifies niches associated with tertiary lymphoid structure (TLS) maturation and ovarian cancer relapse, reconstructs molecular trajectories of breast cancer invasion across 325,112 spots, and uncovers underlying transcriptional programs. Applied to whole-slide images from 898 patients, SQUALL outperforms existing pathology foundation models in outcome prediction while enabling interpretable risk stratification. Together, these results establish spatially aligned multimodal pretraining as a new paradigm for extending molecular insights into pathology images.

bioinformatics↗