AI News Archive: July 21, 2026 — Part 11
Sourced from 500+ daily AI sources, scored by relevance.
- The safety failures we are not instrumenting: a perspective on hidden safety-critical challenges in modern AI systems
Current AI safety discourse still focuses disproportionately on visible failures, including obvious harms, dramatic misuse, and hypothetical catastrophic scenarios. That focus is incomplete. In deployed systems, many of the most consequential failures are quieter: plausible rather than spectacular, ...
- Public perceptions of AI-driven decision-making in healthcare: A structural equation modeling approach
Artificial intelligence (AI) is increasingly integrated into healthcare to support diagnostics, decision-making, and administrative processes. However, the successful implementation of AI depends not only on technical performance but also on public perceptions of its helpfulness, riskiness, and fair...
- Mitigating Matthew Effect: Multi-Hypergraph Boosted Multi-Interest Self-Supervised Learning for Conversational Recommendation
The Matthew effect is a big challenge in Recommender Systems (RSs), where popular items tend to receive increasing attention, while less popular ones are often overlooked, perpetuating existing disparities. Although many existing methods attempt to mitigate Matthew effect in the static or quasi-stat...
- Caring Over Computing: An Ethical and Sociotechnical Perspective on Generative AI for Social Connectedness in Dementia Care
People with dementia in residential care often experience reduced social connectedness. Person-centered care approaches foreground meaningful social interactions to support well-being, but staff and time pressures increasingly constrain opportunities for such care. Generative AI (GenAI) technologies...
- Understanding ADHD Productivity in Construction Work: Toward AI-enabled VR Interventions
Attention-Deficit/Hyperactivity Disorder (ADHD) is identified as the most prevalent neurodivergent condition in the construction industry. While the construction industry may broaden employment opportunities, little is known about how ADHD traits shape workers' performance, sustained attention, and ...
- Data Leakage Prevention in Agentic Applications via Preemptive Hardening
Agentic systems integrate LLM driven planning with interfaces to external tools, making data leakage and tool misuse feasible via instruction/data boundary failures and prompt injection attacks. Enforcing required controls consistently is particularly challenging in workflows spanning many codebases...
- Cross-Agent Campaign Attribution: Linking Asynchronous Attacks Across LLM Agents
LLM-agent defenses are typically evaluated one session at a time. In deployment, however, attacks can be distributed across independent agents, teams, and runtimes, leaving each local guardrail with only a sparse fragment. We formalize cross-agent asynchronous campaign attribution: linking sessions ...
- When to Trust the Map: Confidence-Aware LLM Routing for Automotive CVE-to-ATM Mapping
Public CVE descriptions report the technical conditions and impact of vulnerabilities, whereas the Auto-ISAC Automotive Threat Matrix (ATM) expresses an adversary's tactics and techniques. Because the two representations are not directly aligned, incorrect automated mappings in safety-critical envir...
- CPInj: Uncovering Prompt Injection Risks in Textual Collaborative Prompt Optimization
Textual Collaborative Prompt Optimization (TCPO) extends Textgrad (Yuksekgonul et al., 2025) to a decentralized setting by allowing multiple clients to jointly improve prompts for large language models (LLMs) while keeping their data locally. Its reliance on free-form textual updating and aggregatio...
- ResearchArena: Evaluating Sabotage and Monitoring in Automated AI RD
As AI agents begin to automate AI R&D, we need ways to assess whether their outputs are safe to deploy, even when the agents themselves may be untrusted. AI control offers one such approach: rather than trusting the agent, it treats it as a potential adversary and uses a monitor to detect covert sab...
- GLID: Gated Local Intrinsic Dimension Repairs the Blind Spots of Face-Forgery Detectors
Fine-tuned foundation-model detectors dominate face-forgery benchmarks, yet they stay blind to generator families absent from training. We present GLID, a detector that repairs this blind spot with geometry instead of data. GLID treats the patch tokens of a single image as a sample from a manifold a...
- EmbeddedKittens: An Evaluation of Code Embeddings for Scratch
The trend of embedding source code for machine learning applications also enables new opportunities in learning analytics in programming education, but which code embedding approach is most suitable for learning analytics remains an open question. A common approach to embedding source code lies in t...
- From Collaboration to Regulation: Characterizing Governance Practice in Three Deep Learning Open Source Communities
Collaboration in Open Source Software (OSS) projects creates substantial coordination and quality-control challenges across diverse contributor bases. Projects address these challenges through documented governance rules, yet maintainers have limited systematic guidance on what rules to codify, when...
- TraceDev: A Traceability-Driven Multi-agent Framework for Requirement-to-Code Development
In modern software development, the rapid advancement of Large Language Models (LLMs) has made the end-to-end transformation of Natural Language Requirements (NLRs) into executable repository-level code increasingly feasible. However, existing approaches typically rely on simplified instructions (e....
- LLM-Based Invariant Testing for Software Functional Bugs
Manually writing unit tests to uncover functional bugs in software libraries is not only time-consuming but also requires a deep understanding of the intended semantics of the APIs. Heuristic-based test generation methods suffer from low usability because they cannot reason about program semantics o...
- Graph-Based Agentic AI with LangGraph: Workflow Pathways for Long-Running Stateful Business Processes
This paper is a practitioner guide to graph-based workflow pathways for long-running, stateful, multi-step generative AI systems in business processes. Rather than treating LangGraph, a low-level orchestration framework for stateful agents, as a model-quality benchmark target, we present three execu...
- Summary of DCASE 2026 Task 5: Audio-Dependent Question Answering
DCASE~2026 Task~5 introduces Audio-Dependent Question Answering (ADQA), which tests whether large audio-language models answer from the audio rather than from textual priors. An Audio-Dependency Filtering (ADF) pipeline combines silent-audio probing, per-option perplexity, a large language model (LL...
- PAGE-RAG: Evidence-Grounded Adaptive Graph Retrieval for Long-Document Question Answering
GraphRAG improves long-document question answering by introducing structured representations beyond conventional retrieval. However, automatically constructed graphs are inherently incomplete projections of source documents, and treating them as independent knowledge sources may lead to unreliable r...
- Sequential Learner Modeling Using Multi-Relational Graph Convolutional Networks
User modeling is a critical task in a variety of personalized systems. Recognizing their effectiveness in learning from graph-structured data, Graph Neural Networks (GNNs), particularly Graph Convolutional Networks (GCNs), are increasingly employed for user modeling. However, existing approaches typ...
- Beyond Noisy Signals: Dual-Level Denoising for Multi-modal Sequential Recommendation
Multi-modal Sequential Recommendation (SR) incorporates rich side information (e.g., textual and visual features) to enhance dynamic user preference modeling. However, existing frameworks inevitably suffer from a \textbf{Dual-Noise Dilemma}: (1) \textit{Feature-level redundancy} stemming from the se...
- Exposure-Based Reinforcement Learning to Rank
Reinforcement learning (RL) methods for learning-to-rank (LTR) can optimize (almost) any ranking goal, e.g., from precision or discounted cumulative gain to fairness-of-exposure or ranking distillation. However, standard RL is ineffective and computationally costly due to the enormous action space i...
- Spectral Biclustering-Driven Scalability for Post-Hoc Explainability in Recommender Systems
Explainability in recommender systems is essential for ensuring transparency, accountability, and trust, yet existing post-hoc methods often encounter severe scalability challenges. Observation-level deletion diagnostics offer a counterfactual way to analyze recommendations by retraining models afte...
- Answer-Reconstruction Search Density: Measuring the Query and Source Work Compressed by Conversational Answers
Conversational systems can collapse a visible sequence of web queries, result inspections, and source comparisons into a single synthesized answer. Existing retrieval metrics evaluate ranking, effort, or factual support, but they do not quantify the minimum conventional search work represented by a ...
- TSGR: Taobao Search Generative Retrieval
Generative retrieval (GR) has demonstrated strong promise for industrial e-commerce search by training a single autoregressive model to directly generate the Semantic IDs (SIDs) of target items. However, existing GR systems are primarily optimized for semantic matching and remain insensitive to item...
- An Epistemic Position-Based Click Model: From Interactions to Epistemic Distributions of Relevance and Bias
User interactions with rankings are affected by both items' relevances and display positions. Accordingly, click probabilities are often modeled as a product of relevance and position factors; and for improving recommendation and search, one needs to disentangle relevance from position bias. However...
- Topology-Aware Tokenization for Generative Recommendation
Generative recommendation reformulates sequential recommendation as an autoregressive generation task, yet a critical issue in this paradigm remains overlooked: topology distortion in item tokenization. In particular, we observe that the intrinsic adjacency relationships of items in the pretrained s...
- Extracellular Matrix Proteomic Signatures Associate with Disease-Free Survival in Later Events of Ductal Carcinoma In Situ or Invasive Breast Cancer
Background: Ductal carcinoma in situ (DCIS) is a noninvasive breast lesion with variable risk of progression to invasive breast cancer (IBC). Current transcription and cell marker investigations suggest ECM decreases in later events but are limited in details of ECM proteomic composition, including post-translational modifications. We investigated whether the extracellular matrix (ECM) proteome alters with later breast events of DCIS or IBC. Methods: ECM targeted mass spectrometry imaging and liquid chromatography tandem mass spectrometry (LC MS/MS) were applied to ten tissue microarrays from the Resource of Archival Human Breast Tissue cohort (RAHBT). Primary DCIS specimens (n=136) were analyzed in relation to later events of DCIS (n=40) or IBC(n=30), with a mean follow-up of 192.1 months 95% CI [179.1, 205.1]. Statistical modeling, survival analyses, and exploratory machine learning approaches were used to identify ECM peptide signatures associated with later events. Results: Distinct ECM peptide profiles were associated with later events of DCIS or IBC. Fifteen peptides derived from fibrillar collagens (COL1A1, COL1A2, COL3A1) and elastin, showed significantly reduced abundance in patients who developed IBC. Lower expression of specific collagen peptides associated with overall 19.9% 95% CI [17.92, 21.81] decreased disease-free survival for IBC. Lower expression of these peptides was significantly associated with reduced disease-free survival (age-adjusted hazard ratio [HR] = 2.45, 95% CI: 2.33, 2.57; P < 0.05). Patient-matched samples of primary DCIS, later DCIS, and later invasive breast cancer further demonstrated reduction in ECM peptide detection. Exploratory predictive modeling from patient-matched samples achieved high performance (AUROC >0.98, accuracy >93%) in distinguishing primary from later events. Following prior work in the RAHBT cohort, reduction of certain collagen peptides was also observed in primary DCIS samples from higher risk patient groups. Conclusions: ECM proteomic remodeling, particularly decreases of specific collagen domains, is strongly associated with later events of DCIS and IBC. These findings highlight ECM proteome as a critical regulator of breast cancer emergence with potential as a prognosticator of risk stratification to guide clinical management of DCIS.
- Microvascular pathology of proteotoxic endothelial signature characterizes Progressive Supranuclear Palsy
Cerebrovascular pathology is increasingly implicated in neurodegenerative diseases, yet its pathomechanistic contribution remains poorly defined. Building on prior evidence of dysregulated iron and oxygen homeostasis in early-affected brain regions of progressive supranuclear palsy (PSP), we hypothesized that brain microvascular alterations may play an etiological role in select neurodegenerative proteinopathies. First, we conducted a systematic neuropathological evaluation of 178 brains from the University Health Network Neurodegenerative Brain Collection, including Alzheimer's disease-related neuropathologic change (ADNC; n=30), Lewy body disease with high or intermediate ADNC (n=38) and low ADNC (n=16), multiple system atrophy (MSA; n=14), PSP (n=39), frontotemporal lobar degeneration with TDP-43 proteinopathy (FTLD-TDP; n=10), and controls (n=31). Arteriolosclerosis, microinfarction, and calcification were assessed in the basal ganglia and frontal cortex. Iron burden was correlated by quantification of Perl's staining in MSA and PSP, where vessel pathology was most severe. Single-nucleus RNA-sequencing (snRNA-seq) of frontal cortex tissue from control (n=5) and PSP (n=8) cases with varying arteriolosclerosis severity was performed to characterize the vascular transcriptome, with validation against an independent snRNA-seq evaluation of PSP (n= 11), Pick's disease (n=9), AD (n=10), and control (n=10) brains. Histological analysis revealed disease-specific involvement of microvascular pathology in neurodegenerative diseases, identifying PSP to demonstrate most prominent and widespread vessel wall thickening across regions examined. Regression analysis using demographic, APOE and MAPT genetic risk status, and neuropathological features of cases corroborated the distinct association with PSP pathology. Elevated iron load in early affected regions of MSA and PSP brains correlated with greater vessel wall thickening, suggesting a possible pathomechanistic relationship between the two disease physiologies. snRNA-seq analysis of vascular transcriptome identified robust upregulation of heat shock proteins and hypoxia-related genes in PSP endothelial cells and pericytes across both datasets. Importantly, we found the proteotoxic signature to be strongly associated with higher vessel scores in PSP cases, linking microvascular morphology to endothelial dysfunction. Our comprehensive neuropathological evaluation coupled with correlative snRNA-seq analysis establish PSP-specific arteriolar thickening associated with endothelial proteotoxic state as a candidate pathogenic mechanism. The cerebral arteriolar unit represents a compelling therapeutic target for disease modification in PSP.
- Plate-based ISD-SPE enables dual proteome-secretome concentration-response profiling of TLR signalling in iPSC-derived macrophages
Protein secretion represents a key functional output of cellular signalling, capturing dynamic responses to stimulation and pharmacological perturbation that shape immune behaviour. In macrophages, activation of Toll-like receptors (TLRs) drives tightly regulated secretion programmes that mediate inflammatory responses and provide a biologically meaningful readout of pathway activity. Whilst mass spectrometry (MS)-based secretomics enables unbiased profiling of these processes, broader application in drug discovery remains constrained by sample preparation workflows that limit scalability. Here, we describe a plate-based in-solution digestion and solid-phase extraction (ISD-SPE) workflow that enables 96-well processing of conditioned media for integrated proteome and secretome analysis from the same sample well. Benchmarking against a precipitation-based approach demonstrated comparable proteomic depth with improved quantitative reproducibility and robust performance across multiple plates. Coupled with dia-PASEF acquisition, this workflow enabled in-depth profiling of macrophage responses to TLR activation, resolving receptor-specific secretory programmes following TLR3, TLR4 and TLR7/8 activation. Extension of the approach to concentration-response studies enabled quantitative characterisation of pharmacological perturbation across intracellular and extracellular protein landscapes, revealing both shared and compartment-specific responses to TLR inhibition, as well as differences in apparent potency linked to secretion dynamics. Together, this workflow provides a scalable strategy for integrated analysis of intracellular signalling and downstream protein secretion, enabling systems-level characterisation of inflammatory responses and compound mechanisms of action.
- Timed STING Inhibition Mitigates Gastrointestinal GvHD While Preserving Graft-versus-Leukemia Activity After Allo-HSCT
Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is curative for hematological malignancies but limited by graft-versus-host disease (GvHD), in which donor T cells damage host tissues. Current prophylaxis broadly suppresses donor immunity, often compromising the beneficial graft-versus-leukemia (GvL) response, highlighting the need for strategies that uncouple GvHD from GvL. The cGAS-STING pathway can be strongly activated during conditioning-induced tissue damage and represents a potential therapeutic target. However, its context-dependent roles in inflammation and homeostasis have constrained its clinical translation. Here, we show that ''timed'' administration of the covalent STING inhibitor H151 before conditioning reduced GvHD-associated mortality without impairing GvL in murine allo-BMT models. Donor T cell activation and effector function were preserved indicating that the protective effect operates at the level of GvHD target tissues rather than through systemic immunosuppression. Timed STING inhibition protected the intestinal epithelium by limiting apoptosis, preserving intestinal stem cell function, and sustaining metabolic fitness during conditioning-induced injury, independently of type I interferon (IFN-I) signaling. In allo-HSCT patients, low intestinal STING expression is associated with reduced transplant-related mortality. Together, these findings identify timed STING inhibition as a tissue-protective prophylactic strategy that could be incorporated into existing conditioning regimens to enhance efficacy while minimizing toxicity.
- Increased expression of Cd74 and MHC II genes by aortic macrophages links atherosclerosis with aging in mice
Atherosclerosis is a chronic inflammatory condition of the arteries leading to myocardial infarction, ischemic stroke and peripheral arterial diseases. Atherosclerotic plaques, which obstruct blood flow contain functionally diverse macrophage populations including pro-atherogenic and atheroprotective subsets. Aging is a major risk factor of atherosclerosis, but the mechanism underlying aging associated risk of atherosclerosis is unclear. Here, using integrated single-cell RNA sequencing data analysis, we demonstrate that specific monocyte and macrophage subsets are enriched in atherosclerotic plaques and the aging aorta of mice. We also show that Cd74 and MHCII genes such as H2-Aa, H2-Ab1, H2-Eb1, and H2-DMb1 are consistently upregulated in the monocytes and macrophages from atherosclerotic plaques, the aging aorta, and the aging bone marrow of mice. Our experimental data also show increased expression of CD74 surface protein by aortic macrophages and bone marrow monocytes from aged mice. In addition, we show increased CD74 expression by RAW264.7 mouse macrophages and THP1 human monocytes following ox-LDL stimulation in vitro. Finally, we demonstrate that monocytes from aged mice adhere more to aortic endothelial cells in co-cultures. Thus, increased monocyte adhesion to endothelial cells may explain the enhanced proportion of specific monocyte and macrophage subsets in the aging aorta, inducing a pre-atherosclerotic condition.
- Chemogenetic Inhibition of the Ventrolateral Orbitofrontal Cortex Disrupts Prediction Error and Salience-Guided Memory-Updating at the Hippocampal Engram Level
Rationale: Adaptive behavior requires revising existing memories when outcomes deviate from expectations. The orbitofrontal cortex (OFC) is implicated in representing outcome expectancies, but its role in memory-updating, beyond value-based learning, remains unclear. Objectives: Here, we tested whether the ventrolateral OFC (VLO) contributes to hippocampal-dependent memory-updating in the Objects in Updated Locations (OUL) task, where novelty-driven exploration signals successful updating of spatial information. Methods: Using inhibitory DREADDs, in male and female (Swiss Webster and C57BL/6) mice, we suppressed VLO activity during the Updating session, when new object-location information was presented. Results: Control mice preferentially explored the newly updated location whereas VLO-inhibited mice failed to show this preference during Updating and again at Test. Critically, VLO inhibition when no updating demand was present did not impair novelty preference or memory for original spatial configurations at Test, indicating preserved novelty detection and retrieval abilities. Conclusions: These results demonstrate that VLO activity is selectively required when novelty must be interpreted relative to prior experience, further implicating this region in evaluating change rather than detecting it. To determine whether these impairments reflected a deficit in memory-updating beyond altered salience attribution, we tagged neuronal ensembles in the dorsal dentate gyrus recruited during Updating, followed by quantification of overlap between Updating-tagged cells and ensembles reactivated at Test. We found that behavioral deficits were correlated with reduced ensemble overlap, indicating that VLO activity supports memory-updating at the hippocampal engram level. These findings extend OFC models beyond reward contexts and identify a cortical contribution to predictive spatial memory processes.
- Microglia from Friedreich Ataxia patients are intrinsically primed for neuroinflammation
Friedreich Ataxia (FRDA) is an autosomal recessive neurodegenerative disorder characterized by progressive loss of cerebellar and proprioceptive neurons that control movement and coordination. In most patients, FRDA is caused by homozygous GAA trinucleotide repeat expansions in the first intron of the frataxin (FXN) gene, resulting in reduced expression of frataxin, a mitochondrial protein essential for biogenesis of iron-sulfur clusters and mitochondrial function. Although recent therapeutic advances have provided modest clinical benefit, effective disease-modifying treatments remain lacking. Increasing evidence indicates that microglial cell dysfunction contributes to FRDA pathogenesis, highlighting these cells as potential therapeutic targets. However, the molecular mechanisms underlying FXN-deficient microglial dysfunction remain poorly understood. Here, we show that microglia generated from FRDA patient-derived iPSCs exhibit a cell-autonomous pro-inflammatory phenotype in the absence of exogenous inflammatory stimuli. This phenotype is characterized by coordinated activation of immune transcriptional programs, dysregulated secretion of neuroinflammatory proteins, impaired autophagy-lysosomal function, and activation of inflammasomes pathways involving NLRP2 and NLRP3. These findings demonstrate that FXN deficiency is sufficient to induce intrinsic microglial activation and identify molecular pathways that may represent attractive targets for future FRDA therapies.
- Transglutaminase 2 Deletion Enhances Astrocyte-to-Neuron Metabolic Support and Attenuates Subacute Pathology Following Repetitive Mild Traumatic Brain Injury
Mild traumatic brain injury (mTBI) is the most common form of central nervous system (CNS) injury and is often characterized by persistent neuroinflammation, metabolic dysregulation, and oxidative stress. Repetitive injuries compound these pathologies and lead to multifocal axonal injuries and long-term functional deficits. Despite the prevalence of mTBIs, the cellular mechanisms that facilitate or prevent recovery following injury remain poorly defined. Here, we extend our previous work on the role of the protein transglutaminase 2 (TG2) in CNS injury and we hypothesize that transcriptional regulation by TG2 restricts metabolic versatility in astrocytes following TBI, thereby impairing neuronal energetic support and worsening pathological outcomes. We utilized an established weight-drop model of repetitive mTBI followed by multi-parametric analysis of TBI pathology in complete TG2 knockout (TG2-/-) and wild type mice. At 28 days post-injury, TG2-/- mice showed marked attenuation of TBI pathology, compared to wild type mice, in vulnerable white matter and default mode network (DMN) regions, as assessed by diffusion magnetic resonance imaging (MRI), resting-state functional MRI, and immunohistochemistry. Integrated epigenomic, proteomic, and metabolomic profiling of cortical astrocytes isolated 28 days after injury revealed a pronounced metabolic restriction in wild type astrocytes which was remarkably attenuated in the TG2-/- mice. This rescue was associated with a de-repression of gene networks involved in glutamate recycling, lipid metabolism, and metabolic homeostasis. Together, these studies provide novel mechanistic insights into the metabolic dysregulation that characterizes persistent TBI pathology, and establish a foundation for evaluating TG2 as a therapeutic target for TBI.
- Progressive neuronal network reorganisation in glioblastoma drives pathological activity in vitro
Glioblastoma (GBM) is the most aggressive primary brain tumour and is frequently accompanied by severe neurological symptoms, including epilepsy and cognitive impairment. Neurological symptoms often persist after surgical resection, indicating that GBM induces durable and self-sustaining changes in the surrounding neuronal networks. However, the mechanisms by which GBM reshapes network structure and function in the tumour periphery remain poorly understood. We present a compartmentalised in vitro platform enabling long-term co-culture of iPSC-derived neurons and primary GBM cells to investigate these changes. Placed on high-density microelectrode arrays, the platform permits longitudinal electrophysiological recordings at single-neuron resolution. Using effective network inference, we find that GBM drives a reproducible structural progression: first toward a hyperconnected, hub-dominated architecture, then a collapse of community structure accompanied by a widespread neuron loss. This evolving structure shapes population dynamics, constraining features such as network burst rate and instantaneous synchrony. The reorganisation also carries computational consequences: signal propagation becomes progressively redundant and synergistic rather than unique. As a result, neurons lose the capacity to encode distinct input combinations independently, and the repertoire of accessible network states contracts. Together, these findings reframe GBM as a driver of neuronal network reorganisation rather than uniform hyperexcitability, and establish a compartmentalised, single-neuron-resolution platform for the longitudinal observation, dissection, and ultimately targeting of the network processes that underlie disease progression.
- Elicitation of stem-directed antibodies in rhesus macaques by a conventional hemagglutinin immunogen
Because they can bind many strains of influenza, antibodies targeting the hemagglutinin (HA) stem have been attractive targets for vaccine development. Many monoclonal antibodies (mAbs) directed at the HA stem have been isolated from humans, and these mAbs have mediated broad protection in animal models. We describe here HA stem-directed mAbs isolated from rhesus macaques immunized with an "ordinary" H1 HA trimer. All immunized rhesus macaques developed high serum titers with broad reactivity to diverse H1N1 and H5N1 viruses, and 7 isolated mAbs strongly blocked canonical stem antibody CR6261 binding to H1. MAb DH726.1 robustly protected mice from lethal challenge with H1N1 and H5N1 viruses, and cryo-EM showed the binding footprint overlapped that of some human mAbs. These findings suggest that vaccination with the standard, trimeric HA immunogens may be sufficient to elicit stem antibodies at titers adequate to protect against zoonotic H5N1 influenza.
- Short-term dietary change rapidly remodels microbial community assemblages and reprogrammed systemic immune phenotypes
Diet is a major determinant of the gut microbiome and immune homeostasis, yet the extent to which short-term dietary interventions can remodel established microbial communities and reprogramme immune phenotypes following long-term western diet consumption remains poorly understood. Here, we investigated temporal dynamics of the gut microbiome, microbial metabolites, intestinal barrier function and local and systemic immune responses following diet switching. Mice were fed either a standard chow or a western diet for 8 weeks before remaining on these diets or switching to the alternate diet for 2 or 4 weeks. Long term consumption of chow and western diets resulted in distinct gut microbial communities and differences in intestinal permeability. Diet switching rapidly remodelled microbial community structure within two weeks, with substantial bidirectional changes in community composition. Despite these changes, the relative abundance of several taxa remained influenced by prior dietary exposure. In contrast, faecal SCFA profiles remained largely associated with long-term diet, indicating that microbial metabolic outputs were altered more slowly than microbial community composition. Mass cytometry revealed progressive remodelling of local (MLN) and systemic (PBMC and spleen) immune responses following dietary switching. Activation-associated immune phenotypes, including Ki67+ and PD-1+ B and T cells, inflammatory monocytes and ROR{gamma}t+ regulatory T cells, rapidly responded to diet switching, whereas overall B cells, regulatory T cells and effector memory T cells retained signatures of long-term dietary exposure. Together, these findings demonstrate distinct temporal dynamics across the diet-microbiome-immune axis, whereby gut microbial composition and immune activation states remain highly plastic, while microbial metabolic outputs and several memory and regulatory immune phenotypes exhibit persistent dietary imprinting. These findings highlight the potential utility of short-term dietary interventions to modulate host-microbiome interactions and immune homeostasis.
- Stable Network Homeostasis during Multi-Level Postnatal Maturation of the Mouse Tuberoinfundibular Dopamine-Prolactin Axis
The hypothalamus orchestrates endocrine function via specialized neuronal populations that interface with the pituitary gland. While postnatal circuit refinement is a hallmark of most neural systems, its contribution to hypothalamic neuroendocrine networks remains elusive. Among these populations, tuberoinfundibular dopamine (TIDA) neurons of the arcuate nucleus are the primary source of tonic inhibition of prolactin (Prl), a hormone essential for reproduction, parental physiology, and behavior. While Prl levels surge during early mouse postnatal life, it remains unclear whether the TIDA system is fully developed at birth or undergoes functional maturation. Here, we combined immunofluorescence, slice electrophysiology, and Ca2+ imaging to determine the development of TIDA neurons in mice during the first three postnatal weeks. Expression of dopaminergic markers was sparse at birth but rose substantially after the first week, followed by the onset of median eminence innervation by TIDA axons. Moreover, TIDA firing rate and oscillation frequency progressively increased with age, with action potential properties and rhythmicity maturing in tandem. Strikingly, local network parameters remained stable despite ongoing changes in single-cell properties, as did excitation/inhibition (E/I) balance. These neuronal adaptations paralleled a significant rise in circulating Prl levels. Together, our results delineate a multi-level developmental program within the TIDA system, encompassing molecular, electrophysiological, and endocrine changes. This maturation likely underlies the emergence of functional hypothalamic control over Prl secretion in early life and points to a coordinated controller-effector co-development. Our findings highlight a critical window during which TIDA neuron plasticity may influence long-term neuroendocrine function and reproductive behavior.
- Geometry-based dynamics of the postsynaptic density explain protein capture by an actin-spine-geometry-dependent synaptic tag
The synaptic tagging and capture (STC) hypothesis explains how early-phase plasticity is converted into its late phase through the coincidence of synaptic tagging and plasticity-related protein (PRP) availability. Yet the biophysical basis of this process remains poorly understood. Based on the hypothesis that the interaction of actin and spine geometry implement the synaptic tag, we here investigate the associated PRP capture mechanism. We propose that capture is implemented by PSD remodelling which is gated by local membrane curvature at the PSD periphery. Using computational modelling, we show that curvature variations around the PSD that arise from long-term potentiation (LTP) inducing stimuli indeed enable a PSD growth, reproducing late-phase potentiation and the maintenance of structural LTP. We further explore how the timing of PRP availability relative to tag formation and the initial spine size determine the extent of PSD enlargement, yielding outcomes consistent with experimental findings. Hence, our results support a structural interpretation of synaptic tagging and capture in which a transient, actin-driven geometric state of the spine encodes the tag, and curvature-mediated PRP recruitment stabilises synaptic changes, and thus render spine geometry as a key biophysical regulator of memory consolidation.
- Synaptic Development of Fine Spatial Scale Organization of Neuronal Orientation Tuning in Mouse Primary Visual Cortex
Primary sensory cortices often organize neurons with similar stimulus preference into spatially functional maps. Recent work in mouse primary visual cortex (V1) has established that neuronal tuning to the orientation of visual grating stimuli is organized into `micro-clusters', where physically close neuron pairs (~ 20$ um) share highly similar orientation preferences, but the organization is unstructured beyond this narrow range. This fine-scale organization is seemingly at odds with the underlying intracortical circuitry in mouse V1 whose spatial extent is an order of magnitude broader (100 ~ 200 um). In this study, we explore an activity-dependent synaptic plasticity model of spatially structured thalamo-cortical connectivity. We develop theory under asymptotic conditions specific for mouse V1, and derive concrete circuit conditions under which `micro-clusters' naturally develop. In particular, the recurrent interaction among V1 neurons requires an additional component over a `micro'-spatial scale, while the spatial profiles of balanced excitation and inhibition support an effective `micro'-scale interaction. Together, our results provide a developmental mechanism and analytical framework linking thalamo-cortical development, recurrent circuit structure, and the emergence of functional organization in primary visual cortex.
- HDAC3 inhibitor RGFP966 acts on the NF-kB pathway and enhances memory persistence in a biphasic manner
Nearly five decades ago, it was first observed that long-term memory consolidation requires waves of transcriptional activity and protein synthesis, with the first two waves occurring within hours after learning. While numerous studies have examined the effects of protein synthesis inhibitors, the contribution of epigenetic mechanisms in these waves remains poorly understood. Here, we aimed to determine the role of HDAC3, a key modulator of memory, in these two phases of gene expression, as well as its functional link with the transcription factor NF-{kappa}B, one of its deacetylation targets, and a critical player in memory formation. Pharmacological inhibition of HDAC3 with RGFP966, either immediately or 6 hours after training, enhanced memory persistence in the NOR task in mice. Conversely, inhibition of NF-kB with BAY 11-7082 impaired memory at the same time points. Moreover, RGFP966 injection increased the nuclear proportion of NF-{kappa}B in the CA1 region of the hippocampus, suggesting a functional link between HDAC3 activity and NF-{kappa}B nuclear translocation. To our knowledge, this study provides the first in vivo evidence of this relationship during memory consolidation, extending previous findings from cell culture and electrophysiological studies in brain slices.
- COSMOS: A FAIR-aligned infrastructure for clinical trial data validation, warehousing, and interactive discovery
The Clinical Omics System for Metadata and Outcome Storage (COSMOS) is an open-source, FAIR- and GCP aligned clinical trial unit (CTU) infrastructure designed to streamline the transition of academic clinical and multi-omic trial datasets into curated, analysis-ready repositories within Secure Data Environments (SDEs). By integrating an automated Data Quality and Data Validation (DQ&DV) "Trust Layer" with a relational Structured Query Language (SQL) schema, COSMOS enables programmatic and interactive data access via R Shiny applications. Dynamic integration of omics and clinical data is achieved through analytical data structures (e.g. ExpressionSet objects) linked via relational database identifiers. This lowers technical barriers for researchers and promotes governed data reuse and secondary discovery.
- GeneAutomate: A Browser-Based, Integer-Indexed Platform for Dual-Gene-List Functional Annotation and Interactive Network Visualization
Comparative interpretation of two gene lists, for example, two treatment arms, two tissues, or a discovery and a validation cohort, is a routine task in functional genomics. While several tools offer dual-list comparison (e.g., EnrichmentMap, RRHO packages), they typically require local software installation, R/Bioconductor, or manual reconciliation of separate single-list outputs. Most widely used web-based enrichment tools (DAVID, g:Profiler, Enrichr, ShinyGO, WebGestalt) are built around the analysis of a single gene list at a time, and those that support comparison often lack interactive, publication-ready visualization or depend on server-side query latency. Here we present GeneAutomate, a browser-based tool purpose-built for side-by-side comparison of two gene lists. GeneAutomate performs Over-Representation Analysis (ORA) against Gene Ontology (GO) and Reactome using an exact hypergeometric test with Benjamini-Hochberg false discovery rate correction, and Gene Set Enrichment Analysis (GSEA) when ranked (log2 fold-change) input is supplied, alongside Protein-Protein Interaction (PPI) subgraph extraction from BioGRID physical interactions. All reference data (Gene Ontology, Reactome, BioGRID, and NCBI/Ensembl identifier cross-references) are pre-compiled offline into a single integer-indexed database of approximately 32 MB for Homo sapiens, in which every gene identifier Ensembl ID, Entrez ID, official symbol, or alias is resolved to one canonical integer prior to any user query. This design removes live database round-trips from the runtime path, enabling fast, at-your-desk enrichment without installation or a server-side per-query bottleneck. The tool renders thirteen interactive, D3.js- and Cytoscape.js-based comparative visualizations, including a Rank-Rank Hypergeometric Overlap (RRHO) heatmap, a GO-slim "Radar/Spider" functional fingerprint, and chord/edge-bundled cross-talk diagrams that are, to our knowledge, not offered as an integrated set by any existing academic or commercial ORA/GSEA platform. GeneAutomate is an unfunded, individual student project developed with feedback from a professor, and is in its final stage of development. It requires no installation or login. We describe the tool's architecture, statistical methods, and comparative feature set relative to established academic tools (DAVID, ShinyGO, g:Profiler, Enrichr, WebGestalt, STRING, PANTHER, GeneMANIA, Cytoscape, clusterProfiler, GSEA, Metascape) and commercial platforms (IPA, MetaCore, Pathway Studio, iPathwayGuide, Partek Pathway), and we state candidly the current version's limitations, which are planned to be the added in next version: single-species (human-only) coverage, no upstream regulator analysis, and comparison currently limited to two (occasionally three) concurrent lists. GeneAutomate is available at https://geneautomate.tech/.
- Estimating trial-wise modulation of functional connectivity using event-related fMRI
Understanding the neural basis of human cognition requires measuring not only localized brain activity but also how functional interactions between brain regions change in response to different cognitive demands. Event-related fMRI is an efficient design for linking trial-wise behavioral and computational variables to brain activity, but comparable methods for examining their effects on functional connectivity remain limited. Here, we develop beta-PPI (beta-series psychophysiological interaction), a method that leverages single-trial response estimates from even-related fMRI to quantify how trial-wise variables modulate functional connectivity. This task-based functional connectivity method provides a flexible approach for studying functional connectivity for event-related fMRI designs. We evaluated beta-PPI using comprehensive simulations across several experimental conditions and signal qualities. Beta-PPI can sensitively detect ground-truth effects and exhibited good parameter recovery. Compared with generalized psychophysiological interaction, beta-PPI achieved comparable performance across most conditions while demonstrating improved statistical power under lower signal-to-noise conditions. We further validated beta-PPI using empirical event-related fMRI data. Distinct trial-wise cognitive variables selectively modulated functional connectivity during their corresponding trial epochs, demonstrating the temporal specificity and flexibility of the approach. By testing how trial-wise variables modulate functional connectivity, beta-PPI extends task-based connectivity analysis to model-based fMRI and provides a common single-trial framework that could facilitate the integration of connectivity, activation, and representational analyses in event38 related fMRI.
- Co-option of CENP-A for activity-induced neuronal plasticity
Neuronal activation drives activity-dependent gene expression that underlies experience-associated synaptic modifications, learning and memory. Here we show that the histone variant CENP-A, best known for specifying centromere identity, is dynamically regulated by synaptic activity at both the RNA and protein levels in postmitotic neurons. Neuronal activation increases a non-centromeric nuclear pool of CENP-A while leaving centromeric CENP-A levels unchanged. Downregulation of CENP-A selectively reduces the activity-associated non-centromeric pool, impairs activity-dependent induction of immediate-early genes such as FOS and ARC, and disrupts hippocampus-dependent learning and memory. Furthermore, we find that activity-dependent neuronal responses in human embryonic stem cell-derived forebrain organoids similarly require CENP-A. Our results reveal a mitosis-independent, conserved role of CENP-A for driving plasticity in mammalian neurons.
- Slow stress-load accumulation dominates BDNF-dependent gain in a ten-state computational model of stress biochemistry
Background. Acute stress responses are often reversible, whereas sustained stress can produce coordinated disruption across endocrine, metabolic, inflammatory, antioxidant, and neuroplastic pathways. The Tiered Stress Biochemistry Model (TSBM) is a hypothesis-generating ten-state ordinary differential-equation framework linking a stylized cortisol signal to noradrenergic drive, vitamin C, a phenomenological aldosterone/renin-angiotensin drive, magnesium, normalized BDNF-related and Nrf2-related states, inflammation, and tryptophan-kynurenine metabolism. Methods. Five prespecified scenarios (normal, acute, chronic, depression-like, and low-cortisol PTSD-like) were simulated for 168 hours. Analyses included local stability, output-specific sensitivity screening, structural ablation, a 400-draw Latin-hypercube scan over independent plus/minus 20% parameter ranges, uncertainty distributions for threshold-crossing times, a success-conditioned parameter-trade-off screen, and a wider scan in which hypothesized BDNF-feedback gain magnitudes varied log-uniformly from 0.1 to 10 times nominal. Parameters were classified as literature-derived, literature-constrained/model-defined, calibrated, or hypothesized. Results. Under the specified forcing assumptions, sustained stress produced coordinated changes across several pathways. In the depression-like scenario, removing slow stress-load accumulation increased day-7 BDNF-related activity from 47.2% to 82.6%, reduced inflammation from 14.63 to 2.15 arbitrary units, and lowered KYN/TRP from 0.173 to 0.057. Removing BDNF-dependent gain increased BDNF only to 49.1% and delayed KYN/TRP crossing by 2.5 hours. The stricter multi-output conclusion was retained in 91% of uncertainty draws. Median crossing times retained the nominal sequence, but the complete four-event order occurred in only 43% of all draws. Conclusions. Within this reduced model, a shared slow stress-load process coordinates the high-exposure state, whereas BDNF-dependent feedback acts mainly as an amplifier. The model generates an experimentally testable staging hypothesis: under sustained high-exposure forcing, magnesium changes may precede later BDNF-related and KYN/TRP changes. Longitudinal studies are required to determine whether this sequence occurs biologically, whether it is reversible, and whether it has clinical relevance. The simulations are not diagnostic or treatment recommendations.
- Cross-database validation reveals distinct layers of transportability in ICU delirium prediction
External validation of clinical AI emphasizes discrimination, although deployment requires the endpoint, probability estimates and operating policy to transport. Here we show that these layers diverged in retrospective bidirectional evaluation of five model families across eICU and MIMIC-IV. Coarse-label AUROC fell from 0.87-0.92 internally to 0.66-0.83 during source-only transfer. For assessment-conditioned repeated monitoring of persistence or recurrence, external AUROC reached 0.76-0.94, but removing assessment history reduced it by 0.16-0.32; broader features did not help consistently. Transported scores concentrated future-positive ICU stays 2.4-6.9-fold in the top risk decile. Development-selected cutoffs alerted 0.3-2.0% of prediction rows and captured 9.2-11.0% of future-positive rows; after deduplication, 4.9-12.2% of stays were alerted, capturing 43.9-49.4% of future-positive stays. Thus, ranking can persist while probability and policy transport remain site dependent. Layered validation is a prerequisite for prospective evaluation, not evidence of clinical benefit.
- Thymus Involution across the Human Menstrual Cycle
Importance: Thymic involution in humans has traditionally been conceptualized as a linear, age-associated process. Animal studies suggest transient hormone-dependent fluctuations in thymus size during reproductive transitions, including across the estrous cycle, but translational evidence in humans remains limited. Objective: To determine whether thymus size changes across the menstrual cycle in healthy women and whether these changes are associated with fluctuations in estradiol and progesterone levels. Design: Observational repeated-measures study using quantitative computed tomography (qCT). Data were collected as part of an imaging study examining menstrual cycle-related physiological variation. Participants underwent assessments during menses and the early luteal phase within the same menstrual cycle. Mixed-effects models with robust variance estimation evaluated associations between cycle phase, hormonal contraception, reproductive hormone levels, and thymus size. Thymus segmentation was independently conducted by 2 blinded raters. Setting: Single-center university-based imaging study conducted at the University of Iowa. Participants: Thirty-one non-smoking women with regular menstrual cycles were included, of whom n = 16 used oral hormonal birth control and n = 15 did not use hormonal contraception. Exclusion criteria included pregnancy, breastfeeding, postmenopausal status, diabetes, body mass index greater than 30 kg/m2, hysterectomy, or use of long-term noncyclic hormonal contraception. Participants tracked menstrual cycles using temperature monitoring and ovulation kits before completing visits during menses and the early luteal phase. All participants provided informed consent before study participation. Main Outcomes and Measures: Primary outcome was thymus size measured in mm3 using ultra-low-dose qCT imaging. Secondary outcomes included serum estradiol and progesterone levels. Results: Estradiol levels increased from menses to the early luteal phase independent of hormonal contraceptive status. Progesterone levels were significantly lower among women using hormonal birth control. Thymus size differed significantly by hormonal contraceptive group, with larger thymus volumes observed among women not using hormonal birth control. Among women not using hormonal contraception, 66.7% demonstrated thymic involution from menses to the early luteal phase, compared with 37.5% of women using hormonal contraception. Estradiol and progesterone significantly interacted in predicting thymus size. Conclusions and Relevance: These findings provide first-in-human evidence suggesting that thymus involution demonstrates short-term dynamics across the menstrual cycle and may be influenced by reproductive hormones and hormonal contraception.
- Composite Artificial Intelligence-Enabled Electrocardiogram for Detection and Prediction of Structural Heart Disease
Background Structural heart disease (SHD) drives heart failure and cardiovascular mortality but remains underdiagnosed, and echocardiography is limited as a population-level screening tool. Objectives We evaluated whether a composite artificial intelligence-enabled electrocardiogram (AI-ECG), combining independently developed models for left ventricular systolic (LVSD) and diastolic dysfunction (LVDD), identifies prevalent and predicts incident SHD across diverse populations. Methods In this multinational cohort study, detection was assessed cross-sectionally in a Korean clinical cohort (Incheon Sejong Hospital) and a US dataset (Columbia University Irving Medical Center), and incident risk was assessed in the Korean cohort and the UK Biobank among individuals without baseline SHD or heart failure. Adults with paired ECG and echocardiography were analyzed for detection, with the composite defined as positive on either model. SHD comprised reduced left ventricular ejection fraction, moderate or severe valvular disease, left ventricular hypertrophy, or pulmonary hypertension. Detection was assessed by sensitivity and specificity, and incident risk by Cox models and the C statistic. Results Among 46,082 and 36,286 participants in the two detection cohorts, the composite detected SHD with sensitivity of 71.8% and 76.1% and specificity of 88.3% and 70.1%, with positivity across all phenotypes. Among at-risk individuals, composite positivity was associated with incident SHD (hazard ratios, 3.75 and 2.75), with C statistics of 0.69 to 0.78. Conclusions A composite AI-ECG identified prevalent and predicted incident SHD across multinational cohorts, capturing signals beyond its training targets and supporting its potential as a scalable cardiovascular screening tool; whether ECG-based risk stratification improves outcomes requires prospective evaluation.
- Encoding Discordance in the Alzheimer's Disease A/T/N Framework
INTRODUCTION: The biomarker-based amyloid/ tau/ neurodegeneration (A/T/N) framework has become a popular staging method for Alzheimer's disease (AD) research. Previous studies use the framework either as a rule-based or data-driven approach but typically sacrifice either adaptivity or interpretability. METHODS: We present an interpretable, hybrid method, called Neurosymodal Data Fusion, for predicting incident AD in the ADNI dataset. Specifically, we encode the A/T/N framework as a logic program, where the input biomarker features are extracted by one or more neural networks. RESULTS: Our pipeline predicted four-year incident AD with a sensitivity of up to 0.84. Additionally, our models learned scores for each A/T/N profile, denoting relative importances to model predictions. These scores also indicated that empirically-derived cut-off values for the A and T criteria might be uninformative for the ADNI data. DISCUSSION: Our pipeline provides a novel way to use the A/T/N framework that could potentially improve early AD screening years before clinical manifestations.