2026-07-24 每日论文
| 来源 | 独立页面 |
|---|---|
| arXiv | arXiv |
| bioRxiv | bioRxiv |
| medRxiv | medRxiv |
arXiv
| 标题 | 作者 | 发布日期 | PDF链接 | 摘要 |
|---|---|---|---|---|
| 3D-Aware VLMs with Implicit and Explicit Geometries | Wenhao Li | 2026-07-23 | Despite rapid progress, most existing vision-language models (VLMs) built from 2D visual inputs often struggle when handling various 3D tasks that require fine-grained spatial understanding and reasoning. To bridge this gap, we present VLM-IE3D, a unified framework that enhances the 3D spatial awareness of VLMs by equipping them with both implicit and explicit 3D geometries learned from RGB videos. Our VLM-IE3D introduces Implicit Geometry Tokens (IGTs) that capture high-level geometric priors from input videos, as well as complementary Explicit Geometry Tokens (EGTs) that encode detailed geometric structures from reconstructed 3D attributes. On top of that, VLM-IE3D comes with a 3D-aware adapter that effectively fuses the two types of geometric representations with 2D visual cues. This RGB-only design injects strong 3D inductive biases for fine-grained spatial understanding and reasoning without requiring any additional 3D inputs. Extensive experiments show that VLM-IE3D achieves superior performance consistently across various 3D tasks including 3D video detection, 3D visual grounding, 3D dense captioning, and spatial reasoning. Code and models are available at https://github.com/Vegetebird/VLM-IE3D. | |
| Streaming Multi-Agent Autoregressive Diffusion Model with World State Registers | Sicheng Mo | 2026-07-23 | Multi-agent interactive world models should not only generate consistent observations, but also maintain world states that persist across agents and evolve across views. Existing autoregressive video diffusion pipelines carry forward observation history as conditioning context, which makes shared state difficult to maintain in multi-agent and multi-view settings. We present WorldWeaver (W^2), a streaming multi-agent video diffusion model that augments rollout with cross-agent world state registers: learnable tokens that store shared world information, track individual agent status, and are dynamically updated after each generated chunk. We ground these registers with supervision signals spanning individual agent status, global state views including bird's-eye views, and scene text. We further improve the architecture with a Mixture-of-Transformers design that uses separate weights for world state modeling and visual frame modeling. Extensive experiments in two-agent Minecraft video generation show that explicit world-state modeling improves logical consistency and generation quality. | |
| Unified Video Dense Prediction from Disjoint Data | Yihong Sun | 2026-07-23 | Scene understanding requires simultaneous prediction about geometry, appearance, and semantics. However, existing task-specific annotations are fragmented across incompatible, domain-specific datasets. Current unified systems circumvent this by restricting training to fully co-annotated data, or by incurring the large computational cost of pseudo-labeling. To mitigate this, we introduce UniD, a unified video model that jointly predicts eight dense scene properties-depth, surface normals, semantic segmentation, boundaries, human parts, albedo, shading, and materials-all learned from disjoint, domain-specific datasets. We propose a simple yet effective distillation step in which per-task experts supervise a unified backbone through lightweight task projectors, eliminating the need for annotation overlap or pseudo-labeling. Our key insight is that the strong visual priors of a pretrained diffusion model are sufficient to bridge the domain gaps introduced by disjoint training sources, enabling robust generalization to scene-task combinations never seen during training. UniD achieves competitive performance against per-task specialists and multi-task baselines, with strong generalization to out-of-distribution scenarios and enhanced temporal and cross-task consistency. Code and video results are available at https://unid-video.github.io/. | |
| Inference-Time Scaling of Diffusion Models via Progressive Seed Pruning | Rogerio Guimaraes | 2026-07-23 | Diffusion and flow-matching models dominate conditional image generation, yet inference-time scaling for these models is far less developed than for autoregressive language models. Because final quality is highly sensitive to the initial noise seed, many approaches spend extra compute on seed search or resampling under a black-box reward, but typically maintaining a constant memory footprint throughout inference. We show that relaxing this constraint enables an underexplored inference-time scaling axis: by front-loading exploration, evaluating many seeds early, and pruning aggressively, we can use a fixed compute budget more effectively. \emph{Progressive Seed Pruning} (\PSP) scores intermediate denoised estimates and progressively narrows the candidate set so that only promising trajectories are fully denoised, while keeping the total number of model evaluations fixed. Across diffusion and flow-matching backbones, \PSP \ consistently improves reward-guided selection and achieves higher GenEval scores (automated) and better human evaluation on prompt-alignment than best-of-$N$, importance-sampling, and tree-search baselines at matched compute. Project page: https://www.vision.caltech.edu/psp. Code: https://github.com/rogerioagjr/psp. | |
| Expanding Flow Maps | Sophia Tang | 2026-07-23 | Flow-based generative models have enabled remarkable progress in fast and controllable generation across continuous and discrete state spaces, yet existing parameterizations are constrained to fixed dimensions or fixed sequence lengths. Here, we introduce Expanding Generative Flows (EFlows), which define flows between distributions of increasing dimensionality along an expanding interpolant that grows the state by augmenting it with conditional noise. Building on this construction, we propose Expanding Flow Maps (EFMs), a new class of flow maps that distill the expanding interpolant into efficient few-step generative models. Each EFM factors the map between any two timesteps into two learnable operations: an expand operator, which augments the state space with new coordinates or tokens conditioned on the current state, and a transport map, which pushes the expanded state forward along the interpolant. Composing these operators yields a single map that jointly expands and denoises the state, recovering existing fixed-canvas flows and flow maps as the special case in which the expand operator is the identity. We further extend the framework to the discrete simplex, enabling variable-size graph generation and variable-length sequence generation. Across both continuous and discrete modalities, we establish EFlows and EFMs as a principled framework for settings in which output size is itself a learned, controllable degree of freedom. | |
| Scale Up Strategically: Learning Compositional Generalization via Bias-Aware Evaluation and Data Collection for Robotic Manipulation | Yu Qi | 2026-07-23 | Compositional generalization is essential for robot to follow diverse instructions. However, pretrained policies are known to take shortcuts, deferring to salient cues rather than grounding language. We introduce a diagnostic framework that localizes this failure to individual \textit{instruction factors}, \textit{e.g.,} reusable semantic components such as color, verb, object, size, and spatial attribute. Our framework formalizes instruction factor bias, the tendency of fine-tuned policies to over-rely on dominant factors as shortcuts, and quantifies it through two metrics: Factor Dominance Rate (FDR), capturing pairwise bias between factors, and Factor Dominance Hierarchy (FDH), aggregating these into a global ranking. Evaluation on six foundation policies reveals broadly consistent ordering, \textit{i.e.}, color $\geq$ object $\geq$ spatial $\geq$ verb $\geq$ size, with color dominant, and verb and size most under-grounded. We further show the diagnosis is actionable: a bias-aware data collection strategy that reallocates a fixed budget toward under-grounded factors outperforms baselines in simulation and on a real robot using half the demonstrations, thereby enabling more sample-efficient and generalizable policy learning. | |
| GraphVid: Interactive Graph-Controllable Video Generation | Vedant Shah | 2026-07-23 | Controllable video generation remains challenging due to the difficulty of specifying precise multi-object interactions using text prompts or motion-control inputs that primarily constrain pixel movement. In practice, trajectory-based control often requires users to draw accurate tracks for multiple objects, which scales poorly with scene complexity and becomes ambiguous under occlusion or overlap. To enable flexible yet precise multi-subject control, we introduce $\textbf{GraphVid}$, a graph-conditioned image-to-video generation model that enables interactive control through structured interaction graphs. We further curate $\textbf{GraphVid-Bench}$, a large-scale interaction-centric video dataset with structured relational annotations to enable training of interaction-aware video generation models. Despite using substantially less training data and fewer trainable parameters than prior motion-control methods, GraphVid delivers strong controllability and video quality. Compared with Motion-I2V, GraphVid reduces FID by up to 39.9% and FVD by 37.6%, while improving PSNR (9.87=>15.98) and SSIM (0.38=>0.61). Our results highlight the potential of structured semantic interfaces as a powerful paradigm for controllable video generation. | |
| Barzilai-Borwein Fails Superlinear Convergence on an Open Set of Quadratics for Every Dimension $n\geq 4$ | Dawei Li | 2026-07-23 | Barzilai--Borwein (BB) method has shown strong practical performance in continuous optimization, yet its convergence dynamics remains poorly understood. In particular, a central unresolved question is whether BB converges superlinearly for almost every strictly convex quadratic problem and initialization. We provide a negative answer to this question. Specifically, for every finite dimension $n\geq4$, we construct a nonempty open, hence positive-Lebesgue-measure, family of strictly convex quadratic problems and initial points for which the long Barzilai--Borwein method (BB1) converges but cannot converge root-superlinearly. More precisely, with the explicit constants $ρ_{\min}=10^{-6},ρ_{\max}=0.61$, every spectral component of the gradient is bounded above and below by the corresponding geometric sequence. Consequently, the gradient norm and the energy norm of the error satisfy two-sided geometric estimates with the same rates, while the objective gap satisfies the corresponding estimates with squared rates. In particular, all three quantities are bounded below by geometric sequences, ruling out superlinear convergence. The construction is highly nontrivial, based on a computer-assisted proof of a nonresonant, attracting seven-cycle of the projectivized BB dynamics in dimension four. | |
| Synthetic data generation framework for quality control automation in gravure printing | Korota Arsène Coulibaly | 2026-07-23 | Quality control in printing, particularly in rotogravure printing, still depends on slow, costly, and subjective manual inspection. Automated surface defect detection is critical for maintaining high-quality standards in rotogravure printing. Deep learning models give prospects for automation. However, training robust deep learning models, such as YOLO or Vision Transformers, is heavily hindered by the extreme scarcity of real-world industrial defects images. To overcome this limitation, this paper introduces a novel synthetic data generation framework tailored for rotogravure printing quality control. The proposed pipeline automatically generates high-fidelity images of specific printing defects (creases, streaks, misregistration, etc.) and outputs corresponding bounding boxes and annotations. To validate the framework, a synthetic dataset of 7533 images was generated and used to train the state-of-the-art object-detection model RFDETR. Experimental results demonstrate that the model trained on our synthetic data achieves a Mean Average Precision (mAP) of 80.9\% on real industrial testing samples. This framework provides a zero-cost, rapid-deployment solution for automating defect inspection in printing lines without requiring massive manual data collection. | |
| Self-Supervised Learning of Structured Dynamics from Videos | Lukas Knobel | 2026-07-23 | Understanding motion in video is a fundamental challenge for visual learning, as frame-to-frame change entangles two sources of dynamics: camera motion and object motion. This decomposition has remained underexplored in representation learning, partly because these factors are tightly coupled in natural videos and difficult to supervise separately. Yet recovering it is important for learning robust motion representations that separate meaningful object dynamics from camera-induced variation. We study whether such structured motion representations can be recovered from frozen features of a pretrained image vision transformer. We propose the Structured Dynamics Model (SDM), which explicitly separates the dominant source of temporal change from residual dynamics through future-feature prediction, rather than representing video change with a single entangled latent or with unstructured, spatially dense transition tokens. Training combines self-supervised learning on real video with weak supervision of scene dynamics on synthetic Kubric data. We evaluate SDM on ProbeMotion, a new evaluation suite spanning synthetic and real videos with camera motion, object motion, and combined dynamics. SDM outperforms backbone baselines using global CLS or average-pooled features, and compares favorably to strongly supervised representations such as VGGT on several probes, despite using substantially weaker supervision. These results suggest that pretrained image models can be readily repurposed into structured video-dynamics representations, providing a useful inductive bias for learning and analyzing latent video dynamics. |
bioRxiv
| 标题 | 作者 | 发布日期 | PDF链接 | 摘要 |
|---|---|---|---|---|
| Mid-zone hepatocytes trade proliferation for survival via Atf4-Chop axis in early acute liver injury | Zhu, Y. | 2026-07-24 | Hepatocytes undergo extensive proliferation to facilitate liver repair after injury, yet early adaptive changes prior to proliferation remain unclear. Here, we report that during early acetaminophen (APAP)-induced liver injury, hepatocytes exhibit transient proliferation suppression that is most pronounced in mid-zone hepatocytes, consistent with zonal APAP metabolism. While spatial transcriptomics (ST) provided robust evidence for this arrest in the mid-zone, support for a similar arrest in pericentral hepatocytes was limited. Integrating ST with immunohistochemistry and functional studies, we identified a unique mid-zone stress-response program centered on the Atf4-Chop axis, which suppresses proliferation via the cell cycle inhibitor Btg2. Together, our findings support a model in which mid-zone hepatocytes transiently prioritize stress adaptation over proliferation, thereby preserving regenerative capacity for subsequent liver repair. | |
| Antifungal activity and mechanisms of D-limonene against Fusarium oxysporum, a pathogen of potato dry rot | Xia, Q. | 2026-07-24 | BACKGROUND: Potato dry rot (PDR), caused by Fusarium species, seriously threatens potato production and postharvest quality. Although limonene has demonstrated antifungal activity, previous studies have mainly focused on limonene-containing essential oils or phe-notypic growth inhibition, while its cellular and molecular mechanisms and application potential against PDR remain insufficiently understood. RESULTS: D-limonene inhibited the growth of Fusarium oxysporum in a concentra-tion-dependent manner, with a half-maximal inhibitory concentration (IC) of 8.32 L/mL. It also altered hyphal morphology, reduced pathogenicity and spore germination, decreased spore viability, and changed the chitin-associated fluorescent brightener 28 staining pattern. Tran-scriptome analysis identified 1,884 differentially expressed genes, including 1,027 downreg-ulated and 857 upregulated genes. Kyoto Encyclopedia of Genes and Genomes analysis and targeted annotation screening revealed extensive remodeling of cell-wall-related processes. Gene Set Enrichment Analysis further indicated suppression of ergosterol biosynthesis and ribosome biogenesis. D-limonene also increased sensitivity to temperature, salinity, and oxi-dative stresses and showed additive and synergistic interactions with mancozeb and hymexazol, respectively. CONCLUSION: Our results reveal that D-limonene shows potential as a bio-based component for the integrated management of PDR, providing a theoretical basis for its application. | |
| Early-life medial pulvinar disruption drives schizophrenia-relevant prefrontal inhibitory and cognitive deficits in primates | Scott, J. T. | 2026-07-24 | Schizophrenia is thought to arise from disrupted postnatal maturation of prefrontal circuits, but the developmental events linking early vulnerability to adult cortical dysfunction remain unclear. Here, we tested whether the primate medial pulvinar, a higher-order thalamic nucleus interconnected with prefrontal cortex, contributes to prefrontal maturation. Bilateral medial pulvinar lesions in neonatal marmosets altered adolescent prefrontal diffusion trajectories and produced adult working memory deficits, the latter of which did not follow comparable lesions in adulthood. Early-life lesioned animals showed reduced thalamocortical input to layer 3 parvalbumin interneurons, diminished prefrontal gamma power, reduced parvalbumin expression, and immature-like physiology in fast-spiking interneurons. These findings reveal a developmental window in which thalamic input shapes prefrontal inhibitory maturation, suggesting that some forms of cortical dysfunction in psychiatric disease originate not in the cortex itself, but in its thalamic inputs. | |
| Opposite and complementary roles of the two calcium thresholds for inducing LTP and LTD in models of striatal projection neurons | Trpevski, D. | 2026-07-24 | Synaptic plasticity has been shown to occur when calcium, flowing into the synapse due to incoming stimuli, surpasses a threshold level. This threshold level is modifiable through a process called metaplasticity. Some neurons, such as the striatal projection neurons, use different sources of calcium as the signal for synaptic strengthening (long-term potentiation, LTP) or weakening (long-term depression, LTD), resulting in them having two thresholds for inducing plasticity. In this study, we show that metaplasticity enables synapses undergoing both LTP and LTD during learning to selectively express just one form of plasticity (either LTP or LTD). To show this, we use the linear and nonlinear feature binding problem (FBP and NFBP) because their input patterns share features, exposing synapses to such competing LTP and LTD processes. In particular, we identify opposite and complementary roles of metaplasticity in the two thresholds for inducing LTP and LTD: metaplasticity in one threshold (e.g. LTD) allows synaptic plasticity of the opposite type (e.g. LTP) to be properly expressed. This happens because metaplasticity in the LTD threshold protects strengthened synapses from weakening, thus allowing them to persistently increase during learning (and encode learned patterns). Similarly, metaplasticity in the LTP threhsold prevents weakened synapses from strengthening, thus allowing them to persistently decrease. Under more general conditions for triggering metaplasticity, reversal learning can also be solved, demonstrating metaplasticity's importance for solving the plasticity-stability dilemma. Finally, we show that even though a single calcium threshold is sufficient for solving the FBP, NFBP and reversal learning, two calcium thresholds allow separate control over LTP and LTD. | |
| Perceiving less or perceiving unreliably? Disentangling thermosensory sensitivity and precision in the contexts of ageing and neuropathy | Courtin, A. S. | 2026-07-24 | Thermal perception is determined not only by sensitivity but also by precision. Yet, the latter is often overlooked in thermosensation and pain research. This study examined how ageing and diabetic polyneuropathy (DPN) affect these parameters and whether assessing both sensitivity and precision can aid in distinguishing patients from healthy controls. Using Bayesian hierarchical models, we estimated psychometric function thresholds (sensitivity) and slopes (precision) for cold detection, warm detection, cold pain, and heat pain stimuli delivered at the volar forearm, in a cross-sectional sample of 75 healthy adults (aged 21-80) and 33 patients with DPN. We also estimated these parameters separately for each participant and used the resulting estimates in classification analyses. Ageing was associated with elevated cold and warm detection thresholds, elevated cold pain thresholds, and reduced cold detection slope. Patients with DPN showed similar patterns: higher detection thresholds and lower cold detection slopes while pain-related parameters were largely unaffected. These findings indicate that ageing and neuropathy produce qualitatively similar changes in thermosensory function, particularly affecting cold detection. Classification based on single parameters successfully discriminated patients from controls, except when warm detection or heat pain slopes were used. Combining threshold and slope parameters for a given modality did not significantly improve classification accuracy but combining all parameters across all modalities led to the best performance, with excellent accuracy (AUROCC: .84, 95% CI [.75.91]). Modelling both thresholds and slopes provides a more comprehensive view of sensory decline and may enhance the detection of early or subtle sensory dysfunction. | |
| Electroconvulsive stimulation drives cortical spreading depression dependent immediate early gene expression in mice | Ladret, H. J. | 2026-07-24 | Electroconvulsive therapy (ECT) is a highly effective treatment for several psychiatric disorders, though its biological mechanisms remain unclear. Its therapeutic action has traditionally been attributed to the generalized seizure ECT induces. However, this view is challenged by the recent finding that electroconvulsive stimulation (ECS) can trigger a cortical spreading depression (CSD). Because CSD triggers massive intracellular molecular changes, we hypothesized that it could be a key mediator of ECT's therapeutic, plasticity-inducing effects. We observed similar neuronal oscillations following ECS in mice and patients undergoing ECT. We show that CSD drives increased expression of the immediate early gene Fos, a key marker of neuronal plasticity, and is associated with factors that predict positive ECT therapeutic outcome. Our results suggest that the therapeutic efficacy of ECT may be mediated by CSD. This challenges the seizure-centric model and implies that CSD, a currently unmonitored neurophysiological event, may serve as a more relevant biomarker for predicting and optimizing therapeutic outcomes of ECT. | |
| Rat mediodorsal thalamic subdivisions differentially modulate the sensory and affective components of pain through distinct prefrontal pathways. | Iben-Daoudi, H. | 2026-07-24 | The mediodorsal thalamus (MD) modulates pain through thalamocortical regulation of the mPFC. Yet, MD is often treated as a single anatomical and functional entity despite marked internal heterogeneity. Here, we tested whether medial-central MD (MDmc) and lateral MD (MDl) subdivisions exert dissociable control over sensory-discriminative and affective-motivational components of pain by engaging the anterior cingulate cortex (ACC) and prelimbic cortex (PrL). Using subdivision-selective excitotoxic lesions in rats, combined with anterograde tracing, laminar activity mapping, and projection-specific optogenetic manipulation of MDmc and MDl terminals in ACC or PrL, we determined the contribution of each subdivision and the underlying MD-PFC circuit mechanisms. Behaviorally, MDmc and MDl lesions induced mechanical and thermal hypersensitivity, but only MDmc lesions increased pain-related avoidance. Anatomical analyses showed that MDl preferentially innervated ACC PV cells, whereas MDmc more strongly targeted ACC SOM cells. Lesions further produced subdivision-dependent reorganization of nociception-evoked cFos activity in layers 2/3 and 5 and altered PV/SOM interneuron recruitment. Optogenetic manipulations revealed pathway-specific effects: MDmc-ACC/PrL manipulations enhanced nociceptive gain and avoidance, whereas MDl-ACC inhibition increased hypersensitivity while reducing avoidance, and MDl-PrL inhibition increased both nociceptive sensitivity and avoidance. Together, these findings identify MD subdivision-specific thalamocortical pathways that recruit distinct inhibitory microcircuits within ACC and PrL, thereby differentially shaping sensory-discriminative and affective-motivational components of pain. | |
| Microglia extract neuronal proteolytic organelles via skoupocytosis | Pepper, R. | 2026-07-24 | Neurons face unique challenges in maintaining protein homeostasis due to their tortuous morphology and extended processes. Proteolytic organelles are typically transported retrogradely towards their soma for degradation where lysosomes are enriched, but some organelles are larger than these neuronal processes, questioning how these organelles are degraded. Here we show that microglia, the resident immune cells of the brain, extract proteolytic organelles from neurons both in vitro and in vivo. Microglia make transient contact with neuronal membranes where proteolytic organelles are stationed beneath. At these sites of interaction, microglia pinch off a small portion of the neuronal process containing the organelle, leaving the rest of the process intact. We term this process skoupocytosis after the Greek word for garbage. Phosphatidylserine (PS) lipase ABHD16a accumulates near these proteolytic organelles and converts PS into lyso-PS to initiate microglial skoupocytosis. Skoupocytosis bypasses the need for retrograde organelle transport, providing homeostatic advantages for neurons that must maintain function in processes that extend extraordinarily long distances from the cell soma. | |
| Redistribution of sidechain-sidechain interactions govern ligand-specific binding affinity changes in missense Shank1 PDZ mutants | Santa, A. | 2026-07-24 | Shank proteins represent a family of abundant scaffolds in the postsynaptic density. Their dysfunctions had been identified as possible causes behind autism spectrum disorders and various types of cancer. The remarkably promiscuous PDZ domain of the Shank family is highly conserved through isoforms, and contains a unique dynamic segment, the {beta}2-{beta}3 loop, which is likely to play an important role in ligand selectivity. We used the Shank1 PDZ as a model system to analyze the perturbing effects of five disease-associated missense mutations on the binding of different partner peptides. Using experimental methods and molecular dynamics simulations, we characterized the interactions in detail, focusing on their dynamic aspect. While the investigated mutations in general weaken most interactions, the R736Q mutant, unique in having increased thermal stability, also binds the GKAP peptide with higher affinity than the wild type. Overall, our results show that the perturbing effect of mutations is highly partner-specific and depends on the dynamic rearrangements of both uniformly occurring and ligand-specific residue-residue interactions. | |
| The Anaphase Promoting Complex targets the toxic protein Progerin for ubiquitin-dependent degradation via autophagy | Eskiw, C. H. | 2026-07-24 | The premature aging disease Hutchinson-Gilford Progeria Syndrome (HGPS) results from the accumulation of progerin, a cytotoxic protein generated from a point mutation in the Lamin A/C gene, in the nuclear lamina. Upon the proper stimulation, cells degrade progerin, reversing cellular HGPS phenotypes; however, there is still a gap in our knowledge concerning which pathways are mediating progerin degradation. Previous data has demonstrated that the Anaphase Promoting Complex (APC), a multi-subunit ubiquitin ligase, tagets proteins for degradation, and that a decrease in APC function is linked with cellular aging. To determine if the APC is linked to HGPS disease phenotypes, we performed a meta-analysis of RNA-seq data from skin samples isolated from HGPS patients and identified dysregulation of several genes encoding subunits and substrates of the APC. Stimulation of APC activity decreased progerin protein levels and significantly decreased the number of cells with nuclear blebs. Proximity ligation assays (PLA) demonstrated that APC structure is compromised in HGPS cells and that APC stimulation increases proximity of the APC with progerin. Coimmunoprecipitation revealed that the APC co-activator, CDC20, physically interacted with nuclear lamina proteins. We further demonstrate that APC-mediated progerin degradation occurs through autophagy. Inhibition of the 26S proteasome enhanced progerin degradation, providing additional support for APC mediated-progerin degradation occurring independent of the proteasome. As such, we propose a previously unidentified interaction and mechanism by which cells remove progerin. This finding has impact on potential therapeutic strategies for HGPS, as well as providing further insight into linking the APC with both normal and premature aging. |
medRxiv
| 标题 | 作者 | 发布日期 | PDF链接 | 摘要 |
|---|---|---|---|---|
| Determinants of adolescent fertility among ever-married women in Bangladesh: an analysis of the 2022 Bangladesh Demographic and Health Survey | Khan, M. I. H. | 2026-07-24 | Bangladesh continues to report one of the highest adolescent fertility rates in South Asia, yet most prior national analyses have not fully accounted for the complex survey design of the Demographic and Health Surveys or assessed the robustness of findings to missing data. Using the most recent nationally representative data, this study examined factors associated with adolescent fertility among ever-married adolescent women aged 15-19 years in Bangladesh. We analysed data from the 2022 Bangladesh Demographic and Health Survey. Adolescent fertility was defined as being currently pregnant or having had at least one live birth. Of 2,449 eligible ever-married women aged 15-19 years, 1,601 with complete information on all model variables were included in the primary complete-case analysis. Survey-weighted logistic regression incorporating sampling weights, clustering, and stratification was used to estimate adjusted odds ratios (AORs) with 95% confidence intervals (CIs). Multiple imputation by chained equations among all 2,449 respondents was conducted as a sensitivity analysis. Analyses were performed in R version 4.5.1, and a two-sided p < 0.05 was considered significant. Adolescent fertility was strongly associated with age at first cohabitation. Compared with women who first cohabited before age 15 years, the odds were significantly lower among those cohabiting at 15-17 years (AOR = 0.41, 95% CI: 0.31-0.55) and 18-19 years (AOR = 0.15, 95% CI: 0.10-0.22; both p < 0.001). Women whose husbands had higher education had lower odds of adolescent fertility than those whose husbands had no formal education (AOR = 0.46, 95% CI: 0.27-0.77, p = 0.004), whereas a spousal age gap of 11 years or more was associated with higher odds (AOR = 1.63, 95% CI: 1.20-2.20, p = 0.002). Higher household wealth showed a borderline protective association (rich vs poor: AOR = 0.75, 95% CI: 0.56-1.01, p = 0.057). Respondent educational attainment and current working status were not independently associated with adolescent fertility after adjustment. Estimates from the multiply imputed datasets were broadly consistent with the complete-case findings. Adolescent fertility in Bangladesh is shaped by socioeconomic and relational factors, with age at first cohabitation the strongest determinant, alongside partner's education and spousal age gap. Multisectoral strategies that delay early marriage, strengthen child marriage legislation, engage male partners in reproductive health, and address socioeconomic and power inequalities are needed to reduce adolescent fertility and advance progress toward Sustainable Development Goal target 3.7 | |
| Multi-ancestry Genome-wide Association Study of Inpatient Opioid Dosing Following Knee or Hip Arthroplasty | Jinwala, Z. | 2026-07-24 | Opioids are commonly prescribed to manage acute postoperative pain. However, there is considerable variability in opioid administration patterns and practices, which is partially attributable to patient characteristics. We used electronic health record (EHR) and genotype data from the Million Veteran Program sample to investigate genetic predictors of individual differences in inpatient opioid analgesic exposure following knee or hip arthroplasty (n = 27,896). We extracted data from pharmacy records of administered inpatient opioid medications to derive a measure of average daily opioid exposure during the inpatient postoperative period. We then conducted genome-wide association studies (GWAS) to identify associated genetic variants in individuals of African-like (AFR; nAFR = 4,676), Admixed American-like (AMR; nAMR = 2,126), and European-like (EUR; nEUR = 21,094) genetically inferred ancestries. Models controlled for age, sex, pre-procedure opioid use disorder status, procedure type (knee or hip), length of stay, and the first 10 genetic ancestry principal components. The within-ancestry GWAS were followed by a cross-ancestry GWAS meta-analysis using fixed-effects inverse variance weighting in METAL. No loci reached genome-wide significance in the within- or cross-ancestry GWAS. Five loci were nominally significant (p <5 x 10-6) in the cross-ancestry GWAS, 9 in the AFR GWAS, 4 in the AMR GWAS, and 3 in the EUR GWAS. This study provides a framework for the use of EHR data to examine the genetics of opioid exposure in postoperative care and indicates the need for larger samples and more precise phenotyping to better understand genetic contributors to individual variation in analgesic requirements. | |
| Causal Mediation Pathways in Continuous Postprandial Glucose Monitoring for Type 1 Diabetes Patients | Hilligoss, S. | 2026-07-24 | Managing postprandial glucose in Type 1 Diabetes Mellitus (T1DM) requires understanding how carbohydrate intake affects glucose through both direct pathways and insulin-mediated compensation. Standard analyses treat insulin as a confounder rather than a mediator, obscuring these distinct causal channels and patient-level heterogeneity in carbohydrate response. Using meal-centered continuous glucose monitoring windows from twelve adults in the OhioT1DM cohorts, we apply causal mediation analysis to decompose the total effect of carbohydrate intake on glucose change into average direct (ADE), insulin-mediated (ACME), and total effects, by meal type and across outcome quantiles. To adjust for pre-meal confounding, we introduce a Causally-constrained Linear Autoencoder that learns low-dimensional representations improving adjustment for measured pre-meal state. Across cohorts, carbohydrate intake raises postprandial glucose chiefly through a large direct effect that insulin only partially offsets. The prespecified primary endpoint, the pooled ADE of a +30 g contrast at two hours, is large and highly significant, robust to multiple-testing correction, and replicates in the independent DiaTrend cohort (54 adults); the insulin-mediated effect is comparatively small. Meal- and quantile-specific signals, including greater dinner-time excursions, do not survive false-discovery-rate correction and are reported as hypothesis-generating. These results localize where postprandial insulin dosing may be refined. | |
| Ring and community vaccination for Bundibugyo ebolavirus outbreak response: a stochastic network modelling study | Andrews, J. R. | 2026-07-24 | Background: Vaccination with rVSV-ZEBOV is highly effective against Zaire ebolavirus, but protection against Bundibugyo ebolavirus (BDBV) is unproven. We evaluated the relative population impact and dose efficiency of a partially cross-protective hypothetical vaccine under operationally realistic constraints during a BDBV outbreak. Methods: We developed a stochastic transmission model on a two-layer household-community contact network calibrated to 2026 Democratic Republic of the Congo BDBV outbreak data. Time-varying effective reproduction numbers were estimated using a Bayesian renewal model. We evaluated case detection, isolation, contact tracing, reactive ring vaccination, and community vaccination (20-80% coverage). Base-case vaccine efficacy was 45% and included post-exposure protection against disease and mortality. Primary outcomes were mortality and incidence reductions, total doses, and dose efficiency (doses per death averted) over 90 days, evaluated in a probabilistic sensitivity analysis with 10,000 matched stochastic replicates per strategy. Results: Compared with base operations alone (30% detection, 30% tracing), enhanced operations alone (70% detection, 80% tracing) reduced expected mortality by 81.9% (95% UI 72.3-88.8). Reactive Ring 2 vaccination under base operations reduced mortality by 7.8% (4.7-10.9), requiring 89.5 doses per death averted. Added to enhanced operations, Ring 2 vaccination reduced mortality by 82.5% overall (73.0-88.9), an incremental benefit of 2.7% (0.9-4.9) beyond enhanced operations alone. Community vaccination at 20%, 40%, 60%, and 80% coverage reduced mortality by 44.4% (33.9-51.8), 67.9% (56.7-75.6), 80.4% (70.8-85.9), and 87.2% (79.6-91.3), requiring 40.3 to 80.9 doses per death averted. Interpretation: Strengthened case finding, contact tracing, and isolation averted most deaths even without vaccination. Once these operations were strong, reactive ring vaccination added little further benefit, whereas rapid community vaccination produced the largest reductions in simulated scenarios but required substantially more doses. A partially protective BDBV vaccine's population-level value will depend principally on rapid, broad delivery. | |
| Spectral Validity and Spindle Detection of Wearable Frontal EEG: A Per-Subject Calibration Framework and Systematic Validation Against Polysomnography Using the Wearanize+ Dataset | Parry, Y. D. | 2026-07-24 | Wearable EEG devices enable home sleep monitoring but require systematic spectral validation before their physiological outputs can serve as proxies for polysomnographic features. This study provides comprehensive spectral validation of the Zmax EEG headband against concurrent PSG using the Wearanize+ dataset. Seventy-one participants with adequate signal quality underwent simultaneous home PSG and Zmax recording. Bandpower correspondence, calibration robustness, within-subject reliability, lateralisation, and spindle detection were evaluated across all sleep stages. Zmax systematically underestimates bandpower across all frequency bands (bias -0.41 to -0.74 log units), attributable to the active Fpz reference electrode. A per-subject N2-referenced calibration eliminates this bias; N2 calibration outperformed N3 and REM alternatives (mean post-calibration r=0.601 vs 0.479 and 0.489). Post-calibration spectral correspondence was strong for alpha (N3: r=0.806) and sigma (N3: r=0.752). Within-subject reliability was excellent (split-half r>0.99). Demographic factors explained less than 4% of offset variance. Lateralisation analysis was underpowered (36-39% power; N=194 required for 80% power). Spindle under-detection was traced to YASA's relative sigma power pre-filter; lowering this threshold recovered PSG-equivalent counts with near-zero bias. These findings establish a validated calibration framework and evidence-based feature selection recommendations for Zmax-based sleep biomarker research. | |
| Clinical Impact, Diagnostic Performance, and Prognostic Implications of Plasma Metagenomic Next-Generation Sequencing in Solid Organ Transplant Recipients | Spottiswoode, N. | 2026-07-24 | Background: Plasma metagenomic next-generation sequencing (mNGS) may detect pathogens in solid organ transplant (SOT) recipients, but optimal patient selection and result interpretation remain uncertain. Methods: We studied 145 SOT recipients who received Karius plasma mNGS testing between 2017-2025. Retrospective multi-physician adjudication assessed diagnostic accuracy, clinical impact, and outcomes. We examined whether detection of atypical bacteria, invasive fungi, adenovirus/parvovirus and parasites, termed pre-specified organisms of presumed significance (POPS), predicted positive clinical impact. We applied a GPT-4o large language model (LLM) to limited electronic medical record (EMR) data to identify patients with POPS diagnoses and positive-impact testing. Results: Of 145 SOT recipients, 119 (82.1%) had positive tests, and 42 (29.0%) had [≥]1 POPS organism detected. Twenty-seven of 133 diagnoses (20.3%) were made by mNGS first or mNGS only. Diagnostic performance versus a gold standard of all microbiologic testing varied by organism, ranging from 100% sensitivity and specificity (Bartonella, Nocardia) to 60.0% and 53.3% sensitivity for Coccidioides and Aspergillus, respectively. Of 141 patients with interpretable test impact, 27 (19.1%) had positive clinical impact, associated with POPS detection (P<0.001). The LLM identified patients with POPS diagnoses (area under the receiver operating characteristic curve [AUC] 0.86), and patients with positive-impact testing (AUC 0.71). Conclusions: Plasma mNGS can aid diagnosis and management of infections in SOT recipients, but negative tests do not exclude invasive fungal disease. Positive clinical impact is greatest when POPS organisms are detected, and patients at risk of POPS diagnoses may be identified by an LLM given limited EMR data. | |
| A systematic review to critically appraise methodological rigour in research on ultra-processed food and cardiovascular disease and hypertension | Mekonnen, T. C. | 2026-07-24 | The authors have withdrawn this manuscript because this withdrawal is made in order to correct the title, submit a revised and updated version of the manuscript. The revised submission includes improved analyses, updated figures, and responses to feedback received. The new version supersedes the previous preprint. Therefore, the authors do not wish this work to be cited as reference for the project. If you have any questions, please contact the corresponding author. | |
| Translational Study of using FOCM/TS Metabolites for Supporting Autism Spectrum Disorder Diagnosis | Arici, H. | 2026-07-24 | Purpose Several clinical studies have shown correlations between certain physiological measure-ments and an ASD diagnosis. Such findings, however, have generally not resulted in tangible progress towards practical translation due to a number of factors which this work seeks to address. Methods This paper presents a double-blind case/control trial design in which metabolic profiles, collected at two developmental pediatric clinics, were collected from children on a diag-nostic waitlist for the purpose of developing a blood-based test for ASD. Besides obtain-ing blood samples, the children underwent gold-standard clinical evaluations, including the Autism Diagnostic Observation Schedule (ADOS), Mullens Scale of Early Learning (MSEL), and Vineland Adaptive Behavior Scale (VABS). The analysis, together with a complete medical history and physical exam, allowed to confirm or rule-out suspected ASD using DSM-5 criteria. The study was based on a cohort of 140 children between the ages 18-60 months, that were referred to a developmental pediatrician because of con-cerns in their development. Results 114 of these children received an ASD diagnosis, while 26 were diagnosed with non-ASD related developmental delays. Based on the measured metabolites, artificial intelligence-based classification algorithms allowed for an over 80% accuracy in predicting whether a sample came from a child diagnosed with ASD or not. Conclusion While these results need to be replicated in a larger study, especially involving more chil-dren with non-ASD related developmental delays, this is the first work using physiologi-cal measurements, coupled with AI, to support ASD diagnoses in a clinically relevant set-ting. The clinical trial that was part of this work was registered on clinicaltrials.gov as NCT04672967 and was entitled the Metabolic Autism Prediction (MAP) Study. The study was IRB approved by the Biomedical Research Alliance of New York (BRANY) IRB on August 12, 2021 (approval number: A21-10-282-888). Keywords: Autism spectrum disorder, Folate-dependent one-carbon metabolism path-way, Transsulfuration pathway, Translational study, Machine learning, Clinical study | |
| Social vs individual training: a cross-sectional observational study of perceived motivation, safety, effort, mood, and satisfaction in recreational exercisers | Sorrentino, M. | 2026-07-24 | Background: Training context may shape motivation, perceived effort, confidence, mood, and satisfaction during exercise, but evidence from real-world self-reported samples remains limited. This dataset includes 155 respondents and compares individuals who reported training mainly alone with those who reported training mainly in groups. Objective: To describe the sample, summarize questionnaire responses, and discuss an analytically appropriate framework for comparing perceived psychological and performance-related dimensions across training contexts. Methods: A cross-sectional observational study was conducted using questionnaire data exported to Excel from a convenience sample of physically active adults. Baseline variables included age group, sex, education, training history, weekly training frequency, and preferred training modality. Fourteen Likert-type items were recorded; however, the questionnaire structure was asymmetric, with the first seven items completed only by respondents training mainly in groups and the second seven available across the full sample, including 116 respondents training mainly alone and 39 training mainly in groups. Descriptive statistics were summarized as counts and percentages for categorical variables and as mean, standard deviation, median, and interquartile range for ordinal items. For inferential comparisons on comparable ordinal outcomes, the prespecified approach was the Mann-Whitney U test, with multivariable ordinal regression considered for adjusted analysis. Results: The final sample comprised 155 respondents, of whom 116 mainly trained alone and 39 mainly trained in groups. The distribution of training modality was therefore substantially unbalanced toward individual training. Across the seven items available in the full sample, median responses were generally centered around 3, suggesting moderate or neutral-to-mild agreement overall. Mean values for these items ranged approximately from 2.65 to 3.32, whereas group-specific items in the group-training block were generally higher, with means often above 4.0, although those items were not directly comparable across modalities because they were only completed in one subgroup. Accordingly, the most defensible interpretation is descriptive: the sample suggests broadly moderate perceived benefits of training context, but the questionnaire structure limits strong head-to-head causal or comparative claims. | |
| Adverse and benevolent childhood experiences and depression among women in rural Pakistan | Gallis, J. A. | 2026-07-24 | Adverse childhood experiences (ACEs) are associated with maternal depression, including during the perinatal and postpartum periods when women face elevated risk. Benevolent childhood experiences (BCEs) may buffer negative effects of ACEs and promote mental health, yet evidence on their joint effects remains limited, particularly in low-resource settings. We investigated the joint effects of ACEs and BCEs on maternal depression from the third trimester of pregnancy through eight years postpartum among 804 women in rural Pakistan. Childhood experiences were characterized using both aggregated scores and latent class analysis to capture overall burden and item-level patterns. We tested the linear interaction term of ACEs and BCEs and utilized subgroup analysis to assess non-linear interactions. Overall, more than half (58.5%) of women experienced at least one ACE, and 6.2% of women experienced four or more ACEs. Commonly reported ACE domains were home violence (39.3%), neglect (19.7%) and family psychological distress (15.2%). Nearly half (45.3%) of women experienced all ten BCEs, and over half (51.5%) experienced 6-9 BCEs. Some BCE items had a very high prevalence, including liking themselves/feeling comfortable with themselves (96.6%), having at least one caregiver with whom they felt safe (96.5%), and having good neighbors (94.9%). Adjusting for ACEs, higher levels of BCEs were independently associated with fewer depressive symptoms ({beta} = -0.25; 95% CI: -0.45, -0.05) and the promotive effects were consistent across ACE domains. Evidence of interaction suggested that BCEs buffered the adverse effects of ACEs among women with 1-3 ACEs ({beta} = -0.56; 95% CI: -0.86, -0.26), but not among those with four or more ACEs. The associations between exposure to ACEs and depressive symptoms were weaker among women with higher levels of BCEs. Latent class analysis identified four distinct childhood experience profiles: Low-ACE/High-BCE (58.6%), High ACE/High BCE (19.5%), Low ACE/Low BCE (13.8%), High ACE/Low BCE (8.1%), with women in the High ACE/Low BCE class experiencing the greatest depressive symptom burden. Findings suggest that both adverse and benevolent childhood experiences are important for understanding maternal depression in low-resource settings, and informing interventions that reduce adversity exposure and foster positive developmental resources. |