Browse Articles

Assessing conventionally bred lentil (Lens culinaris Medik) cultivars for adaptation to organic cropping systems in the southeastern USA

Many AI analysts, one dataset: Navigating the agentic data science multiverse

Empirical conclusions depend not only on data but also on analytic decisions. Many-analyst studies have quantified this dependence: independent teams testing the same hypothesis on the same dataset regularly reach conflicting conclusions. But such studies require costly human coordination. We show that fully autonomous AI analysts built on large language models (LLMs) can, cheaply and at scale, produce the analytic dispersion observed in human many-analyst studies. In our framework, each AI analyst independently executes a complete analysis pipeline on a fixed dataset and hypothesis; a separate AI auditor screens every run for methodological validity. Across three datasets, AI analyst-produced analyses exhibit substantial dispersion in effect sizes, P -values, and conclusions. This dispersion can be traced to identifiable analytic choices in preprocessing, model specification, and inference that vary systematically across LLM and persona conditions. Critically, the outcomes are steerable: reassigning the analyst persona or LLM shifts the distribution of results even among methodologically sound runs. These results highlight a central challenge for AI-automated empirical science: when defensible analyses are cheap to generate, evidence becomes abundant and vulnerable to selective reporting. The same capability also helps address it: treating analyst results as distributions makes analytic uncertainty visible, and deploying AI analysts against a published specification can reveal how much disagreement stems from underspecified design choices. Taken together, our results motivate a transparency norm: AI-generated analyses should be accompanied by multiverse-style reporting and full disclosure of the prompts used, on par with code and data.

RISC-V-based YOLOv3-tiny acceleration with runtime-reconfigurable systolic arrays and custom instructions

Sequential evolution of antidote and toxin links genetic incompatibility with immune responses

Toxin-antidote (TA) systems are selfish genetic elements that ensure their own inheritance by eliminating offspring that do not inherit the module, thereby creating postzygotic genetic incompatibilities both within and between species. Despite their ubiquity and substantial fitness costs, the origin and persistence of TA systems remain poorly understood. Here, we report a TA gene pair in the nematode Caenorhabditis nigoni . The antidote gene, Cni-shls-2 , is a C. nigoni -specific F-box gene that arose through recent tandem duplications, leading to three identical copies. Its absence results in embryonic lethality in both C. nigoni and its hybrids with the sister species Caenorhabditis briggsae . This lethality is mediated by a maternally deposited toxin, Cni-hlix-1 , a chimeric gene formed by the fusion of duplicated host sequences with novel sequences that could be derived from bacteria/archaea. Analysis of evolutionary trajectory of the TA genes among various populations suggests that the antidote is more likely to predate the toxin. These results support a possible model of TA origin, in which Cni-shls-2 initially evolved under pathogen pressure, whereas the subsequent emergence of the toxin enforces antidote retention. We speculate that host–pathogen conflict may serve as a key driving force in the evolution and maintenance of TA systems, inadvertently leading to reproductive barriers.

Association of target miRNAs expression in blood plasma and cerebrospinal fluid with Alzheimer’s disease biomarkers level and cognitive decline

A general mobility analysis framework to reveal nonintuitive kinematic characteristics of rigid origami

Rigid origami exhibits intriguing reconfigurability governed by geometric constraints imposed by its creases. This intrinsic relationship gives rise to nonintuitive kinematic characteristics, including the number of degrees of freedom (DOFs) and their corresponding motion paths in complex configurations. Here, we develop a general mobility analysis framework to reveal the indeterminate kinematic characteristics of rigid origami. By first analytically deriving kinematic compatibility conditions up to the fourth order, we clarify the role of high-order compatibility analysis in determining rigid foldability. The results show that origami structures with infinitesimal mechanisms can be classified into three categories: regular origami, bifurcated origami, and shaky origami. These kinematic characteristics can be revealed by second-order analysis in most cases, with only a few cases requiring third-order analysis. Based on the analytical results, we then develop an angle-dominated simulation algorithm for tracing the finite motions of rigid origami, including single- and multi-DOF regular origami as well as bifurcated origami. The proposed framework is validated on 20 previously reported origami structures and 31 newly designed structures, including planar origami, polyhedral origami, and thick origami, confirming the generality of our approach. This work elucidates the nonintuitive kinematic behaviors of rigid origami by integrating analytical derivation, numerical simulation, and prototype validation, advancing the mobility analysis of rigid origami and other spatial linkages.

Underwater biological target detection algorithm based on DVS-YOLO11

Structural basis of membrane engagement and polyreactivity control in HIV-1 MPER broadly neutralizing antibodies

The membrane-proximal external region (MPER) of HIV-1 Env represents a critical target for broadly neutralizing antibodies (bnAbs) due to its conservation and functional importance. However, MPER-targeting bnAbs recognize composite epitopes comprising peptide and viral membrane lipid components, creating an inherent tension between viral neutralization efficacy and polyreactivity. 10E8-class antibodies exhibit high neutralization potency with low polyreactivity, whereas 4E10-class antibodies show comparably broad neutralization but higher polyreactivity, underscoring the need to understand the structural basis of this distinction. We therefore determined crystal structures of DH511.1 (memory B cell–derived), DH511.12P (plasma cell-derived), and VRC42.01 in complex with MPER peptide and phosphatidic acid, along with a cryo-EM reconstruction of DH511.2 bound to membrane-embedded Env. Through integrative analysis taking into account previously determined structures of other MPER bnAbs, we reveal two distinct lipid recognition strategies. Groove-mediated binders, including 10E8 and DH511, engage lipids through antibody–membrane interface grooves with distinct geometries and angular approaches to the membrane. In contrast, heavy chain-mediated binders, including 4E10, PGZL1, and VRC42, utilize positively charged CDR H1 patches for direct lipid headgroup recognition. Importantly, DH511 lineage members exhibited differential cardiolipin polyreactivity linked to their maturation stage. PGZL1 and VRC42.01 employ weaker positive patches at lipid-binding sites than 4E10, and PGZL1 additionally introduces a CDR H3-mediated negative patch that creates electrostatic repulsion with negatively charged lipid headgroups, thereby limiting nonspecific interactions. These findings provide a structural framework for understanding how MPER bnAbs balance lipid binding with specificity and inform immunogen design for inducing safe and effective neutralizing responses.

Message accuracy labels but not source verification labels affect message engagement and persuasion of nutrition-related social media content through information credibility

Spatiotemporally controlled matrix softening facilitates deterministic crypt formation in human intestinal organoids

Spatially controlling morphogenesis is a challenge for many organoid systems that manifests as a limited understanding of self-organization of differentiating cells and leads to a high degree of heterogeneity in organoid morphometrics. Current methods to grow organoids rely on temporal presentation of soluble cues that are not controllably delivered, and little is known about the role of the extracellular microenvironment in this process. Here, we present a material-based strategy to spatiotemporally control morphogenesis of human intestinal organoids (HIOs) with predictable crypt morphometrics and cell composition that match their in vivo tissue counterparts. We first optimize culture conditions to generate more reproducible HIOs with predictable growth in phototunable poly (ethylene glycol)-based hydrogels, and then systematically investigate the role of light-mediated matrix softening in guiding crypt formation. The light dose delivered to crypt-sized regions adjacent to growing organoids is a key factor in maintaining organoid cell viability, as well as crypt budding and elongation. With optimized light doses, predictable epithelial shape changes result in programmable crypt formation, confirmed by the presence of proliferative (Ki67+) and niche-defining Paneth (Lyz+) cells. This methodology could be readily adopted for other budding and branching organoids to facilitate controllable changes in morphogenesis or cell migration. Sequential patterning approaches and more complex pattern designs could further open the parameter space to facilitate modeling of a wide array of engineered tissues for applications ranging from fundamental biology to disease modeling and translational medicine.

A large-scale analysis of open higher education student attrition: variables characterizing study interruption and withdrawal

Raman imaging of the phycosphere reveals sharp gradients of organic matter exuded by single phytoplankton cells

Phytoplankton cells exude a wide array of chemicals in the water column, generating a localized microenvironment known as the phycosphere. Although it is now well accepted that the phycosphere mediates interactions between phytoplankton and bacteria, the chemical gradients around individual phytoplankton cells have never been explicitly measured, and their shape has been classically assumed to be set by ideal diffusion. Here we used Raman microspectroscopy to obtain micrometer-scale measurements of the concentration profile of a phytoplankton metabolite (fucoxanthin) around individual phytoplankton cells of different species, having radii between 2.5 and 60 μ m. We found that fucoxanthin concentration decreases more rapidly with distance from the cell than predicted by ideal diffusion, showing that the phycosphere includes compounds whose diffusion is characterized by nonideal effects. We explain this observation using a space-dependent diffusivity model where nonideality arises from viscosity and solubility gradients in the extracellular environment. Our results suggest an onion-structured model of the phycosphere, in which small hydrophilic solutes that obey ideal diffusion generate broad but weak gradients, whereas insoluble compounds are retained within 10 to 20 μ from the phytoplankton cell surface and yield steep gradients of organic matter. These observations, supported by evidence that fucoxanthin can act as an effective chemoattractant for marine bacteria, show the existence of strong and highly localized chemical cues with potentially far-reaching impacts on microbial interactions in aquatic environments. These findings highlight the importance of directly measuring the microscale chemical landscape experienced by marine microbes.

Green synthesis of graphene oxide–TiO2/Fe2O3/ZnO nanocomposites for photocatalytic degradation of petroleum refinery wastewater

APOBEC2 acts as an enforcer of lymphocyte lineage identity

Fluorescence characteristics and bioimaging application of carbon dots extracted from sugar mill flue gas

Reply to Sun: Behavior affects epidemic outcomes

Emerging XDR and MDR Pseudomonas aeruginosa strains from Oreochromis niloticus reveal high pathogenicity, virulence determinants, antimicrobial resistance genes, and oprL sequence diversity

Enteroendocrine cells wire the gut–brain vagal axis

The vagal sensory nervous system plays important roles in mediating gut–brain communication and maintaining physiological homeostasis. Although recent research has greatly enhanced our understanding of the function of vagal sensory neurons, little is known about what guides the vagal sensory nerve fibers to innervate the intestine and form a complex gut–brain sensory network. Here, using zebrafish genetic models to trace the development of the intestinal vagal network, we found that the development of the intestinal vagal sensory network is coupled with the formation of enteroendocrine cells (EECs). Vagal sensory neurons begin to innervate the intestine at 2 d postfertilization. Vagal nerve fibers branch out toward the newly formed EECs in the intestinal epithelium. The newly formed EECs display active actin filaments at their base, enabling them to physically engage with the vagal sensory fibers and facilitate the formation of the intestinal vagal sensory network. Genetically ablating EECs impairs intestinal vagal network formation and alters the vagal central projection pattern. Ablating EECs induces apoptosis in a subset of vagal sensory neurons and reduces the number of central projecting axons of the intestine innervating vagal sensory neurons. Moreover, ablating EECs completely alters the brain’s response to nutrient ingestion and diminishes nutrient-induced hindbrain and hypothalamus neuron activation. Finally, loss of EECs increases food intake while impairing adipose tissue development and survival. Together, our study revealed that EECs guide vagal sensory neuron development and intestinal vagal network formation. Loss of EECs impairs the anatomical vagal sensory axis and alters gut–brain signaling transmission.

Constructing the conditional association network and identifying central and bridge nodes in aneurysmal subarachnoid hemorrhage

Goal-directed modulation of the default network supports interactions between selective attention, working memory, and prior knowledge

Memory is central to flexible, goal-directed behavior. Prior knowledge shapes our current decisions and future plans and guides our attention to selectively prioritize relevant information in our complex, noisy world. Yet, most research on the neural basis of selective attention has focused on novel stimuli and has defined goal relevance based on perceptual features rather than prior knowledge. We investigated how the brain supports selective attention when goal relevance depends on prior knowledge, and compared selection of novel (memory-independent) versus familiar (memory-enhanced) stimuli. Using multiecho fMRI, we recorded brain activity during a selective working memory task involving pictures of famous (familiar) and anonymous (novel) people and places. We manipulated the relevance of fame to task performance: In “fame relevant” task blocks, participants selectively attended to famous while ignoring anonymous stimuli; in “fame irrelevant” blocks, they attended to anonymous while ignoring famous stimuli. Compared to a nonselective, categorization task, the selective working memory task elicited sustained coactivation of frontoparietal control and dorsal attention networks. Task-relevant contrasted with task-irrelevant stimuli elicited transient coactivation of frontoparietal control and default network regions, irrespective of stimulus fame. Within the DN, this effect was driven by both enhanced activation for task-relevant stimuli and suppression for task-irrelevant stimuli, confirming attention-driven neuromodulation. Further, selective working memory coactivated these high-order, heteromodal brain networks while lowering activity in visual regions. Our findings elucidate the neural underpinnings of prior knowledge-guided selection of perceptual inputs for working memory, suggesting differences from those previously reported for perception-guided selection.