Abstract: Advances in generative AI have brought advanced tools to tax & legal practice, but with them comes AI hallucinations”, which are fabricated citations, quotes, or facts that appear plausible but are entirely false. In 2023, a notable U.S. case (Mata v. Avianca) revealed this risk when attorneys, relying on ChatGPT for research, submitted a brief containing fictitious case law and were sanctioned as a result. This incident revealed substantial risks for the tax & legal profession. Increased AI use in tax & legal submissions and decision drafting subsequently led to numerous similar global incidents. By late 2025, a collection of various datasets logged nearly 800 cases of AI-related citation errors or hallucinations” in at least 25 countries, with a marked increase in 2025 alone. These cases span court filings by lawyers and pro se litigants, as well as orders drafted by judges or tribunals. This development necessitates an examination of professional responsibility and procedural fairness concerning AI-generated falsehoods. This article analyzes how courts and administrative bodies across jurisdictions have responded to AI-generated hallucinations in tax & legal submissions and decisions, and what these responses indicate regarding emerging verification standards under existing law. The analysis compares incidents from the United States, Canada, the United Kingdom, India, Israel, and other jurisdictions, focusing on the imposition or withholding of sanctions, the treatment of various actors (e.g., lawyers, self-represented parties, experts, judges), and the adaptation of tax & legal doctrines to this challenge.

Abstract: Direct-to-consumer (D2C) brands increasingly rely on A/B testing to optimizepaid social advertising, yet common execution errors—inconsistent attribution,underpowered samples, mid-test edits, and metric flexibility—undermine inferentialvalidity.[13] This paper develops a methodological framework for decision-gradeexperimentationonMeta(Facebook)Adsthatbalancesstatisticalrigorwithbusinessguardrails. We synthesize established practices in online experiments[3, 12, 13] andoperationalize them into a prioritized test sequence, integrating power analysis,Sample Ratio Mismatch (SRM) diagnostics,[21, 23] optional CUPED variancereduction,[6] and economic guardrails (e.g., ROAS thresholds, delivery balance,frequency parity).Evidence and scope. All analyses use a synthetic/simulated dataset calibratedto D2C benchmarks; no live-traffic data are analyzed. The simulation demonstratesthe analytical workflow and decision logic without making empirical claims aboutreal-world effectiveness. The contribution is methodological: a reproducibletemplate practitioners can adapt and validate in production.

Abstract: Misdeclared hazardous cargo inside sealed maritime containers continues to drive ship fires, port disruptions, and avoidable losses for crews, carriers, and coastal communities. Ports and carriers already use non-intrusive inspection (NII) imaging and document checks, but these systems are often treated as separate queues rather than a single integrated decision system. This paper proposes a practical multimodal framework that fuses radiographic sensing with shipping document intelligence to flag hazardous misdeclaration risks without opening a container. The approach combines a vision encoder that learns density-aware patterns from X-ray or gamma imagery with a language model that extracts and normalizes claims from bills of lading, manifests, and booking descriptions. A fusion layer learns cross-modal consistency, so the system can react when what the scan suggests does not match what the paperwork claims. The design is grounded in port constraints, including strict throughput targets, noisy and delayed labels, long-tailed hazard categories, and an operational need for clear explanations that can be audited. We define a data lifecycle that turns inspections, holds, claims, and incident investigations into structured feedback, without requiring constant full unpacking of cargo. We describe a low-latency edge deployment pattern that reduces backhaul and helps avoid unnecessary centralized compute, which matters in regions where water and power constrain expansion planning. A simulation-driven evaluation plan is provided, including realistic cost-sensitive metrics that focus on recall at fixed false-alarm rates, because ports pay real costs for every extra secondary inspection. The paper positions the framework relative to DHS work on AI-enabled paradigms for non-intrusive screening and recent industry adoption of AI screening for dangerous goods in booking workflows.

Abstract: The volume of scientific submissions continues to climb, outpacing the capacity of qualified human referees and stretching editorial timelines. At the same time, modern large language models (LLMs) offer impressive capabilities in summarization, fact checking, and literature triage, making the integration of AI into peer review increasingly attractive—and, in practice, unavoidable. Yet early deployments and informal adoption have exposed acute failure modes. Recent incidents have revealed that hidden prompt injections embedded in manuscripts can steer AI-generated reviews toward unjustifiably positive judgments. Complementary studies have also demonstrated brittleness to adversarial phrasing, authority and length biases, and hallucinated claims. These episodes raise a central question for scholarly communication: when AI reviews science, can we trust the AI referee? This paper provides a security- and reliability-centered analysis of AI-assisted peer review. We map attacks across the review lifecycle—training and data retrieval, desk review, deep review, rebuttal, and system-level. We instantiate this taxonomy with four treatment-control probes on a stratified set of ICLR 2025 submissions, using two advanced LLM-based referees to isolate the causal effects of prestige framing, assertion strength, rebuttal sycophancy, and contextual poisoning on review scores. Together, this taxonomy and experimental audit provide an evidence-based baseline for assessing and tracking the reliability of AI-assisted peer review and highlight concrete failure points to guide targeted, testable mitigations.

Abstract: The integration of outdoor camera images with three-dimensional (3D) geographic information on the observed scene has an interest in many applications of video acquisition. To solve this data fusion problem, camera images have to be matched with the 3D geometry provided by the geographic information system (GIS). This paper proposes to use, for a camera of known geographical position, a dense local azimuth-elevation map (LAEM) derived from a gridded digital elevation model (DEM) as data representation to ease the matching operation between GIS data and the image. Such a map assigns to each regularly sampled azimuth and elevation angles pair the geographic point derived from the DEM viewed in this direction. The problem of computing the LAEM from the DEM is closed to the problem of surface rendering, for which solutions exist in computer graphics. However, rendering software cannot be used directly, since their view directions are constrained by the pin-hole camera model and apparent colour rather than position of the viewed point is assigned to the viewing direction. This paper therefore also proposes a specific algorithm for the computation of the LAEM from the DEM. A MATLAB implementation of the algorithm is also provided, which is tailored to process the DEM data set swissALTI3D from the Swiss Federal Office of Topography swisstopo.

Abstract: In the current study, we propose a novel reduced-order model for the drag coefficient of a circular cylinder model that can be either fixed or undergoing an oscillatory linear motion in the cross-flow direction, the streamwise direction, or at an arbitrary tilt angle. Thus, the proposed model is not restricted to a single geometric setting of the cylinder. The model establishes a proper nonlinear coupling between the drag coefficient and the lift coefficient, such that the drag coefficient can be restructured using the simple reduced-order model, given the time signal of the lift coefficient. The proposed model is able to capture both the mean component of the drag coefficient, as well as the oscillatory component of it. We derived closed-form expressions to estimate the model parameters from a training dataset. The model was tested and found to be performing satisfactorily under different motion modes. We generated the training data using computational fluid dynamics simulation for a circular cylinder at a low Reynolds number of 300. The computational fluid dynamics solver used was successfully validated by comparison against independent published data. The current study is viewed as a contribution to the fields of nonlinear dynamics, fluid mechanics, and computational mathematics.

Abstract: Generative diffusion models have emerged as a powerful class of models in machine learning, yet a unified theoretical understanding of their operation is still developing. This paper provides an integrated perspective on generative diffusion by connecting the information-theoretic, dynamical, and thermodynamic aspects. We demonstrate that the rate of conditional entropy production during generation (i.e. the generative bandwidth) is directly governed by the expected divergence of the score function's vector field. This divergence, in turn, is linked to the branching of trajectories and generative bifurcations, which we characterize as symmetry-breaking phase transitions in the energy landscape. Beyond ensemble averages, we demonstrate that symmetry-breaking decisions are revealed by peaks in the variance of pathwise conditional entropy, capturing heterogeneity in how individual trajectories resolve uncertainty. Together, these results establish generative diffusion as a process of controlled, noise-induced symmetry breaking, in which the score function acts as a dynamic nonlinear filter that regulates both the rate and variability of information flow from noise to data.

Abstract: Extracellular vesicles (EVs) reflect their spatiotemporal cellular origins and, when released into the peripheral bloodstream, represent sources of biomarkers for processes occurring in inaccessible tissues, such as the brain. Positive selection by immunoaffinity isolation targeting a specific EV surface marker can be used to preferentially enrich for EVs from a particular cell type over other EVs in circulation. However, the case of L1CAM immunocapture for enrichment of brain-derived neuronal EVs has exemplified the biological and technical influences affecting the outcomes of positive selection, which are discussed here. These include findings from the wider (non-EV) literature showing that soluble L1CAM is a binding partner of the common EV marker CD9. Additional studies show that CD9 is not expressed by the vast majority of neurons in the brain, but it is highly expressed by neurons of the peripheral nervous system, which could affect the interpretation of studies demonstrating the co-localisation of L1CAM and CD9 on EVs to assert isolation of EVs derived from neurons in the brain. Next, emerging negative selection strategies are discussed which, when combined with positive selection, could overcome some of the challenges associated with enriching for EV-based biomarkers. These strategies include depleting EVs on a cell-by-cell type basis, as well as targeting common EV markers to simultaneously deplete diverse, undesired EV populations whilst selecting for the EVs of interest. An example is given of how CD9 depletion could enrich and select for brain-derived neuronal EVs. Additionally, studies that have used negative selection alone to select for EVs, and the advantages of this approach for biomarker detection, are highlighted. Finally, based on these insights, considerations for the choice of future positive and negative selection targets are discussed, including how understanding the nuances of EV heterogeneity could facilitate the process of EV-based biomarker discovery and their translation to the clinic.

Abstract: The continued scaling of large-scale AI and HPC systems increasingly encounters limits that are not imposed by raw compute capacity, but by the dynamics of interconnects that bind distributed components into a single execution fabric. As system size, heterogeneity, and synchronization demands grow, performance degradation manifests in non-linear and often opaque ways, leading to a collapse of effective cost per performance despite sustained investment in additional hardware. Classical performance and network metrics, while necessary, fail to capture the structural origins of these effects and therefore provide limited guidance for architectural or economic decision-making. This article argues that interconnect-induced instability should not be understood as a collection of incidental faults or implementation bugs, but as an emergent structural property of tightly coupled, large-scale runtime systems. We analyze how latency drift, synchronization loss, and non-local coupling effects propagate through operator dependencies and give rise to hidden economic costs, including re-runs, over-provisioning, and diminished result usability. The contribution of this work is a structural problem analysis that reframes stability as a first-order economic variable rather than a secondary performance artifact. The methodology is deliberately conceptual and analytical, avoiding implementation details or prescriptive solutions. By isolating the structural mechanisms underlying cost-per-performance collapse, this analysis establishes a foundation for structure-oriented approaches to runtime stability control in AI and HPC infrastructures.

Abstract:

Chronic lymphocytic leukemia (CLL) is a dynamic malignancy in which intraclonal subfractions differ in activation history and responsiveness to microenvironmental signals. Here, we investigated the expression and inducibility of IL-12 family receptor subunits (IL-23R, IL-12Rβ1, IL-12Rβ2) and the related receptor complexes in recirculating CLL cells, with a focus on CXCR4/CD5-defined fractions: the proliferative fraction (PF; CXCR4^dim/CD5^bright; most recently divided, tissue-emigrated cells) and the resting fraction (RF; CXCR4^bright/CD5^dim; older, quiescent cells). At baseline, IL-12Rβ1 was enriched in the PF and was associated with a higher proportion of cells expressing IL-23R and IL-12R receptor complexes. Concomitantly, RT-qPCR disclosed higher IL-12Rβ1 mRNA levels. Following antigen-independent activation with CpG or CpG + IL-15, there was a marked increase of IL-23R and IL-12Rβ1 but not of IL-12Rβ2 surface expression, resulting in preferential upregulation of the IL-23R complex over the IL-12R complex. Fraction-specific analyses showed stronger induction of IL-23R and IL-23R complex expression in PF compared with RF. These findings identify an intraclonal bias toward IL-23 responsiveness in the CLL cells with a phenotype of recently divided, tissue-emigrated cells and suggest the IL-23/IL-23R axis as a potential therapeutic target.

Abstract:

Follicular lymphoma (FL) is the second most common form of non-Hodgkin’s lymphoma (NHL) and accounts for about 5% of all hematological malignancies. Despite therapeutic advances, FL follicular lymphoma remains an incurable disease, with frequent relapses and increasingly shorter disease control intervals. Bispecific antibodies (bsAbs) are molecules that target two different epitopes or antigens. The mechanism of action is determined by the molecular targets and structure of the bsAbs. Several bsAbs have already changed the therapeutic landscape of hematological malignancies and some solid tumors. In particular, in this article we review the general principles on follicular lymphoma and established and innovative therapies including bsAbs, in particular the bsAb mosunetuzumab, a new bispecific antibody that acts on CD3 epitopes of T lymphocytes and CD20 epitopes of B lymphocytes with the aim of inducing T lymphocyte-mediated elimination of malignant B lymphocytes, its safety and efficacy with the analysis of no. 3 patients who completed treatment with the drug mosunetuzumab in the A.O. Pia Fondazione di Culto e Religione ‘Card. G. Panico’, Tricase (Lecce).

Abstract: BACKGROUND In Kenya, end-stage renal disease is a significant public health burden treated primarily with hemodialysis in county hospitals, yet comprehensive outcome data from these routine settings are scarce. AIM To evaluate one-year clinical outcomes and identify independent predictors of mortality among ESRD patients undergoing hemodialysis at a Kenyan county hospital. METHODS We conducted a retrospective cohort study of all patients who initiated hemodialysis for ESRD at Murang'a County Referral Hospital between January 2024 and January 2025. Data on demographics, clinical characteristics, comorbidities, and treatment parameters were extracted from hospital electronic medical records and dialysis unit records. Cox proportional hazards regression was used to identify factors associated with one-year mortality. RESULTS Of 79 patients analysed (median age 62.0 years, IQR 48.0-74.0; 65.8% male), the one-year all-cause mortality rate was 34.2% (27/79). The cohort demonstrated a heavy reliance on central venous catheters (89.9%, 71/79) rather than arteriovenous fistulas (10.1%, 8/79). 3 Non-survivors were significantly older (median 73.0 vs 58.0 years, p<0.001) and had lower baseline haemoglobin (7.1 vs 8.6 g/dL, p=0.008). In multivariable analysis, older age (aHR 1.05 per year, 95% CI 1.01-1.09, p=0.012) and central venous catheter use (aHR 3.12, 95% CI 1.08-9.01, p=0.036) remained independent predictors of mortality. Lower eGFR and hemoglobin were significant in univariate analysis but not in the adjusted model. Comorbidities, including HIV and diabetes, did not reach statistical significance. CONCLUSION This study found high one-year mortality in Kenyan hemodialysis patients, with older age and catheter use showing strong associations with death. The near-universal use of CVCs is a marker of systemic challenges in pre-dialysis care, underscoring the urgent need for vascular access programs and improved care strategies to improve survival.

Abstract: Background/Objectives: Driving cessation is associated with adverse health outcomes. Proactive support that extends safe driving while preparing for life after driving cessation has been emphasized, but empirical evidence remains limited. This study examined the effects of a proactive class for older drivers on awareness and behavior related to driving and mobility (Study 1) and on longitudinal changes in on-road driving behavior (Study 2). Methods: The proactive class was implemented as a municipal program, including information provision, training activities, group discussions, and optional on-road driving evaluations. Study 1 included 71 older drivers who attended the class at least five times annually and completed an anonymous questionnaire assessing perceived changes in awareness and behavior. Study 2 included 29 participants who completed standardized on-road driving evaluations at baseline and at a 1-year follow-up. Paired t tests or Wilcoxon signed-rank tests with effect sizes were applied. Results: In Study 1, participants reported increased awareness of safe driving, greater confidence in continuing to drive, heightened risk perception, initiation of health-related behaviors, trial use of public transportation, and increased healthcare utilization, particularly ophthalmology visits. In Study 2, total scores on the on-road driving skill test improved significantly at follow-up (Cohen’s dz = 0.805), with reductions in errors related to braking, vehicle control, and overspeeding. No significant changes were observed in physical, cognitive, or daily functioning, except for a reduction in driving simulator accidents. Conclusions: A proactive, continuous driving transition support class may facilitate multidimensional behavioral change and improve on-road driving performance among older drivers, supporting safer mobility and healthier aging. This study provides an initial conceptual and empirical foundation for proactive driving transition support delivered during the driving continuation phase, which will be examined in a future randomized controlled trial.

Abstract: Membrane curvature is a fundamental biophysical property of cellular membranes that underlies essential processes such as vesicle formation, organelle shaping, intracellular trafficking, and membrane scission. While traditionally studied in the context of cell biology and membrane dynamics, membrane curvature is now emerging as a critical, albeit underrecognized, regulator of oncogenic transformation and tumor progression. Curvature not only governs the mechanical properties of the membrane but also influences the spatial localization and activation of key signaling proteins, including Ras family GTPases, whose oncogenic functions are closely dependent on membrane topology. Cancer, is frequently associated with disruptions in the regulation of membrane curvature as a result of aberrant lipid metabolism, overexpression of curvature-modulating proteins, and cytoskeletal remodeling. These changes facilitate the hallmarks of malignancy such as uncontrolled proliferation, enhanced motility, immune evasion, metabolic rewiring, and therapy resistance. Notably, recent evidence reveals that curvature acts as a spatial cue for Ras activation, particularly during epithelial-to-mesenchymal transition (EMT), where curvature-driven Ras relocalization amplifies growth factor signaling and promotes metastasis. This review provides a comprehensive overview of the molecular determinants that generate and sense membrane curvature from lipid shape and membrane asymmetry, BAR domain proteins, and actin dynamics, and explores how these mechanisms are hijacked in cancer. We describe the feedback between membrane architecture and oncogenic pathways such as Ras/MAPK and PI3K/AKT, emphasizing the role of curvature in shaping signal transduction platforms. Furthermore, we examine how these biophysical alterations impact vesicular trafficking, organelle morphology, and secretion, all of which are co-opted to support tumor development. From a translational standpoint, we assess emerging therapeutic strategies designed to target curvature-regulating factors and leverage membrane topology for precision drug delivery. Innovations in nanomedicine, super-resolution imaging, and curvature-sensing biosensors are also discussed as tools for both diagnostics and therapeutic monitoring. By integrating advances in membrane biophysics, cancer signaling, and bioengineering, this review highlights membrane curvature as a central and actionable dimension of cancer biology.

Abstract: Many complex networks have partially observed or evolving connectivity, making link prediction a fundamental task. Topological link prediction infers missing links using only network topology, with applications in social, biological, and technological systems. The Cannistraci-Hebb (CH) theory provides a topological formulation of Hebbian learning, grounded on two pillars: (1) the minimization of external links within local communities, and (2) the path-based definition of local communities that capture homophilic (similarity-driven) interactions via paths of length 2 and synergetic (diversitydriven) interactions via paths of length 3. Building on this, we introduce the Cannistraci-Hebb Adaptive (CHA) network automata, an adaptive learning machine that automatically selects the optimal CH rule and path length to model each network. CHA unifies theoretical interpretability and data-driven adaptivity, bridging physics-inspired network science and machine intelligence. Across 1,269 networks from 14 domains, CHA consistently surpasses state-of-the-art methods—including SPM, SBM, graph embedding methods, and message-passing graph neural networks—while revealing the mechanistic principles governing link formation. Our code is available at https://github.com/biomedical-cybernetics/Cannistraci_Hebb_network_automata.

Abstract: The surface properties and electrical behavior of carbon-based materials can be effectively tailored by energetic ion irradiation. In this study, graphene oxide (GO), cyclic olefin copolymer foils (COC, Topas 112 and 011, respectively) were irradiated with 1 MeV Au ions using a 3 MV Tandetron accelerator at fluences of 1 × 10¹⁴, 1 × 10¹⁵, and 2.5 × 10¹⁵ ions/cm². The irradiation induced systematic modifications in surface chemistry, morphology, wettability, and electrical properties. Compositional changes before and after irradiation were investigated using Rutherford backscattering spectrometry (RBS) and elastic recoil detection analysis (ERDA), while surface morphology and roughness were characterized by atomic force microscopy (AFM), revealing a clear fluence-dependent evolution of nanoscale topography. The vibrational characteristics will be assessed through Raman spectroscopy. Surface wettability was evaluated by static contact angle measurements, and surface free energy was determined using the Owens–Wendt–Rabel–Kaelble (OWRK) method, showing a consistent decrease in water contact angle and an increase in surface free energy with increasing ion fluence in Topas 112/011 but not in GO. Electrical characterization demonstrated a pronounced fluence-dependent decrease in sheet resistivity across all investigated substrates. The results show that 1 MeV Au-ion irradiation enables controlled modification of both surface and electrical properties of carbon-based foils.

Abstract: In today’s digitally connected world, social media has become central to culture, shaping how we interact, see ourselves, and feel. Platforms like Facebook, Instagram, and TikTok are promoted as ways to connect, but growing evidence shows they can also cause anxiety, social comparison, and emotional strain. Many studies explore these positive and negative effects, but fewer examine changes in academic discussion about social media and well-being over time. To address this issue, the present study employs BERTopic, a dynamic topic model, to analyze 7,254 journal articles indexed in the Web of Science between 2010 and 2025. The analysis identifies 110 distinct research topics and reveals that the most prominent themes converge around anxiety-related outcomes, social connection and support, as well as contextual and methodological developments such as COVID-19 communication and AI-based depression detection. Temporal trend analysis indicates a clear shift in scholarly focus. Research published between 2010 and 2016 adopted a relatively balanced perspective, addressing both the connective potential and the psychological risks associated with social media use. However, since 2017—coinciding with the rapid rise of visually oriented platforms—academic attention has increasingly centered on anxiety-related issues, particularly fear of missing out and body image concerns. By mapping the shift from connection to anxiety focus, the study shows how academic research tracks social change. The results also suggest that future research should explore platform-specific mechanisms, identify protective factors against digital stress, and contribute to the creation of healthier online spaces.

Abstract: We present a high-dimensional quantum key distribution protocol by using N-qudits quantum light states, that is, product states with N photons, each of them in a quantum superposition of dimension d which provides a high dimension d^N and accordingly a very high security. We present the implementation of this protocol in different types of optical fibers where the mentioned states undergo perturbations under propagation in optical fibers; such perturbations can be notably reduced in a passive (autocompensation) or active way and importantly the N-qubits present a great robustness against such optical perturbations. Likewise, quantum states also undergo attenuation, that is, some photons are lost under propagation in the optical fibers and then effective N′ (< N)-qudits are obtained which also are used to generate secret keys. In fact, the detection of states combines standard projective measurements along with photon coincidences. Besides, we analyze the security of this high-dimensional protocol under an intercept and resend attack realized by Eve, and the resulting secure key rates are calculated showing a significative increasing with the dimension provided by the number N of photons.

Abstract: This paper presents a comparative analysis of natural gas and electric power prices using visibility graph methodology, a technique from complex network theory that transforms temporal sequences into network representations. We analyze 1,826 daily observations from the Italian energy market (2019-2023), implementing a three-stage preprocessing pipeline (logarithmic transformation, LOESS detrending, and first differencing) before constructing visibility graphs. Our topological analysis reveals striking differences: gas exhibits substantially higher connectivity (6,202 versus 5,354 edges), heavier-tailed degree distributions (maximum degree 117 versus 54), and dramatically longer-range connections (average temporal distance 26.4 versus 11.0 days). Paradoxically, despite power displaying twice the raw volatility, gas generates more structured long-range correlations due to storage-enabled intertemporal linkages. Both series exhibit small-world properties with high clustering (≈0.76), short path lengths (4.59 and 5.36), and positive assortativity (≈0.17). Correlation analysis reveals moderate contemporaneous return correlation (Pearson r = 0.456) with substantial time variation (range 0.173– 0.696), no lead-lag relationships, and partial synchronization of topological properties. Node-level degree and clustering show positive correlations between markets, while closeness centrality exhibits strong negative correlation (r = −0.719), indicating fundamentally different global network organization. Structural similarity (Jaccard coefficient 0.404) confirms 40% shared visibility connections with 60% commodity-specific structure. These findings demonstrate that physical storability fundamentally shapes temporal correlation structure, with direct implications for risk management, forecasting model selection, and portfolio construction in energy markets.

Abstract: α-Mangostin, a natural product from Garcinia mangostana L, presents most antibacterial activity in plant flavonoids against Staphylococcus aureus so far. Recently, it was reported that the quinone pool is a key target of α-mangostin against Gram-positive bacteria. To further confirm this and investigate the detail of α-mangostin killing S. aureus, the interactions between α-mangostin and a key enzyme as type II NADH:quinone oxidoreductase (NDH-2), together with possible non-enzymatic mechanisms, were explored. Through the enzyme kinetic inhibition experiments, it was found that α-mangostin mainly competes with the menaquinone-binding sites of NDH-2, and the half-maximal inhibitory concentration (IC₅₀) of α-mangostin on NDH-2 is 4.95 μM. Fluorescence analyses indicated that α-mangostin can spontaneously bind to NDH-2 to form an α-mangostin–NDH-2 complex. Subsequently, molecular simulation further indicated that α-mangostin can dock to the menaquinone-binding sites of NDH-2. Another, non-enzymatic mechanism showed that α-mangostin can cause membrane potential depolarization and disrupt the proton motive force balance, thereby promoting the cell-membrane destruction of S. aureus. These results suggest that α-mangostin mainly can interact with the amino acid residues at the menaquinone-binding pocket of NDH-2 to block the electron transfer at the quinone pool in the respiratory chain of S. aureus, and which will hinder the energy supply and promote its incidental effect on membrane disruption, ultimately leading to the death of S. aureus. This once again proves that the quinone pool is a key target of plant flavonoids against Gram-positive bacteria.