Advancing Open Science

Class imbalance remains a central problem in machine learning because the rare class is often the event of scientific, clinical, financial, or operational interest. Yet the literature is fragmented across sampling, cost-sensitive learning, ensembles, representation learning, modern generative augmentation, and long-tail learning. This structured critical review examines these method families through the joint perspective of empirical class priors and data distributions. It distinguishes interventions that change the relative frequency of the classes from those that alter class-conditional distributions, meaning how feature values are distributed within each class, including minority support, local density, and class overlap. It also examines interventions that affect the information retained during training, the learning objective, the decision boundary, or the calibration of predicted probabilities, together with changes between the training and deployment distributions. The review draws on a documented pool of 470 bibliographic records, complemented by backward citation tracing and targeted searches to address identified gaps. Classical methods including the Synthetic Minority Over-sampling Technique (SMOTE), editing, Balanced Random Forest, EasyEnsemble, and random undersampling boosting (RUSBoost) are connected to focal and class-balanced losses, logit adjustment, decoupled training, generative adversarial network (GAN) /variational autoencoder (VAE)/diffusion synthesis, language-model tabular generation, graphs, federated learning, and domain shift. The synthesis identifies six recurrent failure mechanisms, namely, prior distortion, synthetic overlap, minority-noise amplification, unreliable high-dimensional neighborhoods, miscalibration, and evaluation leakage. The assumptions behind representative methods are presented, while an intervention map and conditional selection matrices translate the evidence into diagnostics, matched baselines, and leakage-free validation requirements. The resulting guidance relates method choice to the class priors observed during training and expected at deployment, the number and class-conditional distribution of minority examples, the degree of class overlap, the label structure, and whether the required output is a ranking, a class decision, or a calibrated probability.

S-adenosylmethionine (SAMe) depletion may represent a rate-limiting nexus linking oxidative/nitrosative stress to epigenetic failure during neurulation. In this review, we synthesize evidence that diverse teratogenic and metabolic insults, including valproic acid (VPA) exposure, maternal diabetes, and folate-dependent one-carbon imbalance, converge on a shared redox–metabolic–epigenetic axis that lowers SAMe availability, elevates S-adenosylhomocysteine, weakens glutathione buffering, and disrupts DNA and histone methylation/demethylation programs of genes required for neural tube closure. This framework explains how redox stress can impair key developmental pathways such as Wnt/β-catenin, sonic hedgehog, and Pax3 while promoting neuroepithelial apoptosis, abnormal proliferation, and failure of morphogenetic coordination. By integrating redox stress with chromatin regulation and one-carbon metabolism, the proposed model provides a unifying mechanism for neural tube defect pathogenesis and positions SAMe-centered metabolic control as a plausible therapeutic target. Definitive testing will require demonstrating that redox stress lowers embryonic SAMe before closure failure, that this precedes epigenetic dysregulation at neurulation genes, and that restoring SAMe rescues neural tube development in vivo. Indeed, our recent experiments showed that SAMe prevented brain oxidative/nitrosative stress and neural tube defects (NTDs) induced by VPA in ICR mouse embryos. Although most studies alongside our hypothesis are experimental and the entire causal pathway in humans remains to be established, increased oxidative stress, reduced SAMe, and reduced SAMe:SAH ratio were also described in human NTDs.

mRNA vaccines have emerged as a promising platform in vaccinology, enabling rapid and flexible antigen design. By delivering mRNA encoding target antigens, they allow in situ protein expression and induction of adaptive immune responses. Advances in sequence optimization, structural engineering, and delivery systems—particularly lipid nanoparticles (LNPs)—have improved mRNA stability, translational efficiency, and immunogenic performance in preclinical and clinical studies. Beyond infectious diseases, mRNA-based approaches are being explored in cancer immunotherapy, autoimmune disease modulation, and personalized medicine. Emerging platforms, including self-amplifying RNA and circular RNA, expand their potential applications. Despite these advances, challenges remain, such as the durability of immune responses, inter-individual variability, delivery specificity, and cold-chain requirements. Ongoing efforts are focused on improving these aspects through optimized delivery systems, novel administration routes, and the integration of artificial intelligence for sequence and formulation design. Notably, this review provides updated coverage of sequence design, manufacturing processes, delivery mechanisms, clinical applications, challenges, and future directions, with particular emphasis on recent advances in bacterial and parasitic mRNA vaccines and the AI-empowered closed-loop design of mRNA vaccines. Overall, mRNA technology represents a rapidly evolving and versatile platform with significant potential for future therapeutic development.

The large-scale integration of offshore wind power increases the flexibility requirements of the integrated power grid. To address this issue, a low-carbon dispatch strategy is proposed for an offshore-wind-integrated electricity–gas–methanol system with variable hydrogen blending in the gas network. A combustion and carbon emission model of the hydrogen-mixed gas turbine (HMGT) is further developed considering the variation in gas composition caused by hydrogen blending. The dispatch framework also integrates carbon emission allowance (CEA) trading with the China Certified Emission Reduction (CCER) mechanism to account for carbon trading costs and CCER revenues from grid-connected offshore wind power. The results show that, compared with the case without the coordinated operating mechanism, the proposed strategy reduces wind curtailment by 189.88 MWh, gross operating cost by 3.49% and carbon emissions by 9.07%. The proposed strategy improves the flexibility, economic performance and low-carbon operation of the integrated electricity–gas–methanol system, thereby supporting the sustainable operation of offshore-wind-integrated energy systems.

We present a generalized electrostatic SWAP gate realized in a chain of two double quantum dots operated in the single-electron regime. Using a minimalist tight-binding model, we derive analytical results and corroborate them with numerical simulations. We exploit the charge anticorrelation arising from Coulomb repulsion and quantify the resulting entanglement generation. We contrast classical and quantum descriptions and show how device geometry and coupling strengths govern entanglement dynamics and gate performance. The results are relevant to cryogenic, CMOS-compatible quantum technologies and suggest a route toward large-scale semiconductor implementations of quantum logic. Finally, we outline a systematic procedure for translating classical electrostatic logic gates into single-electron quantum gates.

Because of the adverse ecological impact of rapid urbanization, coupling the study of urban land use efficiency (ULUE) and ecological carrying capacity (ECC) is vital for harmonizing socioeconomic growth and environmental protection, and promoting regional sustainability. This research focuses on 16 cities in Shandong Province, utilizing SE-DEA, DPSIR models, and entropy-TOPSIS to study the interaction mechanisms of ULUE and ECC by coupling coordination analysis employing models, relative development indices, and geographic detector. During 2000–2020, the results show that ULUE first increased from 1.11 to 1.15 and then declined to 1.07, exhibiting an overall decreasing trend. ECC nearly doubled, rising from 0.25 to 0.59. The coupling coordination degree between ULUE and ECC improved from intermediate coordination (0.72) to good coordination (0.89). By 2020, the system reached a state of coupling coordination, although ECC development lagged behind (K = 0.58 < 0.8). Key driving factors for ULUE, ECC, and their coupling coordination were capital investment scale, resource and environment regulation, and capital investment scale, respectively. This reveals the changes in the interaction mechanism of ULUE and ECC in the three stages, indicating that ULUE and ECC will gradually form imbalance to coordinated development driving by these factors in different stages of urban development.

Electroencephalography (EEG) is a cost-effective and noninvasive modality for identifying neurophysiological alterations associated with Alzheimer’s disease (AD). To robustly extract disease-specific features, this study presents contrastive multiscale EEGNet (CM-EEGNet), a domain-specific deep learning framework that systematically integrates multi-scale temporal convolutions with supervised contrastive learning. By explicitly coupling parallel temporal convolutions of varying kernel sizes—calibrated to capture both spectral slowing (4–12 Hz) and transient dynamics—with a joint contrastive and cross-entropy optimization objective, the model effectively maximizes interclass separability within the latent representation space. Furthermore, we introduce channel prototype attribution (CPA), which causally evaluates channel-wise contributions based on shifts in contrastive evidence scores following channel occlusion. Evaluated under a strictly subject-independent protocol, CM-EEGNet achieved a segment-level accuracy of 96.26% for AD classification, significantly outperforming conventional baselines. Moreover, CPA highlighted the occipital (O2) and parietal (P4) channels as key AD discriminators, where occlusion of the top three channels reduced accuracy to 77.11%. This specialized integration establishes a highly discriminative and interpretable framework for EEG-based diagnosis, demonstrating both robust classification performance and clinical transparency.

Crude petroleum oil (CPO), acting as an abiotic soil factor, significantly influences the chemical properties of tropical riverine mangrove roots; however, there has been a lack of systematic explanation regarding the effects of varying concentrations of CPO on reactive oxygen species synthesis, antioxidant activity, lipid peroxidation, and non-structural carbohydrates in the roots of young and mature mangrove trees, particularly concerning their adaptation to contaminated acidic organic soils. Thus, levels of H₂O₂, enzymatic and non-enzymatic antioxidants, proline, glucose, fructose, sucrose and malondialdehyde (MDA) were studied in roots of young (≤five years) Laguncularia racemosa (Lr1) and Rhizophora mangle (Rm) trees and mature (≥40 years) Laguncularia racemosa (Lr2) trees. The roots of all trees were sampled in soil that was non-toxic (1.2 to 1.6 g/kg) and toxic (6.8 to 6.9 g/kg) with respect to total petroleum hydrocarbons (TPHs), respectively. The roots of Lr2 and Lr1 trees responded to toxic CPO with increased levels of phenolic acids and phenylalanine amino lyase (PAL) activity in addition to increased FRAP (ferric reducing antioxidant power) and total antioxidant levels. A significant increase in MDA levels, indicating oxidative stress conditions in Lr2 exposed to CPO, proved important, despite elevated antioxidant levels; this suggests that the activated antioxidant response fails to completely prevent membrane damage. Remarkably, most chemical indicators of antioxidant activity and carbon mobilization analyzed in this study were significantly depressed in roots of young Rm trees. Thus, Lr2 trees, followed by Lr1, appear to have adapted better than their Rm counterparts to exposure to CPO contaminants, a trait that may provide a competitive advantage in contaminated riverine–lagoon forests.