Accurate prediction of posttransplant survival is critical for optimizing liver allocation under organ scarcity. Although the Model for End-Stage Liver Disease (MELD) is widely used for prioritization, it was not designed to estimate posttransplant outcomes, limiting its utility for individualized risk stratification. We developed and temporally validated machine-learning models to predict 12-month survival after liver transplantation using routinely collected donor-recipient variables from a real-world cohort. Sequential temporal validation was used to approximate prospective deployment. Performance was evaluated in terms of discrimination (receiver operating characteristic-area under the curve [ROC-AUC] and precision-recall area under the curve), calibration (Brier score), and decision-analytic evaluation using decision curve analysis. Class imbalance was addressed through cost-sensitive learning, and predicted probabilities were recalibrated with Platt scaling. Across temporal validation windows, discrimination remained modest, with ROC-AUC values below 0.70 and reaching 0.599 in the final window, reflecting the complexity of posttransplant outcomes in heterogeneous clinical populations. Calibration remained relatively stable (Brier score ≈ 0.18). Within the evaluated threshold range, the machine-learning models yielded a positive net benefit and generally exceeded the transformed MELD baseline in decision-analytic comparisons. However, these findings should be interpreted cautiously given the modest discrimination and the exploratory nature of the MELD transformation used for cross-model comparison. Logistic regression provided the most consistent balance of discrimination, calibration, and net benefit. Overall, calibrated models with modest discrimination may still offer complementary risk information. However, further external validation, threshold-specific evaluation, and prospective testing are required before such models can be considered for operational decision support.
The transition toward a circular bioeconomy relies on identifying microorganisms with traits suitable for sustainable industry. This study explored the biotechnological potential of two Papiliotrema laurentii strains isolated from contrasting environments: urban São Paulo (FBU001) and the Atacama Desert (FBU003). The research focused on the safety profile of the strains, environmental stress resistance, and ability to produce photoprotective compounds. Taxonomic identity was confirmed through ribosomal DNA sequencing and MALDI-TOF Biotyper. Crucially, biosafety assessments, including growth temperature limits and the absence of common fungal virulence factors, combined with in vivo infection models, indicated that both strains have low pathogenic potential and are safe for industrial applications. Physiological tests showed that while both strains resist UVC radiation, the Atacama strain possesses additional extremophilic adaptations like osmotic stress tolerance. Both yeasts produced mycosporine-glutaminol (MG), a natural antioxidant and UV filter. Genomic and transcriptional analyses confirmed that the genes responsible for this compound are organized in a cluster (MYC BGC) activated by UV exposure, suggesting that MG production is a conserved trait in the species regardless of its habitat. These findings position these yeasts as robust candidates for sustainable biotechnology. By offering a bio-based alternative to synthetic sunscreen ingredients, this work supports the development of renewable, biodegradable cosmetics within a circular bioeconomy framework. • P. laurentii from diverse geographic origin are avirulent and produce MG under UVR. • P. laurentii from the Atacama Desert are highly resistant to multiple stress factors. • MYC BGC transcription is UVR induced and it is associated with the production of MG.
Abstract Dye-containing effluents remain a concern in wastewater treatment because many synthetic dyes are persistent, chemically stable, and potentially toxic. In parallel, the construction industry, although closely linked to economic and social development, generates large amounts of construction and demolition waste (CDW), which often lacks adequate disposal. Mollusk shells are another locally abundant residue with limited reuse. Converting these abundant residues into low-cost adsorbents offers a promising strategy for wastewater treatment and waste valorization. This study evaluates the efficiency of an innovative hybrid adsorbent composed of CDW and Mytella charruana (MC) shells for removing methylene blue (MB) from aqueous solution. The composite CDW-MC (75:25, m m–1) produced a synergistic effect, increasing the adsorption capacity to approximately 8.0 mg g–1 under initial screening conditions, a gain of about 14% and 37% relative to CDW alone (7.00 mg g–1) and MC alone (5.85 mg g–1), respectively. Under equilibrium conditions, the Toth model estimated a maximum adsorption capacity of 20.83 mg g–1 at 60 °C for the CDW-MC composite. Characterization by scanning electron microscopy (SEM), Fourier transform infrared spectroscopy (FTIR), X-ray diffraction (XRD), X-ray fluorescence (XRF), Brunauer–Emmett–Teller (BET) analysis, and point of zero charge (pHPZC) indicated that the composite had a heterogeneous surface containing carbonate-, silicate-, and metal oxide-related sites. The adsorption process was found to be spontaneous, exothermic, and predominantly physical in nature, with a positive entropic contribution. Regeneration of the saturated adsorbent via the Fenton reaction, optimized through a factorial experimental design, demonstrated that Fe2+ concentration was the decisive factor governing process efficiency, maintaining regeneration performance above 70% over five successive cycles. Overall, these results show that the CDW-MC composite offers a technically feasible and environmentally sustainable route for treating dye-contaminated wastewater while promoting waste valorization, adsorbent reuse, and circular-economy principles.
Institutions face growing regulatory complexity, and traditional Governance, Risk, and Compliance (GRC) approaches are often reactive and error-prone, while the ambiguity of legal language weakens compliance traceability. This study introduces CFR2SBVR, an automated method that transforms natural-language financial regulations into structured vocabularies and rules aligned with the Semantics of Business Vocabulary and Business Rules (SBVR) standard. The method combines Natural Language Processing (NLP) techniques, Large Language Models (LLMs), and mappings to the Financial Industry Business Ontology (FIBO) to extract, classify, and formalize regulatory content from the United States Code of Federal Regulations (CFR). Grounded in the Design Science Research framework, the method presents each stage as a reviewable checkpoint, preserving traceability from the original text to the transformed statements. The evaluation combines semantic similarity metrics (SemScore and LLM-as-a-Judge) with comparison against a manually constructed gold standard. With gold-standard correction applied at each checkpoint to simulate Subject Matter Expert intervention, all stages achieve average scores of 0.85 or higher across element types, while the estimated end-to-end success rate without such intervention is approximately 0.73. A reference-free probe on a regulatory section unseen during development matches the transformation indicators of the main experiment. Unlike earlier deterministic and rule-based approaches, the method is designed to accommodate ambiguity and regulatory change, and all datasets, code, and results are shared for reproducibility. Beyond automation, the study argues that traceable SBVR representations support the justification and contestability of regulatory interpretations, connecting compliance engineering with normative reasoning in artificial intelligence and law.
Modern internal combustion engines must satisfy increasingly strict and often conflicting requirements: high torque demand must be delivered without sacrificing fuel efficiency, while emissions compliance depends on maintaining suitable exhaust thermal conditions for effective aftertreatment operation. These objectives are strongly coupled and highly nonlinear, so improving one metric can deteriorate others, and the feasible region is further constrained by complex actuator interactions and operating-regime variability. To address this challenge, this study applies a suite of recent bio-inspired metaheuristic algorithms to multi-target engine calibration, leveraging their derivative-free global search capability to handle multimodality, nonconvexity, and black-box constraints that limit conventional gradient-based tuning. A high-fidelity surrogate model was first constructed from a public engine dataset using Principal Component Analysis (PCA) and Gaussian Process Regression (GPR) and validated via k-fold cross-validation, enabling fast and accurate prediction of torque, brake thermal efficiency (BTE), and exhaust gas temperature as the fitness function. Five optimizers were then benchmarked in terms of solution quality and convergence behavior in this realistic calibration setting: Meerkat Optimization Algorithm (MOA), Dumbo Octopus Algorithm (DOA), Pufferfish Optimization Algorithm (POA), Hybrid Jellyfish Search-Particle Swarm Optimization (HJSPSO), and Dendritic Growth Optimization (DGO). Finally, an automated DOA-based calibration workflow was used to generate efficiency-oriented control maps. Across all tested operating conditions, the DOA-based calibration maintained percentage errors of up to 6% relative to all reference targets, with the analysis focused on high-efficiency setpoints, as indicated by BTE reference values of at least 30%.
