Search Results (1,074)

To mitigate high expert-annotation costs, domain-preference misalignment, and the inherent trade-off between sensitive-data protection and training utility in petrochemical dataset construction, an iterative framework combining human-feedback-aligned reinforcement learning (RLHF) with post hoc data de-identification is proposed. Direct scoring and pairwise preference feedback are generated using two high-capability language models. A reward model is subsequently trained via a joint Bradley–Terry and mean-squared-error loss, followed by three rounds of closed-loop proximal policy optimization (PPO) constrained by a fixed supervised fine-tuning reference model. Retained high-quality samples are then processed through a four-stage post-RLHF de-identification pipeline. Experimental results demonstrate that the PPO-V3 model achieves a reward score increase of 1.183 over the baseline alongside a 96.2% pairwise win rate, while the sensitivity-aware adaptive differential privacy with context-aware token-level injection (SA-ADP-CTI) post-RLHF de-identification method attains a composite score of 0.9636. The reliability of both the automated feedback and privacy-preservation mechanisms is further validated through blind expert review and manual spot checks. Full article

Coastal plains are low-relief environments where subtle topographic variations govern sedimentary processes, landscape structure, and hydrological dynamics. Along the southeastern Brazilian coast, barrier–lagoon systems record Quaternary coastal evolution and sea-level oscillations; however, detailed mapping of these environments is constrained by the spatial resolution of conventional cartographic bases. This study presents high-resolution morphological mapping of the Maricá coastal plain (Rio de Janeiro State, Brazil) based on digital elevation models (DEMs) derived from UAV photogrammetry, combined with geomorphological interpretation following the IBGE classification framework. The DEM resolved subtle altimetric variations associated with beach ridges and relict lagoonal depressions not represented in conventional topographic maps, providing a surface proxy for the underlying stratigraphic architecture. UAV-derived elevation models significantly improve the identification of morphological units in low-relief coastal environments, establishing a spatial framework for future stratigraphic investigations along the southeastern Brazilian coast. Full article

The Twaite shad (Alosa fallax Lacépède, 1800) is an endangered anadromous fish listed as Vulnerable on the IUCN Red List. In northern Catalonia (NE Iberian Peninsula), the species maintains a small reproducing population in the Fluvià River, while in the Ter River only sparse, irregular records exist and the status of the species has remained essentially unknown. This study aimed to assess the status of Twaite shad in the lower Ter River through a combined approach of scientific rod-and-reel fishing and eDNA analysis, with the specific objectives of confirming adult presence during the spawning season and detecting juveniles as evidence of effective reproduction. Water samples for eDNA analysis were collected at multiple sites along the lower Ter, across five sampling campaigns conducted between May and September 2025. Duplicate samples of mean 32 L each were filtered in situ and amplified using the MiFish 12S marker with twelve PCR replicates per site and sequenced using Illumina high-throughput technology at a 40 M read sequencing depth per sample. Complementary scientific fishing was carried out using a rod-and-reel technique with demonstrated effectiveness for Twaite shad, comprising five spring sessions (40 h of effort) and four autumn sessions (80 h of effort) at key habitat sites. The eDNA approach proved markedly more sensitive than conventional fishing. Notably, high-resolution eDNA confirmed the presence of the species below the Ullà weir, demonstrating its capacity to move upstream past the Torroella de Montgrí gauging station, most likely facilitated by elevated discharge events during spring. A confirmed eDNA detection in late August provides strong evidence of juvenile presence and, consequently, of effective reproduction in the Ter River. Conclusions: This study provides the first evidence that Twaite shad reproduce in the Ter River, confirming the existence of a previously undescribed breeding population. The eDNA approach demonstrated substantially greater sensitivity than conventional fishing methods for this cryptic migratory species. Priority actions include the identification of spawning grounds, the extension of eDNA surveys to reaches upstream of Ullà—now feasible given the current breach in the weir—and the genetic characterisation of this population to determine its origin, diversity, and potential introgression. Full article

Background/Objectives: This study examined whether the high-sensitivity modified Glasgow Prognostic Score (hs-mGPS) and daily weight change by percentile (DWtCP) are associated with overall survival in patients with gastrointestinal (GI) cancers, overall and across BMI categories. Methods: We analyzed de-identified electronic health record data from 176 patients with GI cancers who had albumin, high-sensitivity C-reactive protein (hs-CRP), and longitudinal weight measurements. hs-mGPS was calculated from baseline albumin and hs-CRP. DWtCP, derived from serial weight measurements, was categorized as weight gain, moderate weight loss, or severe weight loss. Survival was evaluated using Kaplan–Meier and Cox proportional hazards models. Results: Nearly 70% of patients had hs-mGPS = 2, which was associated with poorer survival in the earlier part of follow-up (Breslow p = 0.011; log-rank p = 0.062), a pattern also observed in the high-BMI group (Breslow p = 0.045; log-rank p = 0.059). DWtCP was strongly associated with survival (log-rank p < 0.001; Breslow p < 0.001). Patients with moderate weight loss had the most favorable 1200-day Kaplan–Meier survival estimate (54.6%), compared with 5.1% for severe weight loss and 10.9% for weight gain. In multivariable analysis, moderate weight loss was associated with substantially lower mortality risk than weight gain (HR = 0.194, p < 0.001). In BMI-stratified models, DWtCP and primary cancer site remained independent predictors, whereas albumin and hs-CRP did not reach statistical individual significance, supporting the value of the composite hs-mGPS score. Conclusions: In this real-world retrospective cohort, hs-mGPS and DWtCP provided complementary prognostic information. Routine monitoring of systemic inflammation and longitudinal weight trajectories may enable pragmatic risk stratification and early identification of high-risk patients, particularly those with elevated BMI. Full article

This study introduces decahydronaphthalenes (C₁₀- and C₁₁-decalins; C₀De and C₁De) as a novel class of hydrodesulfurization (HDS)-associated forensic markers. Diesel-contaminated soil residues from 12 gas station spills and fresh ultra-low-sulfur diesel (ULSD) reference samples (2018–2025) in Taiwan were evaluated. The decalin abundances, derived from gas chromatography/mass spectrometry (GC/MS) peak areas, were converted into diagnostic ratios (DRs) using a conserved C₁₅ sesquiterpane (BS3) internal reference. These ratios were correlated with conventional fingerprint parameters and alternative HDS products to support source identification and manufacturing-era estimation. Notably, Pearson correlations revealed that C₀De and C₁De weathering behaviors mimic those of short- (CH-6/CH-7) and medium-chain (CH-9) n-alkylcyclohexanes, respectively. Furthermore, C₁De’s negative correlation with C_xD proxies indicates the premise that rigorous HDS processing lowers sulfur concentration while elevating these hydrogenated side products. Expanding on a framework using dimethylbiphenyls (C₂B) and trimethyltetralins (C₃T), adding these novel decalin markers in principal component analysis (PCA) maintained manufacturer-specific source discrimination, with collinear loading vectors confirming C₁De and C₃T as codependent HDS markers. Cross-validated (CV) linear discriminant analysis (LDA) showed that replacing C₃T with C₀De or C₁De improved discrimination among broad sulfur-regulatory manufacturing-era groups, although residual classification uncertainty remained for strongly weathered samples. These results support C₀De and C₁De as complementary HDS-associated markers for semi-quantitative manufacturing-era assessment rather than exact chronological dating. Full article

Background: Early identification of patients at risk of severe acute pancreatitis (AP) remains a clinical priority. We evaluated whether the admission Systemic Immune-Inflammation Index (SII, ×10⁹/L) adds prognostic value to the BISAP score, alone and in combination. SII integrates neutrophil, lymphocyte, and platelet counts from the routine complete blood count. Methods: In this retrospective single-center cohort, 523 adults with AP (January 2020–December 2025) were classified by the Revised Atlanta Classification. Admission SII and BISAP were related to severe AP, intensive care unit (ICU) admission, and in-hospital mortality using logistic regression, ROC analysis with DeLong comparison, 1000-sample bootstrap validation, and decision curve analysis. Results: Severe AP occurred in 48 patients (9.2%), ICU admission in 61 (11.7%), and death in 26 (5.0%). Admission SII was markedly higher in severe than non-severe disease (4085 vs. 1499; p < 0.001). Both SII and BISAP independently predicted all three outcomes after mutual adjustment (severe AP: SII odds ratio 2.12 and BISAP odds ratio 4.92 per standard deviation). For severe AP, the combined BISAP + SII model outperformed either marker (AUC 0.954 vs. 0.911 and 0.917), as it did for mortality (0.957); for ICU admission SII was the strongest single marker (0.869) and the combined model did not significantly improve discrimination compared with SII alone (DeLong p = 0.287). Internal validation showed adequate calibration with modest optimism and suggested potential net clinical benefit across the evaluated threshold ranges, pending external validation. Conclusions: Combining the admission SII with the BISAP score provides incremental prognostic information beyond BISAP alone for the early risk stratification of severe acute pancreatitis and in-hospital mortality using data already available at presentation. External, prospective validation is warranted before routine adoption. Full article

Historical mining districts are long-term sources of environmental contamination, especially where agriculture and grazing occur near mining areas. This study evaluates mercury (Hg) and other potentially toxic elements (PTEs) in soils from Solana de Peñarrubia, within the Almadén mining district (Spain). Thirty-four soil samples collected using a regular staggered grid were analyzed for total Hg and major oxides and selected trace elements. Spatial patterns were evaluated using Kriging interpolation, hierarchical clustering, PCA, and the geoaccumulation index, after standardizing variables to minimize differences in magnitude among them. Thermal desorption and microscopy were performed on three selected high-Hg samples, with thermal assignments interpreted as operational rather than definitive mineralogical identifications. Hg concentrations ranged from 10 to 994 mg kg⁻¹, with the highest values in the northern sector. The PCA separated cropland soils, associated mainly with lithological variables, from soils with anthropogenic inputs related to Hg, Pb, SO₃, and SiO₂. The geoaccumulation index indicated Hg enrichment, reaching the extremely polluted category at the maximum concentration. Thermal desorption profiles were dominated by fractions within the reference range of α-HgS, accounting for 78.27–87.44% of the estimated relative abundance, while microscopic examination revealed particles consistent with cinnabar in high-Hg samples. The Hg anomaly showed no straightforward relationship with the mapped local lithology and may partly reflect, as a working hypothesis, the possible redistribution of Hg-bearing mineralized materials associated with historical mining or metallurgical activities. Full article

Identification of protease cleavage sites is essential for understanding biological regulation and disease mechanisms, yet many predictive approaches rely on annotated substrates and curated databases, limiting performance for poorly characterized proteases. We present PhageScout, a framework for database-independent generation of protease-specific features to predict cleavage sites using de novo experimental substrate phage display screening. We screened a randomized 5-mer phage display library against two neutrophil serine proteases (cathepsin G, elastase). Cleaved peptides generated position weight matrices (PWMs) and peptide enrichment scores to evaluate cleavage-site likelihood across substrate sequences. Sequence-derived scores were integrated with structural features, including accessibility and flexibility, using XGBoost classification models. Performance was benchmarked against annotated cleavage sites from the MEROPS peptidase database as reference data. Phage-derived PWM scores alone captured protease preferences and discriminated cleavage sites from background sites. Without model fitting, PWM scores achieved an area under the curve (AUC) of 0.756 (95%CI: 0.714–0.797) (cathepsin G) and 0.787 (95%CI: 0.753–0.821) (elastase). Combining broad and specific phage-derived scores improved cathepsin G prediction (AUC = 0.783), whereas this improvement was not observed for elastase. Compared to only phage-derived features, XGBoost models integrating phage sequence and structural features provided modest gains for elastase (AUC = 0.775 to 0.806), with phage-derived features ranking among the strongest predictors, but not cathepsin G (AUC = 0.702 to 0.710). Our findings demonstrate that PhageScout can use experimentally derived cleavage signatures to generate protease-specific predictive features and prioritize protease cleavage sites, providing a framework that warrants further validation across diverse proteases and biological contexts. Full article

Large Language Models (LLMs) have been rapidly evolving lately, resulting in the need for strong, explainable models to detect the difference between human-generated and machine-generated articles. Existing approaches which are mostly based on fine-tuned transformers suffer from several drawbacks such as rapid obsolescence, paraphrasing attacks, and lack of interpretability. To improve their ability to detect, this paper proposes a novel paradigm called Human vs. LLM Identification (HLI) which introduces a Retrieval-Augmented Generation (RAG)-inspired evidence-based detection strategy alongside a fine-tuned transformer classifier. Our core model, DeBERTa-Sentinel, is built on top of a fine-tuned Microsoft DeBERTa-v3-small model, which uses a disentangled attention mechanism to better capture subtle syntactic and stylistic deviations characteristic of AI-generated text. We evaluate our framework on a balanced dataset of 43,456 text samples, curated from the OpenGPTText corpus and covering AI-generated and human-authored content across diverse domains including news, education, and creative text. The experimental results show improved performance over the selected baselines, with our framework achieving an accuracy of 97.53%, precision of 95.89%, recall of 99.34%, and ROC-AUC of 99.53%. In addition, explainability is integrated into our framework through Local Interpretable Model-agnostic Explanations (LIME) analysis, providing token-level insight into classification decisions. This study establishes a benchmark for scalable, explainable AI text detection, with implications for academic integrity, content moderation, and combating misinformation. Full article

Background/Objectives: Early identification of complicated appendicitis in children remains challenging. We developed and internally validated laboratory-based machine-learning models for severity stratification using age and routine admission laboratory data. Methods: This retrospective study included 628 children with surgically confirmed appendicitis treated between 2020 and 2024. Complicated appendicitis was defined by operative or pathological evidence of perforation, gangrene, abscess, phlegmon, diffuse peritonitis, or comparable advanced inflammation. Fifteen candidate predictors were evaluated using five prespecified models. Models were tuned in the training set and evaluated once on an isolated test set. Pairwise DeLong comparisons, decision curve analysis, SHAP, and permutation importance were performed. Results: Complicated appendicitis occurred in 93 patients (14.8%). The prespecified primary CatBoost model achieved a ROC AUC of 0.867, a precision-recall AUC of 0.677, a sensitivity of 0.750, a specificity of 0.863, a positive predictive value of 0.488, and an F1-score of 0.592. Formal comparisons did not demonstrate statistically significant AUC superiority over the other algorithms after Holm correction. Exploratory decision curve analysis showed a greater net benefit than treat-all and treat-none strategies across threshold probabilities of 0.06–0.35. ESR, age, CRP, and CRP-derived indices were the most influential model features. Conclusions: Routine laboratory data may provide adjunctive information for severity stratification, but the modest event count, limited positive predictive value, and absence of external validation preclude stand-alone clinical use. Full article

Background/Objectives: CT-based trabecular attenuation in Hounsfield units (HU) is used as a surrogate marker of bone quality, but most approaches rely on a single vertebral level, usually L1, where local abnormalities may limit reliability and availability. This study evaluated whether multilevel lumbar, reference-adjusted HU assessment beyond isolated L1 improves the identification of QCT-defined osteoporosis and the discrimination of prevalent fragility fracture status. Methods: Between 2021 and March 2024, 800 patients referred for evaluation of bone mineral density underwent QCT of the lumbar spine with a calibration phantom. Cancellous attenuation was measured in HU at L1–L3 using manually positioned ellipsoid regions of interest, with predefined adjacent levels (T12, L4) substituted when a target vertebra was unsuitable. Single-level HU, the mean multilevel HU value and the lowest valid HU value were compared with QCT-defined osteoporosis and prevalent fragility fracture status (vertebral or sacral) using ROC analyses and pairwise DeLong testing. Results: Mean multilevel HU correlated closely with mean vBMD (Spearman ρ = 0.988; p < 0.001); as both derive from the same acquisition, the same vertebral bodies and the same phantom, this reflects a strong association between two related measurements rather than validation against an independent standard. Osteoporosis was present in 483 patients (60.4%). Mean multilevel HU was associated with osteoporosis with an AUC of 0.996 (95% CI, 0.993–0.998) at an internally derived cut-off of ≤99.6 HU. In the 699 patients with a valid L1 measurement, in whom a paired comparison was possible, mean multilevel HU exceeded isolated L1-HU (0.992; p = 0.0028). For fracture status, the AUC was 0.975 (0.966–0.983) at ≤80.5 HU, without advantage over L1-HU (p = 0.816). A valid value was obtainable in all 800 patients versus 699 (87.4%) with L1 alone. All patients with a fracture had a vBMD below the osteoporosis threshold, so this endpoint is structurally dependent on the densitometric one. Conclusions: Multilevel lumbar HU assessment provides a practical marker of trabecular bone quality beyond isolated L1. Its principal advantage is measurement availability rather than a clinically meaningful gain in discrimination. The reported thresholds are internally derived, cohort- and protocol-specific, and require external validation. Full article

Background/Objectives: The identification of novel therapeutic targets is imperative to overcome the limitations of current anti-toxoplasmosis treatments. This study aims to investigate the potential of repurposing Pentamidine as an inhibitor against the HMG-Box domain-containing protein (TGARI_247020) in Toxoplasma gondii, a protein hypothesized to be essential for the parasite’s genomic stability. Methods: The study utilized a multi-layered in silico approach. First, the biological essentiality of the target gene was validated by analyzing CRISPR-Cas9-based phenomics data from the ToxoDB database. Second, the structural properties of the HMG-Box domain (ID: A0A139YAG1) were characterized using AlphaFold models. Finally, molecular docking simulations were conducted via the SwissDock server to evaluate the binding affinity and interaction dynamics between Pentamidine and the target protein. Results: Genomic analysis revealed a phenotype score of −1.2, confirming the indispensable role of the TGARI_247020 gene for parasite viability. Structural analysis identified a well-defined binding pocket within the HMG-Box domain. Molecular docking results demonstrated a high binding affinity for Pentamidine, yielding an optimal AC Score of −44.93, supported by a FullFitness value of −1134.13 kcal/mol. The interaction was primarily stabilized by a network of hydrogen bonds and favorable steric fits within the catalytic groove of the protein. Toxoplasmosis is widely classified as a neglected parasitic disease, posing persistent public health challenges and veterinary economic concerns globally. Traditional de novo drug discovery is often hindered by high costs and prolonged timelines, making drug repositioning (repurposing) a highly attractive and cost-effective strategy to identify novel therapeutics from established clinical agents over the past decade. Computer-Aided Drug Design (CADD), particularly Structure-Based Drug Design (SBDD), has provided a robust molecular framework to prioritize candidate drugs against essential parasitic targets. In apicomplexan parasites, high-mobility group box (HMGB) proteins, such as TgHMGB1a, serve as critical nuclear architectural factors that bind to distorted DNA structures and modulate genomic transcription, disrupting these essential DNA–protein interactions, representing a promising, yet under-explored, therapeutic target. The hypothesis for evaluating Pentamidine—an aromatic dicationic diamidine traditionally used in African trypanosomiasis—lies in its established ability to interact with nucleic acids and block critical molecular targets in other protozoa, providing a logical biochemical rationale for testing its potential as a structural inhibitor of the T. gondii HMG-box protein. Conclusions: Our findings provide preliminary in silico evidence that Pentamidine targets the HMG-Box protein, suggesting its potential for drug repurposing. However, due to established clinical limitations of Pentamidine (such as nephrotoxicity and poor blood–brain barrier permeability), further experimental in vitro and in vivo validation is strictly required to evaluate its therapeutic efficacy. Full article

Hymenopteran venoms contain diverse proteins and peptides that shape envenomation, defense, predation, and allergic responses. Honeybee venom from Apis mellifera is well characterized, while molecular resources for the Asian giant hornet Vespa mandarinia remain less curated. We compared protein- and peptide-fraction LC-MS/MS identification datasets from V. mandarinia venom (VM-V) and A. mellifera venom (AM-V), supported by a de novo V. mandarinia transcriptome-derived database. Protein-level identification yielded 197 protein groups in VM-V PRO and 164 protein groups in AM-V PRO. AM-V contained well-recognized honeybee venom components, including phospholipase A2, hyaluronidase, venom acid phosphatase, venom dipeptidyl peptidase 4, melittin precursor, mast cell degranulating peptide precursor, secapin, and allergen Api m 6. VM-V PRO contained transcriptome-supported candidate venom-associated proteins, including venom dipeptidyl peptidase 4-like, hyaluronidase-like, venom allergen 5-like, serine protease-like, apolipophorin-like, and hexamerin-like entries. Peptide-fraction annotation was strongest in AM-V PEP, led by melittin precursor, whereas VM-V PEP remained largely unannotated. The study focuses on identification, annotation, and hypothesis-generating functional summaries, not replicate-level differential abundance. GO, KEGG, and STRING analyses organized functional annotation patterns and prioritized candidate protein groups. These data provide an identification-based comparative venomomics resource for VM-V and AM-V and a foundation for targeted validation of candidate V. mandarinia venom-associated components. Full article

Background/Objectives: Early identification of trauma patients requiring massive transfusion (MT) is essential for timely activation of massive transfusion protocols. Existing models have limitations: the Assessment of Blood Consumption (ABC) score includes penetrating injury, which limits its use where blunt trauma predominates; the Trauma-Associated Severe Hemorrhage (TASH) score requires numerous weighted variables; and the Traumatic Bleeding Severity Score (TBSS) requires blood pressure measured after crystalloid infusion. This study aimed to develop and validate a simplified MT prediction score using only variables available during initial assessment. Methods: We retrospectively analyzed 1462 adult trauma patients (366 MT, 1096 non-MT) treated at a regional trauma center in Korea between January 2020 and December 2025. Candidate predictors were identified by logistic regression, and five candidate models were compared. The final model, designated the Simplified Massive Transfusion (SMT) score, was compared with the ABC, TASH, and Modified TBSS scores in the same complete-case cohort using the area under the receiver operating characteristic curve (AUC) and the paired DeLong test. Results: The SMT score comprises five variables—systolic blood pressure ≤ 110 mmHg, lactate ≥ 4.0 mmol/L, ionized calcium ≤ 1.05 mmol/L, positive FAST, and pelvic fracture—each assigned one point (range, 0–5). In a common cohort of 1352 patients, the SMT score (AUC 0.846, 95% CI 0.822–0.870) performed similarly to the TASH score (0.841; p = 0.668) and significantly better than the Modified TBSS (0.821; p = 0.015) and the ABC score (0.783; p < 0.001). MT probability increased with increasing score, from 2.8% at a score of 0 to 82.9% at a score of 4. The observed rate at a score of 5 was 81.2%, but only 32 patients had this score. Conclusions: The SMT score, based on five readily available variables, showed discriminatory ability comparable to the TASH score and superior to the Modified TBSS and ABC scores. Since none of its components depends on the mechanism of injury, it may be broadly applicable across trauma populations. Further external validation in multicenter cohorts, especially those with a greater prevalence of penetrating trauma, is needed. Full article

Species of the genus Hepatozoon are among the most widespread haemogregarine parasites infecting reptiles; however, information regarding their occurrence in native snakes from Mexico remains scarce. This study investigated the molecular occurrence of Hepatozoon spp. in free-ranging Mexican black-tailed rattlesnakes (Crotalus molossus nigrescens) rescued within the Reserva Ecológica del Pedregal de San Ángel (REPSA), an urban protected area in central Mexico. Opportunistically sampled squirrels were also included to explore the occurrence of these parasites in sympatric mammals. Blood samples from 26 rattlesnakes, six Mexican gray squirrels (Sciurus aureogaster), and three rock squirrels (Otospermophilus variegatus) were screened by PCR targeting a partial fragment of the 18S rRNA gene. Hepatozoon spp. DNA was detected in six rattlesnakes (23.1%) and two Mexican gray squirrels (33.3%), whereas all rock squirrels tested negative. Phylogenetic analyses suggested the presence of two distinct Hepatozoon lineages: one showing phylogenetic affinity with Hepatozoon cf. ophiosauri, detected in all positive rattlesnakes and one squirrel, and another closely related to Hepatozoon griseisciuri, identified in a second squirrel. These findings represent the first molecular detection of Hepatozoon in C. molossus nigrescens, the first record of infection in S. aureogaster, and the first report of a H. griseisciuri-like lineage in Mexico. Although parasite identification was based exclusively on partial 18S rRNA gene sequences, this study expands current knowledge of Hepatozoon diversity in Mexican wildlife and provides important baseline information for future studies integrating molecular, morphological, and ecological approaches to better understand parasite diversity and transmission in protected urban ecosystems. Full article