Misurata University (also known as MU) is a public research university in Misurata, Libya. It was founded in 1984. The campus of Misurata University spans approximately 1,500 acres and includes the Misurata Central Hospital (MCH), named one of the Libyan's best hospitals. Misurata University alumni network exceeds 60,000. Misurata University is ranked as one of the top three universities in Libya.
Gastric remnant or duodenal perforation after Roux-en-Y gastric bypass (RYGB) or one-anastomosis gastric bypass (OAGB) is rare and may be missed radiologically. We systematically reviewed PubMed, Scopus, and Web of Science and included 26 reports (34 patients), alongside a video-illustrated late OAGB case. Mean age was 49 years, 73.5
Arabic spam detection presents technical challenges due to linguistic variability, feature sparsity, and the limited availability of interpretable classification systems. Several machine learning models lack intrinsic interpretability, which reduces transparency in decision-making. This study proposes a spam detection pipeline that combines LightGBM with an integrated explainability layer. Interpretability is incorporated directly into the evaluation process by embedding SHAP and LIME to quantify explanation behavior. The system targets characteristics of Arabic text data by operating in a compressed feature space. From an initial 100-dimensional TF-IDF representation, the top 20 statistically stable features are selected for the experimental design. Under this configuration, the model achieves 93.01 % accuracy, 94 % precision, 93 % recall, and an F1-score of 93 %, maintaining balanced precision and recall. Compared to baseline classifiers that exhibit asymmetric precision-recall behavior, the proposed model produces consistent classification performance on Arabic text. The explanation layer is evaluated using a coherence metric that measures rank agreement between intrinsic LightGBM feature importance and attribution scores. The resulting coherence values (SHAP = 0.726, LIME = 0.3806) indicate that SHAP explanations show higher rank agreement with the model's internal feature ranking, while LIME captures more localized variation. The findings demonstrate that predictive performance is maintained under feature reduction and that explanation behavior can be evaluated quantitatively.
Cancer remains a major global health challenge, emphasizing the need for new therapeutic agents. A series of tetrahydrobenzo[b]thiophene derivatives was synthesized from a beta-enaminonitrile precursor and designed to target tubulin, a key protein in mitosis. The purity of compounds obtained was checked and confirmed with HPLC analysis. Antiproliferative activity was assessed against MCF-7 (breast) and HepG2 (liver) cancer cell lines, and cytotoxicity against the non-cancerous WI-38 cell line identified N-phenylamino- and N-benzyl-tetrahydrobenzo[b]thienopyrimidines as the most potent, likely due to their hydrophobic nature. The selectivity index showed the more selective toxicity of these derivatives toward cancer cells rather than normal cells. Molecular docking revealed strong binding affinities and favorable interactions to tubulin protein (PDB ID: 5NM5), predicted to resemble the co-crystallized ligands GTP and Colchicine (LOC). Pharmacokinetic modeling confirmed compliance with Lipinski's, Veber's, and Ghose's rules, supporting favorable drug-likeness and bioavailability. These results point to the synthesized compounds as promising candidates for the development of novel anticancer agents.
Semi-active vehicle suspension systems with integrated energy-harvesting mechanisms offer the potential to improve ride comfort and road holding while recovering energy. However, achieving these objectives simultaneously is challenging due to nonlinear dynamics, hybrid operating modes, and mixed constraints. Many existing approaches treat comfort, stability, and energy harvesting separately or rely on simplified linear models that do not fully capture the behaviour of energy-harvesting suspensions. A neural-network-based multi- objective control framework is proposed for a joint-internal semi-active vehicle suspension systemequipped with an energy harvesting mechanism. A high-idelity multibody model is irst developed in ADAMS to simulate the nonlinearsuspension dynamics and generate training data. A data-driven NARX neural network is then identiied to approximate the nonlinear dynamicsand replace the computationally intensive physical model in closed-loop simulations. The trained neural model is embedded in a neural-network-based model predictive control (NN–MPC) scheme, where sprung-mass acceleration, tire force, and harvested energy are explicitlyshaped in the cost function. Hybrid constraints associated with torque–velocity limits and harvesting-mode transitions are incorporated directlyinto the predictive optimization problem. The controller is evaluated via simulations under standardized stochastic road excitations. Simulation results demonstrate that the proposed control framework effectively mediates the trade-off between ride comfort, road holding, and energy recovery. Compared with a passive suspension, the comfort-oriented coniguration reduces the RMS sprung-mass acceleration byapproximately 40–60
The phytochemical profile and antioxidant capacity of Ruta graveolens L. leaves, collected from Msallata, Libya, are investigated in this study. Four solvents were used to extract the bioactive compounds: petroleum ether, ethanol, water, and chloroform. Moisture, ash, total proteins, total alkaloids, total phenols, total flavonoids, antioxidant activity, and mineral content in the leaves were measured. Ethanol was the extract with the highest extraction efficiency (16.32%) and the highest concentrations of antioxidant activity (6.47 mg/g), total flavonoids (3.77 mg/g), and total phenols (37.65 mg/g). According to phytochemical screening, all extracts lacked saponins but contained proteins, carbohydrates, phenols, flavonoids, alkaloids, coumarins, glycosides, steroids, and terpenes. Copper was the least common element, according to mineral analysis, whereas calcium, sodium, and magnesium were the most abundant. Iron was the most prevalent heavy metal, followed by zinc, with the overall order being Ca > Na > Mg > Fe > Zn > Cu. These results show the abundance of bioactive chemicals and natural antioxidants in R. graveolens L. leaves, demonstrating their potential uses in the production of functional foods and health promotion.
