×
The submission system is temporarily under maintenance. Please send your manuscripts to
Go to Editorial ManagerThe expansion of chemical and pharmaceutical data increases the need for enhanced computational systems that can interpret molecule structures and forecast chemical properties. Due to the fact that molecular visualization or prediction of chemical properties are presently performed separately during the chemical research and initial drug development processes, chemical research and the initial drug development process are not as efficient as possible. As such, the research presents a fully integrated model that predicts chemical properties with a single input method (SMILES strings) for both 2D and 3D representations of molecular structure, while at the same time utilizing SMILES strings to reconstruct the molecule and extract the features from the molecule. The research develops physicochemical descriptors and circular (Morgan) fingerprints which scientists use to create features for training ensemble-based machine learning models that predict essential molecular properties. The framework enables users to create interactive two-dimensional and three-dimensional molecular visualizations which show the molecular structure in an intuitive way while displaying the results of quantitative prediction methods. The proposed method enhances usability and analytical efficiency through its unified visualization system for prediction and interpretation which outperforms current standalone tools. The Python prediction and visualization pipeline with MATLAB verification tool enables independent assessment of three-dimensional molecular geometries to confirm spatial structure representation and conformer generation while maintaining model training and inference integrity. The framework enables organizations to incorporate new datasets and molecular descriptors and learning algorithms through its modular scalable and extensible design which demonstrates the successful application of cheminformatics-based representations and machine learning to drive data-based molecular research and decision-making in chemical and pharmaceutical fields.
Predicting the risk of operative delivery, especially Cesarean section, is important for improving outcomes for both mothers and newborns. In this study, a machine-learning-based framework is developed using explainable artificial intelligence techniques across three clinical datasets: fetal cardiotocography, maternal health parameters, and pregnancy outcome data.Three classification models—Logistic Regression, Decision Tree, and Random Forest—were evaluated to identify the most effective approach. Among these, the Random Forest model achieved the best performance, with an F1 Score of 0.88 and an Area Under the Curve of 0.99 on the fetal health dataset.To enhance interpretability in a clinical context, a dual-layer explanation strategy was adopted. SHapley Additive exPlanations was used to analyze global feature importance, while Local Interpretable Model-agnostic Explanations was applied to explain individual predictions. The results indicate that abnormal short-term variability is a key factor influencing operative delivery risk. The explanations generated by both methods were consistent with established clinical understanding, making the model both accurate and interpretable.
In precision agriculture, crop disease detection can be a highly valuable undertaking in which scalable and correct solutions may save considerable amounts of money and loss of yield. This paper introduces a comparative analysis of state-of-the-art deep learning models with special attention to EfficientNetB3 hybrids, which are trained on a balanced subsample of the PlantVillage dataset with 33 classes based on nine crops. To overcome the shortcomings of the previous studies, which used unbalanced sample, a leakage-free balancing approach was used, resulting in 13,200 training and 3,300 validation samples. Custom head transfer learning was used where it was tested using two strategies; FreezeUnfreeze fine-tuning, and Singlephase training. MobileNetV2, InceptionV3, DenseNet121, GhostNet, in addition to other baseline CNNs, were compared to baseline Convolutional Neural Networks (CNNs). The findings indicate that EfficientNetB3 hybrids are superior with an accuracy of ≥99.5% and 99.9% Area Under the Curve (AUC) and specificity than the previous CNN-based systems. The paper logically defines a performance ladder between model options and real-life deployment demands, such as lightweight mobile applications to precision agriculture systems, and points out future trends in the field-based validation.