GRNs represent one of the most significant models to explain biological processes and decipher pathogenicity. However, the current methodologies largely separate network topology or gene expression, thus ignoring the combinative interaction that forms the basis of network structure and regulation processes. We present an in-depth framework in this study whereby network topology analysis is combined with machine learning to detect hub genes and potential biomarkers in lung adenocarcinoma. By using the correlation-based inference algorithms with the data of RNA-sequencing of tumorous and normal tissue samples, we assembled the gene regulatory networks, and then obtain four traditional network centrality measures degree, betweenness, closeness, and eigenvector centrality to be combined with the gene expression properties. Three machine learning models, namely, Random Forest (RF), Support Vector Machine (SVM), and XGBoost, have been trained using stratified five-fold cross-validation with a large-scale hyperparameter evolution. The fifteen hub genes that were found show significant topological changes in cancerous samples. The XGBoost model was shown to be the best in terms of classifying the data as indicated by the area under the ROC curve of 0.912(95%, CI: 0.891 - 0.933), exhibits superiority to expression only models (p < 0.001, DeLong test). The five hub genes namely TP53, MYC, EGFR, CDKN2A and PTEN are characterized by resounding relationships with patient survival. (log-rank p < 0.001) and were statistically significant following multivariate adjustment of the clinically relevant covariates. Combination of network topology and expression profiles boosts the identification efficacy of biomarkers, which results in a 6.2% improvement in the AUC compared to methods that use only expression. Our framework hence provides a generalizable approach that can be used to precision oncology initiatives.

Globally, lung cancer continues to rank among the leading causes of mortality. The lack of sophisticated technology choices for screening and the fact that the sickness is typically discovered at the final stage are the main causes of this. Consequently, this work creates a hybrid AI-based system that combines machine learning (ML) and deep learning (DL) methods to make lung cancer detection quick and accurate. A machine learning classifier such as Random Forest or Support Vector Machine (SVM) is used for the final diagnosis after the created system uses convolutional neural networks (CNN) for automatic feature extraction from computed tomography (CT) images. Such a hybridization allows the model to leverage the deep networks' superior feature extraction ability and also, maintain the interpretability and the efficiency of traditional ML algorithms. The proposed model through rigorous experiments on benchmark datasets is reported to achieve classification accuracy, sensitivity, and specificity of 97.8%, 96.5%, and 98.2%, respectively, thus, outperforming the conventional standalone methods. Besides, the designed scheme is perfect in differentiating malignant and benign lung nodules under various imaging conditions.