Background/Objectives: Glioblastoma multiforme (GBM) is an aggressive and heterogeneous brain tumor with poor prognosis, emphasizing the need for reliable molecular biomarkers to improve patient stratification and treatment planning. This study aimed to identify key genes associated with overall survival in GBM by employing and comparing machine learning (ML) and deep learning (DL) approaches using RNA-Seq gene expression data. Methods: RNA-Seq expression and clinical data for primary GBM tumors were obtained from The Cancer Genome Atlas (TCGA). A univariate Cox proportional hazards regression was used to identify survival-associated genes. For survival prediction, ML-based feature selection techniques—RF, GB, SVM-RFE, RF-RFE, and PCA—were used to construct multivariate Cox models. Separately, DeepSurv, a DL-based survival model, was trained using the significant genes from the univariate analysis. Gradient-based importance scoring was applied to determine key genes from the DeepSurv model. Results: Univariate analysis yielded 694 survival-associated genes. The best ML-based Cox model (RF-RFE with 90% training data) achieved a c-index of 0.725. In comparison, DeepSurv demonstrated superior performance with a c-index of 0.822. The top 10 genes were identified from the DeepSurv analysis, including CMTR1, GMPR, and PPY. Kaplan–Meier survival curves confirmed their prognostic significance, and network analysis highlighted their roles in processes such as purine metabolism, RNA processing, and neuroendocrine signaling. Conclusions: This study demonstrates the effectiveness of combining ML and DL models to identify prognostic gene expression biomarkers in GBM, with DeepSurv providing higher predictive accuracy. The findings offer valuable insights into GBM biology and highlight candidate biomarkers for further validation and therapeutic development.
Triple negative breast cancer (TNBC) presents substantial challenges in clinical oncology because of its aggressive characteristics and fewer treatment options relative to other breast cancer subtypes. Understanding the molecular mechanisms underlying TNBC response to therapy-induced senescence-derived small extracellular vesicles (sEVs) is crucial for improving treatment outcomes. In this study, we employed weighted gene co-expression network analysis (WGCNA) and gene ontology enrichment analysis (GOEA) to identify key gene modules and functional pathways associated with TNBC cell response to sEV treatments. Utilizing gene expression data from MDA-MB-231 cells treated with three types of sEVs (No sEV, Cont sEV, TIS sEV), we identified two modules exhibiting significant correlations with treatment conditions. Genes within these modules, such as Sema3g, Hoxc8, Fgf8, Myb, Clusterin, Trim28, and Arf1, were found to play crucial roles in modulating TNBC cell response to sEVs through various molecular mechanisms. GOEA revealed their involvement in biological processes such as mitochondrial function, respiratory processes, membrane transport, and oxidoreduction activities. This study provides insights into the complex molecular mechanisms underlying TNBC response to sEV treatments and lays the groundwork for developing targeted therapies to enhance treatment effectiveness in TNBC.
Parkinson’s disease (PD) is a progressive neurodegenerative disorder for which dopamine transporter single-photon emission computed tomography (DaT-SPECT) is widely used to support clinical diagnosis. Recent convolutional neural network (CNN)-based studies have reported high classification accuracy on DaT-SPECT datasets. However, the relative contributions of network architecture and training protocol design to these results remain insufficiently explored, particularly for small and moderately sized medical imaging datasets. In this study, a training-oriented evaluation of CNNs for PD classification is conducted using two DaT-SPECT datasets derived from the Parkinson’s Progression Markers Initiative (PPMI). First, a previously published experimental setup is faithfully reproduced on a curated dataset of 645 DaT-SPECT images using identical preprocessing procedures, data splits, and model architectures. Under the reproduced experimental setting, previously reported classification accuracies for individual CNN architectures ranged from 93.02% to 95.34%, while an ensemble approach achieved 98.45% accuracy. The same architectures are then evaluated using a unified training protocol incorporating standardized optimization and regularization strategies. Using this protocol, ResNet50 achieves 100% classification accuracy, with all evaluation metrics reaching 1.0, while VGG16, Inception V3, and Xception each achieve an accuracy of 99.22%. On a larger, independently constructed PPMI-derived dataset with higher spatial resolution, previously reported classification accuracies ranged from 93.27% for PD vs. SWEDD to 95.33% for PD vs. control. Using the proposed unified training protocol, the evaluated CNN architectures achieve classification accuracies of 98% for PD vs. SWEDD and 100% for PD vs. control. These results indicate that training protocol design has a stronger influence on DaT-SPECT-based PD classification performance than the specific choice of CNN architecture.