Incorporating elastic-net in survival regression for highly dimensional gene expression datasets

High-dimensional gene expression Cox regression involves using gene expression data, which can have thousands of features (genes), to predict survival outcomes. This type of analysis is often used in bioinformatics and computational biology to identify genes associated with survival times in patients, for example, those with cancer. The analysis of high-dimensional gene expression data poses significant challenges for traditional survival regression models due to the curse of dimensionality and multicollinearity among gene expressions. The Cox elastic-net model (Coxnet) is an advanced approach that combines the Cox proportional hazards model with the elastic-net regularization technique, enhancing survival analysis in the presence of high-dimensional covariate data. Elastic-net regularization addresses these challenges by incorporating both L1 (Lasso) and L2 (Ridge) penalties, promoting sparsity and ensuring stability in variable selection. Our study provides a thorough application of Coxnet on publicly available lung adenocarcinoma dataset, incorporating the concordance index and cross-validated survival curves for performance evaluation. This approach shows the reduced number of covariates in Cox regression not only contribute better performance but also parsimony regression. The results demonstrate that the elastic-net regularized Cox model outperforms traditional methods in terms of predictive accuracy and interpretability, identifying key gene signatures associated with survival. The Coxnet model improves the prognosis for lung adenocarcinoma patients by increasing the C-index value and handling high-dimensional gene expression data. Cox regression with elastic-net regularization was found to be the best model for predicting survival. Key influencing factors include clinical stage (hazard ratio 1.603) and gene expressions: RFTN1 (0.788), PNP (1.270), TMSB4X (0.988), and PRKACB (0.899). The genes were categorized into those accelerating or slowing patient mortality.