The Effective Gene on Survival Time for Diffuse Large B-cell Lymphoma: A Comparative Study

Background: Extensive biomedical studies have been designed to discover subsets of genes that are associated with survival risks for diseases and construct predictive models using those selected genes. Materials and Methods: In this study, elastic net method was used for gene selection that is competitive with other dimension reduction methods. In this paper, two variable selection methods were presented for dimension reduction of time to event data to predict the survival time with the Cox proportional hazards. First, the Lasso approach was used. Then, an elastic net approach was employed. Moreover, the most effective gene for the two variable selection methods was selected. Finally, the two methods were evaluated by plotting area under the ROC curve over time. Results: Based on our findings the elastic net method identified more number of genes than Lasso method. The most effective gene based on Lasso method was GENE ۱۲۱۲X (P=۰.۰۰۶) and based on the elastic net method was GENE۳۳۲۵X (P=۰.۰۰۴). In addition, the first gene selected by elastic net-Cox method is more effective than the first gene selected by Lasso. Conclusion: This study showed that the elastic net-Cox method outperformed the Lasso-Cox method in terms of predictive power. Moreover, the elastic net selected more important genes which were highly correlated.

کلیدواژه ها:

Survival analysis ، Microarray ، Variable selection ، Lasso ، Elastic net