Development of a prediction model for radiosensitivity using the expression values of genes and long non-coding RNAs

Radiotherapy has become a popular and standard approach for treating cancer patients because it greatly improves patient survival. However, some of the patients receiving radiotherapy suffer from adverse effects and do not obtain survival benefits. This may be attributed to the fact that most radiation treatment plans are designed based on cancer type, without consideration of each individual's radiosensitivity. A model for predicting radiosensitivity would help to address this issue. In this study, the expression levels of both genes and long non-coding RNAs (lncRNAs) were used to build such a prediction model. Analysis of variance and Tukey's honest significant difference tests (P < 0.001) were utilized in immortalized B cells (GSE26835) to identify differentially expressed genes and lncRNAs after irradiation. A total of 41 genes and lncRNAs associated with radiation exposure were revealed by a network analysis algorithm. To develop a predictive model for radiosensitivity, the expression profiles of NCI-60 cell lines along, with their radiation parameters, were analyzed. A genetic algorithm was proposed to identify 20 predictors, and the support vector machine algorithm was used to evaluate their prediction performance. The model was applied to 2 datasets of glioblastoma, The Cancer Genome Atlas and GSE16011, and significantly better survival was observed in patients with greater predicted radiosensitivity.