Transcriptomic Analysis of Radio-resistant A549 Cells Identify Stemness Gene Signature to Predict Radiotherapy Response in Lung Adenocarcinoma Patients

Abstract Despite major technical advancements, the prediction of radiotherapy clinical outcome is a major challenge to radiation oncologists due to lack of suitable predictive biomarkers. To address this, a radio-resistant cell line (RR) has been generated from A549 cells (CC) after fractionated doses followed by clonogenic assay. On evaluation these cells showed cancer stem like features. The transcriptomic data of RR versus CC cells were analysed. Out of 658 differentially expressed genes, RNA expression of 60 genes were found to be correlated (p < 0.05) with transcriptomic data of radiotherapy treated lung adenocarcinoma patients obtained from the TCGA database. Binary logistic regression of these 60 selected genes resulted in identification of seven genes (KCNB1, UNC13A, RIMS2, KCNH3, TOX2, SYTL3 and NR3C2) which showed significant (p < 0.05) association with response to radiotherapy. Data was employed to predict radiotherapy response in patients using machine learning algorithms [KNN Cosine]. Algorithms were trained on 80% and tested on 20% of patient’s data. Accuracy was 90% for the model in predicting radiotherapy outcome. When nomogram analysis was performed based on the results of KNN Cosine model, it showed a positive likelihood ratio of 3.45, suggesting potential prognostic nature of this gene signature for radiotherapy outcome in lung cancer patients.