Machine Learning Model-based Detection of Potential Genetic Markers Associated with the Diagnosis of Small-cell Lung Cancer

Abstract

Background: Small-cell lung cancer (SCLC), which is in the category of intractable cancers, has a low survival rate. It is essential to understand the pathophysiological pathways underlying its development to create powerful treatment alternatives for the disease. Objectives: This study aimed to classify gene expression data from SCLC and normal lung tissue and identify the key genes responsible for SCLC. Methods: This study used microarray expression data obtained from SCLC tissue and normal lung tissue (adjacent tissue) from 18 patients. An Extreme Gradient Boosting (XGBoost) model was established for the classification by five-fold cross-validation. Accuracy (AC), balanced accuracy (BAC), sensitivity (Sens), specificity (Spec), positive predictive value (PPV), negative predictive value (NPV), and F1 scores were utilized for performance assessment. Results: AC, BAC, Sens, Spec, PPV, NPV, and F1 scores from the XGBoost model were 90%, 90%, 80%, 100%, 100%, 83.3%, and 88.9%, respectively. Based on variable importance values from the XGBoost, the HIST1H1E, C12orf56, DSTNP2, ADAMDEC1, and HMGB2 genes can be considered potential biomarkers for SCLC. Conclusion: A machine learning-based prediction method discovered genes that potentially serve as biomarkers for SCLC. After clinical confirmation of the acquired genes in the following medical study, their therapeutic use can be established in clinical practice.