Plasma miRNAs and the risk of cancer: Data mining model in lung cancer

Abstract

AbstractData mining(DM) has been widely used in researching the auxiliary diagnosis of cancer. Circulating miRNAs are related to the occurrence and development of various cancer types. For this reason, they have the potential to be used as biomarkers for early tumor diagnosis. Previously, we found that SVM model combined with plasma miRNAs biomarkers could be a method for lung cancer prediction; However, it still has some limitations. So this study further enlarges the analysis to other DM techniques, and explores more accurate methods for auxiliary diagnosis of lung cancer. Univariate analysis showed the differences had statistical significance in the smoking, fever, chest tightness or pain, cough, bloody phlegm, haemoptysis, and 10 plasma miRNAs (miR-21, miR-20a, miR-210, miR-145, miR-126, miR-223, miR-197, miR-30a, miR-30d, and miR-25) between the lung cancer group and normal control group (P < 0.05); Logistic regression analysis showed that fever, chest pain or tightness, cough, miR-21, and miR-223 could be considered as indicators of the presence of cancer (P < 0.05). According to the univariate and multivariate analysis results, two sets of models were constructed using data mining models. The results showed that the sensitivity was 88.6%, the specificity reached 86.7%, the accuracy value was also the highest, and AUC was 0.877 for the GBDT (Gradient Boosting Decision Tree) 16-model, indicating that its predictive effect was the best. Conclusively, this study was to further explore the high accuracy data mining model of lung cancer prediction using plasma miRNAs.