The Value Of Human Plasma MicroRNAs In Predicting The Prognosis Of Lung Cancer

Research ObjectiveUsing the plasma miRNAs related to the prognosis of lung cancer,the data mining model of prognosis prediction was established to accurately predict the prognosis of lung cancer,providing the basis for making accurate diagnosis and treatment plan and improving the quality of life of patients.Subjects and Methods1.Subjects From June 2016 to February 2017,patients with primary lung cancer were initially diagnosed in respiratory department of the First Affiliated Hospital of Zhengzhou University,thoracic surgery of Henan Cancer Hospital,respiratory department of Henan thoracic hospital and thoracic surgery.2.Research methodsThe relative expression levels of 7 kinds of miRNAs(miR-125b,miR-146a,miR-92a,miR-25,miR-195,miR-21 and miR-204)in human peripheral blood plasma were detected by real-time fluorescence quantitative PCR,and the influence of these miRNAs on the prognosis of lung cancer patients was analyzed;the clinical indexes of lung cancer patients were pre ana-lyzed;the prognosis of lung cancer patients were followed up to collect the survival information.3.Statistical analysisSPSS 21.0 statistical software was used to analyze data,survival time was expressed as median and quartile.When comparing the quantitative data,the data that do not conform to the normal distribution are tested by Mann Whitney U test.Kaplan Meier was used to calculate the survival rate,and log rank test was used to analyze the single factor survival.COX proportional risk regression model was used to analyze the statistically significant factors in single factor analysis.Nomogram was drawn by rms package in R language,and COX prognosis model was visualized.C-index was used to evaluate the effect.Using BNN,ANN,LDA,DT,SVM and other data mining technologies,the accurate prognosis prediction model of lung cancer patients was established.Inspection level ?=0.05.All statistical charts are completed by AI,R language and Graplipad Prism 7 software.Results1.Kaplan-Meier single factor survival analysis showed that six miRNAs(miR-125b,miR-146a,miR-92a,miR-25,miR-195 and miR-21)had significant influence on the prognosis of lung cancer.Further multivariate Cox analysis showed that three miRNAs(miR-125b,miR-146a and miR-195)were independent prognostic factors of lung cancer(log-rank P<0.05).2.Multivariate Cox analysis showed that three miRNAs(miR-125b,miR-146a and miR-195)were independent risk factors for lung cancer prognosis.According to the risk scores established by three miRNAs,lung cancer was divided into high-risk and low-risk groups.The survival rate of low-risk group was significantly better than that of high-risk group(log rank P<0.0001).The C-index of Nomogram was 0.843(95%CI:0.78-0.90),indicating that Nomogram had a good predictive power for the-prognosis of lung cancer.3.Five prognostic models of lung cancer were established by means of data mining.The accuracy of BNN model was 65.52%,and the accuracy of AUC was-0.608.The accuracy of ANN model was 68.97%,and the AUC area was 0.674.The accuracy of Fisher's discriminant analysis model was 82.76%,and the AUC area was 0.828.The prediction effect of C5.0 model is good,with the accuracy of 89.66%and the AUC area of 0.900.The prediction effect of SVM model was the best,the accuracy and AUC area were 96.55%and 0.971,respectively.The sensitivity and negative predictive value were 100%.Conclusions1.Three plasma miRNAs(miR-125b,miR-146a and miR-195)were independent risk factors for the prognosis of lung cancer.2.Among the prediction models of lung cancer prognosis based on plasma miRNAs,SVM model has the best prediction effect on lung cancer prognosis.