Construction Of Artificial Neural Network Model For Predicting The Efficacy Of First-line FOLFOX Chemotherapy For Metastatic Colorectal Cancer

Objective: To explore and establish an artificial neural network(ANN)model for predicting the efficacy of first-line FOLFOX chemotherapy for metastatic colorectal cancer by analyzing the gene chip of patients with metastatic colorectal cancer treated with FOLFOX.Methods:GSE104645 dataset was downloaded from the GEO database and expression matrix was constructed.The previous chip data(GSE69657)of our department was analyzed and expression matrix was constructed.The R 3.5.1 software with Combat package was used to perform batch effect correction for the expression values of the two sets of matrices.According to the FOLFOX protocol,the efficacy was divided into two groups: the chemo-sensitive group(including CR and PR)and the chemo-resistant group(including SD and PD).The GSE104645 dataset was assigned as the training set,and the GEO2R platform was used to analyze the differences in gene expression between the two groups.P<0.05 and the absolute value of log2 FC >0.33 were used as the thresholds value to screen the drug resistance and sensitive genes of FOLFOX regimen.The online tool STRING(functional protein association networks)was used to perform the GO function enrichment of differential expressed genes,so as to explore the biological processes involving in chemo-resistance related genes.The IBM SPSS 22 was used to construct ANN of the efficacy of the FOLFOX regimen on the GSE104645 dataset.The training set was randomly divided into training samples and test samples at a ratio of 7:3.Two hidden layers were set using the Multilayer Perception(MLP)method.After the model training was stable,the XML format model(named FOLFOXpredict.xml)was exported for subsequent test set replacement verification.After the model was constructed,using the predicted value of the output of the model together with the efficacy data(sensitive group or resistant group)of the FOLFOX program,the receiver operating characteristic curve(ROC curve)was drawn to verify the prediction accuracy of the model.Set our chip dataset as the test set(GSE69657),the constructed GSE69657 expression matrix and clinical efficacy parameters are loaded into IBM SPSS 22 software,and FOLFOX predict model was used to verify the test set by back substitution.With the predicted outcome,the ROC curve was used to evaluate the test results and prediction performance.Results: 1 The chip data of the FOLFOX sensitive group and the FOLFOX drug resistance group in the training set(GSE104645)were compared.GEO2R was used to screen the differentially expressed genes.A total of 2,076 differentially expressed genes were selected,of which 822 genes were up-regulated and 1,254 genes were down-regulated in the chemo-resistance group.The down-regulated genes were sensitive genes.2 GO analysis of the biological processes in which the differentially expressed genes were involved,and revealed that they were mainly involved in the regulation of substance metabolism.Including the primary metabolic process of cell biosynthesis,RNA,nitrogen compounds and macromolecular biosynthesis,and etc.,which further suggesting the important role of substance metabolism regulation process in the drug resistance of FOLFOX regimen.3 It was determined that 37 patients(68.5%)were training samples and 17 patients(31.5%)were test samples.A total of 39 genes were included in the final model construction.This was a neural network model with two hidden layers.The accuracy of predicting training samples and test samples are 75.7% and 76.5% respectively,and the area under the ROC curve was 0.875.4 The chip data set of our department(GSE69657)was set as the test set,and the constructed expression spectrum matrix with batch effect correction was imported into IBM SPSS 22 software.Prediction was made by the joint loading of FOLFOX.mxl,and the monitoring was based on the actual outcome(sensitivity or drug resistance).The results indicate that the prediction efficiency of the model is well,and the area under the ROC curve was 0.778.Conclusions:In this study,an artificial neural network model was successfully constructed to predict the efficacy of first-line FOLFOX regimen for metastatic colorectal cancer based on chip data,and an independent external verification was also conducted.The model has good stability and well prediction efficiency.In addition,the results of this study suggests that the gene functions related to oxaliplatin resistance were mainly enriched in the regulation process of substance metabolism.Including the primary metabolic process of cell biosynthesis,RNA,nitrogen compounds and macromolecular biosynthesis,and etc.,which further suggesting the important role of substance metabolism regulation process in the drug resistance of FOLFOX regimen.