中国医科大学学报2024，Vol.53Issue(4)：295-301,7.DOI:10.12007/j.issn.0258-4646.2024.04.002
结合铁死亡建立遗传算法优化的反向传播神经网络脓毒症预后模型
Combined machine learning with ferroptosis to establish the prognosis model for sepsis based on backward propagation neural network optimized using genetic algorithm
摘要
Abstract
Objective Combining ferroptosis with machine learning,a backward propagation network model based on a genetic algo-rithm(GA-BPNN)was established to predict the 28-day survival of sepsis patients.Methods Data related to sepsis were downloaded from gene expression omnibus(GEO).A total of 65 genes associated with ferroptosis were downloaded from molecular signatures database(MSigDB).The gene set enrichment analysis(GSEA),gene set variation analysis(GSVA),weighted gene co-expression network analysis(WGCNA),and protein-protein interaction(PPI)network analysis were then used to screen the genes associated with ferroptosis in sepsis patients.The short time series mining analysis(STEM)was used to screen the genes related to the development of sepsis.The intersection of genes related to ferroptosis in sepsis patients was made to obtain the key genes for the prognosis of sepsis.Based on the traditional back propagation network(BPNN),a genetic algorithm(GA)was used to optimize the weights and thresholds to establish a prognostic model of sepsis,namely GA-BPNN.Results GSVA showed a significant difference in the activity of the ferroptosis-related gene set between the sepsis survival group and nonsurviving group.Ninety-seven genes related to ferroptosis in sepsis were screened.A total of 191 genes that were significantly up-regulated or down-regulated in the development of sepsis were identified.A GA-BPNN model for predicting the prog-nosis of four genes in sepsis was established by the intersection of genes related to ferroptosis and differentially expressed genes in sepsis.The mean square error of the training after the iteration in the training set is below 0.05,and the AUC value is 0.98.In order to verify the classification effect of the GA-BPNN model,the GA-BPNN model was compared with the support vector machine(SVM),backpropaga-tion network(BPNN),and random forest(RF)in the verification set.The results show that the AUC value of the GA-BPNN model in the validation set were all 0.92,higher than that of the SVM,RF,and BPNN.Conclusion The four-gene model based on the GA-BPNN is reliable and stable in predicting the prognosis of sepsis patients,which provides a new insight into ferroptosis in sepsis.关键词
脓毒症/铁死亡/生物标志物/预后模型Key words
sepsis/ferroptosis/biomarker/prognostic model分类
生物科学引用本文复制引用
曾媛媛,常莉,池晴佳,封顺,田菲菲..结合铁死亡建立遗传算法优化的反向传播神经网络脓毒症预后模型[J].中国医科大学学报,2024,53(4):295-301,7.基金项目
国家自然科学基金(22174117) （22174117）
