Skip to main content
Fig. 5 | BMC Cancer

Fig. 5

From: Prediction of radiosensitivity and radiocurability using a novel supervised artificial neural network

Fig. 5

Trained ANN-SCGP using a priori gene interaction matrix. A Gene interaction matrix from STRING database. B Illustration of autoencoder. C SCM identified by autoencoder. D & E Two connected group genes in gene interaction matrix showed distinct connected patterns in SCM. F Interlinked genes had higher Kappa coefficients in SCM than unlinked genes. G Nodes of Gene interaction matrix and SCM had similar closeness. H & I ANN-SCGP model predicted OS in TCGA LUAD (training) and GSE68465 (testing). J Occlusion test suggested that IL18 had highest contribution. IL18 was an unfavorable prognostic factor in TCGA LUAD K and GSE68465 L. M High IL18 was linked to advanced patients. RT, radiotherapy; SCM, selectively connected matrix

Back to article page