Figure 1.
We identified a glioma core survival module based on sample-matched miRNA, mRNA expression and pathways structure. First, we utilized Kaplan-Meier (K-M) survival analysis to identify glioma survival related miRNAs and genes. Then, we integrated these miRNAs and genes to further identify KEGG pathways. Finally, we developed a pathway-based random walk method to identify glioma core survival genes from each pathway, and constructed a glioma core miRNA-gene survival module.
Table 1.
Clinicopathological characteristics of patients in the training set, the testing set, and entire patient set.
Figure 2.
The glioma core survival module and distribution of miRNAs and genes within the module.
(A). The triangles and rectangles in the core survival module correspond to miRNAs and genes, respectively. MiRNA node size is proportional to the degree of the node. Gene nodes are colored according to their categories, which include 13 pathway classes from KEGG pathway database. (B). Distribution of survival related miRNAs with respect to the number of their regulatory genes. (C). Distribution of core survival genes with respect to the number of times the gene is regulated by miRNAs.
Table 2.
The detailed composition information of four sub-modules in the glioma core survival module.
Figure 3.
Module signatures predict glioma patient clinical outcome.
(A-D). Four sub-modules (moduleS1-S4). K-mean clustering representation of module signatures in the 80 glioma patients of the testing set. The columns represent tumor samples and rows represent genes and miRNAs in corresponding module. Red indicates high relative expression levels, whereas green low levels. Horizontal bars above the heat map indicate the grade status and class of the patient (cyan, deepblue and red box indicated grade II, III and IV; green and darkred box indicated low-risk and high-risk class). The low-risk and high-risk groups were derived from K-mean clustering (K = 2) and estimated by Kaplan-Meier survival analysis. P-values were calculated by the log-rank test.
Table 3.
The P-value performance of four sub-modules using Kaplan-Meier survival analysis in the testing set.
Figure 4.
The gene signature from moduleS3 predicts clinical outcome of glioma datasets from GEO.
The glioma datasets were respectively extracted from the studies of (A) Freije et al., (B) Phillips et al., (C) Murat et al., and (D) Lee et al.. The significance of clinical outcome difference between the low-risk and high-risk groups was estimated by K-M survival analysis. P-values were calculated by the log-rank test.
Figure 5.
The survival prediction performance of the 9-gene signature from moduleS3.
(A). The identification of top 9 genes from the moduleS3. (B). The 9 genes in moduleS3. Gene nodes were colored according to the gene class colors used in Figure 2. (C). The survival prediction power comparison between 9-gene and 26-gene signatures.