Fig 1.
Analysis workflow of the study.
The microarray data were manually curated from four studies (GSE113740, GSE112264, GSE106817, GSE113486) and combined for miRNA selection. The selected miRNAs were then used to classify and validate cancer subjects. miRNA-mRNA interaction network, protein-protein clustering analysis, and KEGG analysis were performed.
Table 1.
Clinical summary for cancer samples.
Fig 2.
Heatmap of the expression value of the top 5 miRNAs selected from highest frequency miRNAs in 100 random forest models.
The X-axis represents the samples, and the Y-axis represents the miRNAs. Each of the boxes represents the normalized expression value of each miRNA in the corresponding sample.
Table 2.
Frequency of top miRNAs in 100 random forest models.
Fig 3.
ROC and AUC analysis of the top 5 selected miRNAs and the 4 miRNA combination ROC and AUC values.
Panel A is the analysis for the discovery set. Panel B is the analysis for the Validation Set. Both panels achieved the highest ROC and AUC value using 4 miRNAs: has-miR-663a, has-miR-6802, has-miR-3184-5p, and hsa-miR-8073.
Table 3.
Classification statistics of selected miRNAs.
Fig 4.
miRNA-mRNA interaction network for the selected 5 miRNAs.
The blue squares represent the miRNAs. The purple and yellow circles represent the mRNAs. The yellow circles represent mRNAs directly associated with cancer, with the bigger yellow circles indicating that the mRNA is more associated with the selected 5 miRNAs. The edge between two nodes indicates their interaction.
Table 4.
KEGG analysis using mRNAs associated with the 5 selected miRNAs.
Fig 5.
Cluster analysis of the mRNA presented in the miRNA-mRNA interaction network.
Top clusters with MCODE value >5 from cytoscape were chosen and the clusters along with their interactions with the 5 selected miRNAs were shown. The miRNAs are highlighted in yellow, and the mRNAs are in blue.