Figure 5
The correlation analysis of gene expression profiles between misjudged tumors and tumors in truth and predicted classes. (A) The heat map showed the unsupervised clustering analysis of gene expression in LUAD, LUSC and HNSC tumor samples. The intensity of gene expression for 6261 DEGs are presented in red color and blue color. To the top panels, three colors indicate LUAD (light yellow), LUSC (gold color) and HNSC (peach color) respectively. And the frequency of error identification is presented in grey (≥ 1 misjudgment) and purple color. In the four scatter plots (B–E), each dot represents the each misjudged tumor sample, and the orange dots means their frequencies of error identification ≥ 50% among the number of total prediction times, otherwise the ones were presented with blue dots. The plots illustrated the averaged gene expression correlation (i.e., Pearson’s r) between the each misjudged tumor of (B) LUAD, (C, D) LUSC and (E) HNSC to every tumor in truth class cancers and in predicted class cancers. If the dots are below the diagonal that indicate the gene expression profiles of misjudged tumors are more similar to tumors in predicted classes than in truth classes.
