Fig. 2

Principal component analysis (PCA) and consensus clustering. A Before eliminating the batch effect. B Elimination of batch effect through PCA, data from the three datasets were evenly distributed in the same region. C Gene expression clustering map of two subgroups. The deeper blue rectangle in the subgroup, the stronger the genetic correlation. D Nine classification methods are obtained by consensus clustering analysis. The abscissa represents different groups, and the ordinate indicates consensus clustering score. E Cumulative distribution graph of nine classification methods. F Delta area of nine classification methods