## 3D PCA ### 應用場景 3D PCA和2D PCA本質類似，可能看上去更酷炫些 > 所有數據和2D一樣 ```R # 增加一個pro3 library(gmodels) mat_a <- t(as.matrix(rt)) mat_pca <- fast.prcomp(mat_a, scale = T) # do PCA sum_a <- summary(mat_pca) tmp <- sum_a$importance # a include 4 sections which contain importance pro1 <- as.numeric(sprintf("%.3f", tmp[2,1]))*100 pro2 <- as.numeric(sprintf("%.3f", tmp[2,2]))*100 # fetch the proportion of PC1 and PC2 pro3 <- as.numeric(sprintf("%.3f", tmp[2,3])) pc <- as.data.frame(sum_a$x) # convert to data.frame # 增加一個zlab pc$group <- c(rep('tumor', 57), rep('normal', 57)) pc$color <- c(rep('#6D9EC1', 57), rep('#E46726', 57)) pc$group <- factor(pc$group, levels = unique(pc$group)) xlab <- paste("PC1(", pro1, "%)", sep = "") ylab <- paste("PC2(", pro2, "%)", sep = "") zlab <- paste("PC3(", pro3, "%)", sep = "") ``` > 利用scatter3D作圖 ```R library(plot3D) # 基本展示 scatter3D(x = pc$PC1, y = pc$PC2, z = pc$PC3, xlab = xlab, ylab = ylab, zlab = zlab, colkey = FALSE) # 改變下顏色及分組 scatter3D(x = pc$PC1, y = pc$PC2, z = pc$PC3, xlab = xlab, ylab = ylab, zlab = zlab, bty = 'g', cex = 1.5, pch = 20, col = as.character(unique(pc$color)), colvar = as.integer(pc$group), groups = pc$group, colkey = FALSE) ```