pheatmap是繪制熱圖的經(jīng)典R包视搏。其中一些細(xì)節(jié)參數(shù)設(shè)置趋艘，之前每次遇到都是網(wǎng)上搜索。這次系統(tǒng)整理下常用用法凶朗，為以后繪圖提供方便瓷胧。

0、示例數(shù)據(jù)與R包加載
1棚愤、聚類相關(guān)參數(shù)
2搓萧、熱圖的顏色
3、行宛畦，列的注釋
4瘸洛、熱圖的格子相關(guān)
5、行名與列名的調(diào)整
6次和、熱圖的分割
7反肋、轉(zhuǎn)為ggplot2對象
8、按行按列歸一化

0踏施、示例數(shù)據(jù)與R包加載

（1）模擬示例數(shù)據(jù)

• 假設(shè)有兩組樣本石蔗，每組5個的30個基因的表達(dá)數(shù)據(jù)（15個上調(diào)，15個下調(diào)）

exp = matrix(rnorm(300), nrow = 30, ncol = 10)
exp[1:15, 1:5] = exp[1:15, 1:5] + matrix(rnorm(75,mean = 4), nrow = 15, ncol = 5)
exp[16:30, 6:10] = exp[16:30, 6:10] + matrix(rnorm(75,mean = 3), nrow = 15, ncol = 5)
exp = round(exp, 2)
colnames(exp) = paste("Sample", 1:10, sep = "")
rownames(exp) = paste("Gene", 1:30, sep = "")
head(exp)
#       Sample1 Sample2 Sample3 Sample4 Sample5 Sample6 Sample7 Sample8 Sample9 Sample10
# Gene1    0.89    5.00    2.37    4.59    2.71   -0.89    0.21   -0.98   -0.41     0.19
# Gene2    4.05    5.58    3.51    5.45    5.47   -0.77    2.34   -0.67    0.83    -3.54
# Gene3    5.39    6.53    5.25    4.08    5.76    0.98   -0.56   -0.37   -0.56     1.21
# Gene4    4.16    2.29    2.58    3.87    5.17    0.92   -0.80    1.00   -1.23     0.29
# Gene5    3.66    2.30    2.54    2.90    5.37   -0.29    2.29    0.10   -0.33     0.81
# Gene6    2.07    4.19    3.87    1.82    4.82   -0.49    0.37    1.48    0.01    -0.22

（2）加載R包

# install.packages("pheatmap")
library(pheatmap)

packageVersion("pheatmap")
# [1] ‘1.0.12’

（3）基礎(chǔ)繪圖

pheatmap(exp)

1畅形、聚類相關(guān)參數(shù)

如上圖默認(rèn)會分別對行养距、列計算兩兩間的距離，再進(jìn)行聚類

1.1 聚類算法

• 對于兩兩間距離計算參數(shù)：clustering_distance_rows=日熬，clustering_distance_cols=棍厌。默認(rèn)為"euclidean"，備選方法有"correlation"
• 聚類方法的參數(shù)：clustering_method=。默認(rèn)為"average"耘纱，備選方法有"ward.D", "ward.D2", "single", "complete", "mcquitty" (= WPGMA), "median" (= WPGMC) or "centroid" (= UPGMC)

1.2 不聚類

pheatmap(exp, cluster_row = FALSE)

1.3 聚類但不想顯示樹

pheatmap(exp, treeheight_row = 0)

其實treeheight_row參數(shù)是用來調(diào)整樹的顯示尺寸的敬肚；設(shè)置為0，也就是不顯示樹了束析。

1.4 提取熱圖的表達(dá)矩陣

由于聚類會調(diào)整原始數(shù)據(jù)的行列順序帘皿，如果想要獲得熱圖里的行列順序數(shù)據(jù)，可如下調(diào)整

ph = pheatmap(exp)
ph$tree_row$order
ph$tree_col$order
ph_exp = exp[ph$tree_row$order, ph$tree_col$order]
ph_exp[1:4,1:4]
#         Sample3 Sample2 Sample4 Sample1
# Gene2     3.51    5.58    5.45    4.05
# Gene12    6.12    5.19    4.04    3.73
# Gene3     5.25    6.53    4.08    5.39
# Gene10    5.36    4.47    4.42    4.54

2畸陡、熱圖的顏色

• color=

#Default
colours = colorRampPalette(rev(RColorBrewer::brewer.pal(n = 7, name = "RdYlBu")))(100)
str(colours)
# chr [1:100] "#4575B4" "#4979B6" "#4E7DB8" "#5282BB" "#5786BD" "#5C8BBF" "#608FC2" ...

# 個性化修改
colours = colorRampPalette(c("navy", "white", "firebrick3"))(10)
str(colours)
# chr [1:10] "#3288BD" "#5FA2CB" "#8DBCDA" "#BAD7E9" "#E8F1F7" "#FAE9EB" "#F1BEC4" ...
pheatmap(exp, color = colours)

#colours = colorRampPalette(c("#3288bd", "white", "#d53e4f"))(10)

3鹰溜、行，列的注釋

• 為annotation_col=丁恭，annotation_row=參數(shù)提供data.frame曹动，行名與表達(dá)矩陣的行名與列名相同；列的內(nèi)容為分組信息牲览。

# 構(gòu)建列注釋信息（行名與表達(dá)矩陣的列名col保持一致）
annotation_col = data.frame(
  group = rep(c("Group_A", "Group_B"), each = 5),
  row.names = colnames(exp))
head(annotation_col)

# 構(gòu)建行注釋信息（行名與表達(dá)矩陣的行名row保持一致）
annotation_row = data.frame(
  Type = rep(c("Up", "Down"), each = 15),
  row.names = rownames(exp))
head(annotation_row)

pheatmap(exp, 
         annotation_col = annotation_col,
         annotation_row = annotation_row)

image.png

• 為ann_colors=參數(shù)提供一個list對象墓陈，可修改注釋分組的顏色

#修改注釋標(biāo)簽的顏色
ann_colors = list(
  group = c(Group_A = "#e66101", Group_B = "#5e3c99"),
  Type = c(Up = "#e7298a", Down = "#66a61e"))
pheatmap(exp, 
         annotation_col = annotation_col, 
         annotation_row = annotation_row, 
         annotation_colors = ann_colors)

4、熱圖的格子相關(guān)

• 格子的邊框顏色border_color=第献；無邊框border=F
• 格子的長寬cellwidth = 15, cellheight = 12

pheatmap(exp, border_color = "white",
         cellwidth = 9, cellheight = 9)

• 格子內(nèi)的文本注釋

pheatmap(exp, display_numbers = TRUE,
         number_color = "blue", number_format = "%.1e") #default "%.2f"

pheatmap(exp, display_numbers = matrix(ifelse(exp > 5, "*", ""), 
                                        nrow(exp)))

pheatmap(exp, display_numbers = matrix(ifelse(exp > 5, exp, ""), 
                                       nrow(exp)))

5贡必、行名與列名的調(diào)整

• （1）不顯示行名與列名

pheatmap(exp,show_rownames=F,show_colnames=F)

• （2）僅特定顯示部分行名
  注意：由于聚類的原因，一定要注意對應(yīng)的順序庸毫！

labels_row = rep("", nrow(exp))
labels_row[c(5,8,16)]=c("gene_A","gene_B","gene_C")
pheatmap(exp, labels_row = labels_row)

• fontsize_row = 12,fontsize_col = 8可修改行名與列名的大凶心狻；
• angle_col = 45可修改列名的角度飒赃；但只能是修改行名利花，并且只能是270”, “0”, “45”, “90”, “315”中的一種

6、熱圖的分割

• （1）根據(jù)聚類樹载佳，按指定數(shù)目進(jìn)行分割

pheatmap(exp, 
         cutree_cols = 2,
         cutree_rows = 4)

• （2）指定行/列數(shù)進(jìn)行分割炒事，前提是要取消聚類

pheatmap(exp, cluster_rows = FALSE,cluster_cols = FALSE,
         gaps_row = c(10, 20),
         gaps_col = 5)

7、轉(zhuǎn)為ggplot2對象

library(ggplot2)
library(ggplotify)
g = as.ggplot(pheatmap(exp))
g + ggtitle("This is a ggplot object")

8蔫慧、按行按列歸一化

• 默認(rèn)按照原始值可視化挠乳，可通過scale =參數(shù)設(shè)置按照行或者列進(jìn)行歸一化之后的結(jié)果繪圖

pheatmap(exp, scale = "row")
# scale_rows = function(x){
#   m = apply(x, 1, mean, na.rm = T)
#   s = apply(x, 1, sd, na.rm = T)
#   return((x - m) / s)
# }