目标是使用ggplot2绘制出类似maftools中somaticInteractions()模块产生的三角热图。 maftools产出如下：
maftools这副三角热图用于展示共突变基因的相关度。首先构建出一个文本文件，格式类似下面这个表格，这里使用的是fisher精确检验，绘制热图时，主要使用的是P值，但同时会根据OR值是否大于1来判断是正相关还是负相关。
x y value OR
PMS2 TP53 0.0036828676433501074 0.2
PMS2 TET2 0.4199168731045784 0.6547619047619048
PMS2 SH2B3 0.5570301431667071 1.7105263157894737
PMS2 SETD2 0.5895552875452755 2.4922279792746114
PMS2 SETBP1 1.0 1.2195121951219512
PMS2 PMS2 NA 1
SETBP1 TP53 1.0 0.8695652173913043
SETBP1 TET2 0.5199504039165167 2.074074074074074
SETBP1 SH2B3 0.7342468809923415 1.7142857142857142
SETBP1 SETD2 NA 1
SETBP1 SETBP1 NA 1
SETD2 TP53 1.0 0.5467980295566502
SETD2 TET2 NA 1
SETD2 SH2B3 NA 1
SETD2 SETD2 NA 1
SH2B3 TP53 0.4397247870116061 0.6043956043956044
SH2B3 TET2 0.7364614903174054 1.8313253012048192
SH2B3 SH2B3 NA 1
TET2 TP53 0.23820858847966137 0.4230769230769231
TET2 TET2 NA 1
TP53 TP53 NA 1
注意，因为是三角热图，只画半边，最好在处理数据时，绘图前，就对基因进行排序。下面是ggplot2绘图脚本。
library(ggplot2)
# 读入数据
data <- read.csv("test.txt", sep="\t", header=T)
data_log <- data
data_log$value <- log10(data_log$value) * -1
data_log <- as.data.frame(data_log)
data_log$value <- ifelse(data_log$OR < 1, data_log$value * -1, data_log$value)
# 绘图
ggplot(data = data_log, aes(x = x, y = y)) +
geom_tile(aes(fill = value)) +
geom_text(data = subset(data, value < 0.05), aes(label = "*"), color = "black") +
geom_text(data = subset(data, value < 0.1 & value >= 0.05), aes(label = "·"), color = "black") +
coord_equal(clip = "off") +
xlab("") + ylab("") +
theme(panel.background = element_blank(),
axis.text.x = element_text(angle = 90, hjust = 0.2)) +
scale_fill_gradientn(colors = c("#8B6914", "#F5F5F5", "#00868B"),
na.value = "#FFFFFF",
limits = c(-3, 3),
oob = scales::squish,
labels = c(">3(Mutually exclusive)", "2", "1", "0", "1", "2", ">3(Co-occurence)"),
name = "-log10(p_value)") +
scale_x_discrete(position = "top") +
geom_text(x = Inf, y = 5, label = "* P < 0.05", hjust = 0.2, vjust = 1, size = 3, color = "black", show.legend = FALSE) +
geom_text(x = Inf, y = 4, label = "· P < 0.1", hjust = 0.2, vjust = 1, size = 3, color = "black", show.legend = FALSE)
可能需要微调一下最后加上去的两个P值图层的y值。
