计算散点图中的标签数量

这是相当直接的,只要你不试图在ggplot中这样做。事先做几次数据总结,并将其用于绘图。在这种情况下,我认为最直接的方法是首先定义标签打印的坐标,然后计算每个象限的点(您需要决定将“连接”的点放置在哪里)。

library(tidyverse)

## assuming you know your cut off values
y_cut <- .16
x_cut <- .5
## basic data frame modificiation for easier coding
data <- data %>%
  mutate(
    x = 1 - Model2,
    y = Model1
  )
## first, get the label positions.
## We will define the quadrants clockwisefrom top right to top left
label_pos <-
  data %>%
  summarise(
    x_right = mean(c(x_cut, max(x))),
    x_left = mean(c(x_cut, min(x))),
    y_top = mean(c(y_cut, max(y))),
    y_bottom = mean(c(y_cut, min(y)))
  )
label_coord <- 
  data.frame(matrix(unlist(c(rep(label_pos[1:2], each = 2), c(label_pos[3:4], label_pos[4:3]))), nrow = 4))
names(label_coord) <- c("x_lab", "y_lab")
## now, summarise how many dots are per quadrant based on those cut offs
## and add the label positions
data_count <- data %>%
  mutate(quad = case_when(
    x > x_cut & y > y_cut ~ "quad1",
    x > x_cut & y < y_cut ~ "quad2",
    x < x_cut & y < y_cut ~ "quad3",
    x < x_cut & y > y_cut ~ "quad4"),
    ## convert both groups to factors
    cancer_type = as.factor(cancer_type), 
    quad = as.factor(quad)
  ) %>%
  count(cancer_type, quad, .drop = FALSE)
## now merge both
data_annot <- data_count %>%
  bind_cols(label_coord[rep(1:nrow(label_coord), 3), ])
## do the data transformation before
ggplot(mapping = aes(x, Model1)) +
  geom_point(data = data, aes(fill = stage), size = 4, pch = 21) +
  ## now you can simply use geom_text for annotation
  geom_text(data = data_annot, mapping = aes(x_lab, y_lab, label = paste("n=", n))) +
  facet_wrap(. ~ cancer_type) +
  scale_fill_manual(values = c("#E69F00", "#56B4E9", "#999999", "#999999")) +
  xlab("Model 2") +
  ylab("Model 1") +
  theme_classic() +
  geom_hline(yintercept = y_cut, linetype = 2) +
  geom_vline(xintercept = x_cut, linetype = 2)

您可以访问ggplot处理的数据,直到通过提取 data 从的输出 ggplot_build(p) 哪里 p 是前者的结果 ggplot 跑示例:

ggplot_build(p)$data[[1]]

这些数据很容易制成表格,例如使用{dplyr},使用您的示例数据:

x_intercept = .5
y_intercept = .16

p <-  df |> ## save ggplot object as "p"
  ggplot(aes(1 - Model2, Model1, fill=stage)) +
  geom_point(size=4,pch=21) + theme_classic()+
  facet_wrap( ~ cancer_type)+
  scale_fill_manual(values=c('#E69F00', '#56B4E9','#999999','#999999')) +
  xlab("Model 1") + ylab("Model 2") +
  geom_hline(yintercept = y_intercept, linetype=2) +
  geom_vline(xintercept = x_intercept, linetype=2)  

p ## display ggplot

将ggplot数据制成表格:

library(dplyr)

ggplot_build(p)$data[[1]] |>
                count(PANEL, ## the facet_wrap panel index (~ cancer_type)
                      fill, ## the fill (~ stage)
                      x > x_intercept, 
                      y > y_intercept
                      )

输出(截断):

+    PANEL    fill x > x_intercept y > y_intercept n
1      1 #56B4E9           FALSE            TRUE 1
2      1 #999999           FALSE           FALSE 1
3      1 #999999           FALSE            TRUE 1
4      1 #E69F00           FALSE           FALSE 1

您可以使用 case_when() 定义每个点所在的象限,然后使用 count() ,分组依据 stage 和 cancer_type

x=0.5
y=0.16
df %>% mutate(q = case_when(
  Model2<x & Model1>y~"ul",
  Model2>=x & Model1>y~"ur",
  Model2<x & Model1<=y~"ll",
  Model2>=x & Model1<=y~"lr",
)) %>% count(stage,cancer_type,q)

以下是另一个使用 ggrepel :注意 df 是原始数据帧,并且 df1 仅用于标记:

工作原理:

1.创建数据帧仅用于标记目的。注意,我们必须消除所有重复!

library(tidyverse)

df1 <- df %>% 
  mutate(helper_x = ifelse(Model2 <= 0.5, "a", "b"),
         helper_y = ifelse(Model1 <= 0.16, "a", "b")) %>% 
  group_by(cancer_type, helper_x, helper_y) %>% 
  arrange(.by_group = TRUE) %>% 
  add_count(stage) %>% 
  mutate(label = paste(stage, "N=", n, sep=" ")) %>% 
  group_by(cancer_type, helper_x, helper_y, stage) %>% 
  slice(1)

2.绘图和标签:

library(tidyverse)
library(ggrepel)

df %>%
  ggplot(aes(x=1-Model2, y=Model1, color=stage))+
  geom_point(size=4, alpha=0.8)+
  facet_wrap(.~cancer_type)+
  scale_color_manual(values = c("#E69F00", "#56B4E9", "#999999", "#999999")) +
  geom_vline(xintercept = 0.5, linetype = "longdash")+
  geom_hline(yintercept = 0.16, linetype = "longdash")+
  theme_classic()+
  geom_label_repel(data=df1, aes(label = label),
                   box.padding   = 0.35, 
                   point.padding = 0.5,
                   segment.color = 'grey50',
                   show.legend = FALSE) +
  xlab("Model2")