本文档演示如何用 ggplot2 + gghalves 复现 Nature Communications 图中常见的“云雨图（raincloud plot）”效果：半小提琴（云）+ 箱线（体）+ 抖点（雨）。
关键词：半小提琴 geom_half_violin()、半箱线 geom_half_boxplot()、抖点 geom_jitter()、自定义顺序与配色、白填充箱线。

0. 环境准备

# 建议的基础包
if (!requireNamespace("tidyverse", quietly = TRUE)) install.packages("tidyverse")
if (!requireNamespace("gghalves",  quietly = TRUE)) install.packages("gghalves")

library(tidyverse)
library(gghalves)

set.seed(20250923)  # 保证可复现

1. 构造示例数据（与提问一致）

df <- data.frame(
  subtills   = c(rnorm(100, 14, 1.3), rnorm(20, 18, 3.2), rnorm(30, 11, 2.5)),
  cereus     = c(rnorm(60,  11, 1.6), rnorm(40,  8.5, 1.2), rnorm(50,  5.5, 1.2)),
  megaterium = c(rnorm(130, 6.5, 1.1), rnorm(10,  9.0, 2.2), rnorm(10,  4.5, 0.8)),
  circulans  = c(rnorm(100, 4.2, 0.9), rnorm(40,  2.6, 0.6), rnorm(10,  1.3, 0.4))
) %>%
  tidyr::gather() %>%
  dplyr::rename(Species = key, value = value)

head(df)

默认顺序可能与论文图不一致。我们可自定义排序：

variable <- c("subtills", "cereus", "megaterium", "circulans")
df <- df %>% mutate(Species = factor(Species, levels = variable))

自定义调色板（与示例中的色调保持一致）：

pal <- c(
  subtills   = "#5E81AC",  # desaturated blue
  cereus     = "#D08770",  # salmon
  megaterium = "#A3BE8C",  # green
  circulans  = "#EBCB8B"   # sand
)

2. 基础：全小提琴 + 箱线（快速复现）

ggplot(df, aes(x = Species, y = value, fill = Species)) +
  geom_violin(alpha = 0.9, linewidth = 0.3, trim = FALSE) +
  geom_boxplot(width = 0.18, linewidth = 0.4, outlier.shape = 21, outlier.size = 0.8, fill = "white") +
  scale_fill_manual(values = pal) +
  coord_cartesian(ylim = c(0, 25)) +
  labs(x = NULL, y = "NO. of BGCs/genome") +
  theme_test(base_size = 15) +
  theme(
    legend.position = "none",
    axis.text = element_text(color = "black"),
    axis.text.x = element_text(face = "italic")
  )

3. 论文风格“云雨图”：半小提琴 + 半箱线 + 抖点分列

思路： - 半小提琴和半箱线放在 右侧（side = "r"）； - 抖点轻微左移（position_nudge(x = -0.2)），形成“云雨”分列； - 箱线白填充、轮廓着色以匹配小提琴颜色； - 适度加 alpha，避免遮挡。

ggplot(df, aes(x = Species, y = value)) +
  # 云：半小提琴（右侧）
  gghalves::geom_half_violin(
    aes(fill = Species, color = Species),
    side = "r", alpha = 0.9, linewidth = 0.3, trim = FALSE
  ) +
  # 体：半箱线（右侧，白填充+彩色边）
  gghalves::geom_half_boxplot(
    aes(color = Species),
    side = "r", width = 0.2, outlier.shape = NA,
    position = position_nudge(x = 0.15),  # 轻微右移，与小提琴并排但不完全重合
    fill = "white", linewidth = 0.5
  ) +
  # 雨：抖点（左侧）
  geom_jitter(
  aes(color = Species),
  width = 0.12, height = 0, size = 1.4, alpha = 0.75
) +
  scale_fill_manual(values = pal) +
  scale_color_manual(values = pal) +
  coord_cartesian(ylim = c(0, 25)) +
  labs(x = NULL, y = "NO. of BGCs/genome") +
  theme_test(base_size = 15) +
  theme(
    legend.position = "none",
    axis.text = element_text(color = "black"),
    axis.text.x = element_text(face = "italic")
  )

做出的图更多需要微调，使用数值化 x 精准微调

“把因子转成数值 + 手动偏移”的做法来微调，下面给出对应实现：

x_num <- as.numeric(df$Species)

ggplot(df) +
  # 半小提琴（右）
  gghalves::geom_half_violin(
    aes(x = x_num + 0.10, y = value, fill = Species, color = Species),
    side = "r", alpha = 0.9, linewidth = 0.3, trim = FALSE
  ) +
  # 半箱线（右、白填充）
  gghalves::geom_half_boxplot(
    aes(x = x_num + 0.10, y = value, color = Species),
    side = "r", width = 0.18, outlier.shape = NA, fill = "white", linewidth = 0.5
  ) +
  # 抖点（左）
  geom_jitter(
    aes(x = x_num - 0.18, y = value, color = Species),
    width = 0.08, size = 1.4, alpha = 0.75
  ) +
  scale_fill_manual(values = pal) +
  scale_color_manual(values = pal, guide = "none") +
  # 恢复离散刻度标签
  scale_x_continuous(
    breaks = sort(unique(x_num)),
    labels = levels(df$Species)
  ) +
  coord_cartesian(ylim = c(0, 25)) +
  labs(x = NULL, y = "NO. of BGCs/genome") +
  theme_test(base_size = 15) +
  theme(
    legend.position = "none",
    axis.text = element_text(color = "black"),
    axis.text.x = element_text(face = "italic")
  )

4. 小贴士（避免常见坑）

  • 箱线白填充：将 fill = "white" 放在 geom_half_boxplot()（或 geom_boxplot()）里，同时使用 aes(color = Species) 让边框着色，达到“彩色边 + 白色内部”的审稿友好风格。
  • 不要用 cex：在 ggplot2 里，请用 size（点/文本大小）、linewidth（线宽）。
  • 限制 y 轴但不截去图层：优先用 coord_cartesian(ylim = ...)，而非 scale_y_continuous(limits = ...)，后者可能会裁剪统计计算。
  • 多图例：如需隐藏重复图例，可在某个 scale_*_manual() 里加 guide = "none"。

5. 导出图片（PNG + PDF）

p <- ggplot(df, aes(x = Species, y = value)) +
  gghalves::geom_half_violin(aes(fill = Species, color = Species), side = "r", alpha = 0.9, linewidth = 0.3) +
  gghalves::geom_half_boxplot(aes(color = Species), side = "r", width = 0.2, outlier.shape = NA, position = position_nudge(x = 0.08), fill = "white", linewidth = 0.5) +
  geom_jitter(aes(color = Species), position = position_nudge(x = -0.18), width = 0.08, size = 1.4, alpha = 0.75) +
  scale_fill_manual(values = pal) + scale_color_manual(values = pal) +
  coord_cartesian(ylim = c(0, 25)) +
  labs(x = NULL, y = "NO. of BGCs/genome") +
  theme_test(base_size = 15) +
  theme(legend.position = "none", axis.text = element_text(color = "black"), axis.text.x = element_text(face = "italic"))

ggsave("raincloud_demo.png", p, width = 8, height = 5, dpi = 300)
ggsave("raincloud_demo.pdf", p, width = 8, height = 5, device = cairo_pdf)

补充：
- 这种组合（半小提琴 + 箱线 + 抖点）就是常说的 雨云图 / 云雨图（raincloud plot）。
- 也可考虑 ggdist::stat_halfeye()、ggbeeswarm::geom_quasirandom() 等扩展实现，视美观与性能取舍。更多请关注xhs：飞高高啦