我们可以按Maker分面，并将Cell_type放在“x”轴上

mock_data |> 
  ggplot(aes(x = Cell_type, y = Value)) +
  geom_boxplot(aes(fill = Cell_type), alpha = 0.2) +
  geom_point(aes(col = Subject, group = Cell_type), position = position_dodge(width = 0.75), size = 3) +
  geom_line(aes(group = Subject), lty = 2, col = "black") +
  facet_wrap(~Marker, strip.position = "bottom") +
  theme(
    axis.title.x = element_blank(),
    axis.text.x = element_blank(),
    axis.ticks.x = element_blank(),
    strip.background = element_blank()
    )

创建于2023-04-26带有reprex v2.0.2

对于一般的兴趣，这里有一个替代的解决方案，不需要刻面。然而，它确实需要“人工”生成线段的x坐标。

segment_positions <- mock_data |> 
  pivot_wider(id_cols = c(Subject,Marker), names_from = Cell_type, values_from = Value) |>
  mutate(x1 = as.numeric(Marker == 'B') + 1 - .75/4, x2 = as.numeric(Marker == 'B') + 1 + .75/4)

mock_data |> 
  ggplot(aes(x = Marker, y = Value)) +
  geom_boxplot(aes(fill = Cell_type), alpha = 0.2) +
  geom_point(aes(col = Subject, group = Cell_type), position = position_dodge(width = 0.75), size = 3) +
  geom_segment(data = segment_positions, aes(x = x1, xend = x2, y = B, yend = T), lty = 2, col = "black")

请注意，x坐标的偏移量是0.75/4，因为总减淡宽度是0.75。这意味着每个箱形图在减淡的每一侧都偏移了总宽度的一半（0.75/2）。因为这些点位于箱形图的中间，所以它们偏移了每个箱形图偏移宽度的一半（0.75/4）。