Suppose I have this dataframe:

# packages
library(ggplot2)
library(dplyr)

# data
genes <- c("gene1", "gene2", "gene2")
tissue <- c("Blood", "Nerve", "Brain")
pval <- c(0.8, 0.6, 0.005)

df <- data.frame(col_x = tissue,
                 col_y = genes,
                 p = pval)

I create a column for labeling purposes in the plot:

df <- df %>%
  mutate(`Post. Prob.` = ifelse(`p`>= 0.5, "P > 0.5",
                         ifelse(`p`>= 0.01 & `p`<= 0.5, "P > 0.01","Not sig")))

The remaining code is for different aesthetics in the plot.

# black ring if P>0.5
# gray ring if between 0.01 and 0.5
# white ring if not sig
ring_df <- df %>%
  select(`Post. Prob.`) %>%
  mutate(`ring` = ifelse(`Post. Prob.` == "P > 0.5", "black",
                         ifelse(`Post. Prob.` == "P > 0.01", "gray", "white")))
rings <- ring_df$ring
names(rings) <- ring_df$`Post. Prob.`

Now I create the plot:

ggplot(df) 
  geom_point(aes(x=`col_x`, y=`col_y`,
                 size=`p`,
                 color=`Post. Prob.`),
             shape=21, # so I can fill and change the border color
             fill="#6565ff", # light blue
             stroke = ifelse(df$`Post. Prob.` == "P > 0.5",2, # stroke is the thickness of the border
                             ifelse(df$`Post. Prob.` == "P > 0.01", 2, 0.5))) 
  scale_color_manual(values=rings,
                     breaks=c("P > 0.5", "P > 0.01", "Not Sig."),
                     labels=c("P > 0.5", "P > 0.01", "Not Sig."),
                     name="Post. Prob.") 
  scale_size(range=c(5,10)) 
  xlab("tissue") 
  ylab("genes")

However, my question is that in the legend, the plot only shows P > 0.5. I know that none of the points follow the other parameters, but how do I make it so that the gray legend marker would show?

CodePudding user response：

One option to achieve your desired result would be

1. to convert your post prob column to a factor with the appropriate levels
2. set drop=FALSE and limits=force inside scale_color_manual to avoid that unused factor levels get dropped

Notes:

1. Instead of using nested ifelse I would suggest to switch to case_when.
2. There is no need to wrap each column name inside backticks. Backticks are only needed if you use awkward column names including e.g. spaces.
3. I would suggest to avoid using column names including spaces such as Post. Prob. as there are easier ways to get nice labels e.g. using labs()

library(ggplot2)
library(dplyr)

genes <- c("gene1", "gene2", "gene2")
tissue <- c("Blood", "Nerve", "Brain")
pval <- c(0.8, 0.6, 0.005)

df <- data.frame(
  col_x = tissue,
  col_y = genes,
  p = pval
)

df <- df %>%
  mutate(post_prob = case_when(
    p >= 0.5 ~ "P > 0.5",
    p >= 0.01 & p <= 0.5 ~ "P > 0.01", 
    TRUE ~ "Not sig"),
    post_prob = factor(post_prob, levels = c("P > 0.5", "P > 0.01", "Not sig"))
  )

rings <- df %>%
  select(post_prob) %>%
  mutate(ring = case_when(
    post_prob == "P > 0.5" ~ "black",
    post_prob == "P > 0.01" ~ "gray", 
    TRUE ~ "white")
  ) %>% 
  tibble::deframe()

ggplot(df)  
  geom_point(aes(
    x = col_x, y = col_y,
    size = p,
    color = post_prob,
    stroke = case_when(
      post_prob == "P > 0.5" ~ 2,
      post_prob == "P > 0.01" ~ 2,
      TRUE ~ 0.5
    )
  ),
  shape = 21,
  fill = "#6565ff"
  )  
  scale_color_manual(
    values = rings,
    name = "Post. Prob.",
    drop = FALSE, 
    limits = force
  )  
  scale_size(range = c(5, 10))  
  xlab("tissue")  
  ylab("genes")