R ggrepel for overlapping text in the whole plot

Question

[camillab.]

Hi!

I am trying to make a scatterplot with ggplot . I have subset my point into 4 groups based on p.value and range of log2fc. I used the geom_text_repel to avoid overlapping text and it works within the same "group" but among different groups some text overlaps.

here my code (apologies if your eyes are bleeding):

#scatterplot
library(ggplot2)
library(ggrepel)

#with pvalue vbut less logf2fc highlighted
dfup <- text %>%  filter(log2FC >= 1.5 & p_value <= 0.05) #ok
pvalup<- text %>%  filter(p_value <= 0.05 &  log2FC > 0 & log2FC <= 1.4)
pvaldown<- text %>%  filter(p_value <= 0.05 & log2FC >= -1.4 & log2FC < 0)
dfdown <- text %>%  filter(log2FC <= -1.5 & p_value <= 0.05) #ok


ggplot(data = text, aes(x= log2FC, y= -1*log10(p_value))) + 
  geom_point() + 
  geom_point(data= pvaldown, aes(x= log2FC, y= -1*log10(p_value)), color = "dodger blue")+
  geom_point(data= pvalup, aes(x= log2FC, y= -1*log10(p_value)), color = "magenta")+
  geom_point(data= dfup, aes(x= log2FC, y= -1*log10(p_value)), color = "red")+
  geom_point(data= dfdown, aes(x= log2FC, y= -1*log10(p_value)), color = "blue")+
  geom_vline(xintercept = 1.5,linetype = "dashed", color = "dim gray")+ 
  geom_vline(xintercept = -1.5,linetype = "dashed", color = "dim gray")+
  geom_hline(yintercept = -1*log10(0.05),linetype = "dashed", color = "dim gray") +
  geom_hline(yintercept = -1*log10(0.01),linetype = "dotted", color = "gray")+ 
  geom_text(mapping=aes(x=-10,y=2.0),label=paste("0.01"), color = "gray", size = 2.5, vjust=-0.5)+
  geom_hline(yintercept = -1*log10(0.001),linetype = "dotted", color = "gray") +
  geom_text(mapping=aes(x=-10,y=3.0),label=paste("0.001"), color = "gray", size = 2.5, vjust=-0.5)+
  geom_text_repel(data=dfdown, aes(label=`Associated.Gene.Name`), color = "blue")+
  geom_text_repel(data=pvalup, aes(label=`Associated.Gene.Name`), color = "magenta")+
  geom_text_repel(data=pvaldown, aes(label=`Associated.Gene.Name`), color = "dodger blue")+
  geom_text_repel(data=dfup, aes(label=`Associated.Gene.Name`), color = "red")+
  theme_bw(base_size = 12)+ #size of the labelling +
  theme( title = element_text(hjust = 0.5), axis.title = element_text(color = "black"),panel.grid.major = element_blank(),
         panel.grid.minor = element_blank())

what I can do to avoid the overlapping of the plots?

Thank you!

Camilla

[Meisam]

If you are using multiple geom_text_repel in one plot you can try to set a specific direction for each set so they don’t collide. Check this Stackoverflow post to see if it answers your question.

[Basti]

Could you share a small example of your dataset ? dput(head(your_df)) would be great to reproduce your issue

[camillab.]

sure. row are the genes and the columns the samples.

# A tibble: 6 × 12
  ID        Descr…¹ Assoc…² 2-3dp…³ `8dpg` `14dpg` `18dpg` hu1_u…⁴ hu2_u…⁵ hu3_u…⁶ p_value
  <chr>     <chr>   <chr>     <dbl>  <dbl>   <dbl>   <dbl>   <dbl>   <dbl>   <dbl>   <dbl>
1 ENSG0000… tetras… TSPAN6  1.31e-5   6.94 4.39e-6   2.75    0.536    1.18   0.827 0.455  
2 ENSG0000… dolich… DPM1    1.97e+2  33.2  3.73e+1  29.6    24.1     22.9   30.0   0.362  
3 ENSG0000… SCY1-l… SCYL3   6.08e+0   4.86 5.73e+0   6.24    2.02     3.30   2.84  0.00154
4 ENSG0000… chromo… C1ORF1… 7.01e+0   2.35 2.99e+0   3.76    1.30     1.19   1.52  0.0801 
5 ENSG0000… feline… FGR     6.37e-1   2.81 1.17e+0   1.54    5.08     1.27   0.403 0.613  
6 ENSG0000… comple… CFH     5.13e-4   2.12 4.67e-2   0.753  17.8      2.45   3.82  0.139  
# … with 1 more variable: log2FC <dbl>, and abbreviated variable names ¹​Description,
#   ²​Associated.Gene.Name, ³​`2-3dpg`, ⁴​hu1_untreated, ⁵​hu2_untreated, ⁶​hu3_untreated
# ℹ Use `colnames()` to see all variable names

Answer 1

[Trivas]

I'd recommend combining all of your data into one dataframe/tibble, then adding a column to denote their category. You can then use color = category within the aes() call of geom_point. This will also fix your geom_text_repel issue since the call to that function will be able to see all text labels and not just the one within each separate call.

For example:

#with pvalue but less logf2fc highlighted
dfup <- text %>%  filter(log2FC >= 1.5 & p_value <= 0.05) #ok
pvalup<- text %>%  filter(p_value <= 0.05 &  log2FC > 0 & log2FC <= 1.4)
pvaldown<- text %>%  filter(p_value <= 0.05 & log2FC >= -1.4 & log2FC < 0)
dfdown <- text %>%  filter(log2FC <= -1.5 & p_value <= 0.05) #ok

Could become

text %>% mutate(category = case_when(log2FC >= 1.5 & p_value <= 0.05 ~ "dfup",
                                     p_value < 0.05 & log2FC > 0 & log2FC <= 1.4 ~ "pvalup",

etc.