How to make a lot of boxplot s in the same figure

Question

[Bioiris]

Hello,

I have a table fpkm (20 tissues)

gene_id category tissue1 tissue2 tissue3 tissue4 …... tissue20 g1 unkown g2 Known g3 Known g4 unkown . . . . g3000

[url=https://postimg.org/image/4nvn96jdn/][img]https://s9.postimg.org/757egg3a7/Screen_Shot_2016_09_14_at_12_10_11_PM.png[/img][/url][url=https://postimage.org/]photo uploading[/url]

I would like to make boxplot (in the same figure) of expression in two categories across all tissues

like in the picture

Thank you

[url=https://postimg.org/image/5l3cmfqjx/][img]https://s22.postimg.org/5l3cmfqjx/Picture1.png[/img][/url]

[nadimintikrish]

You can certainly use par(mfrow=c(x,y)) where x= rows and y=columns. Run the box plots the x*y array and finally set it to par(mfrow=c(1,1)). par(mfrow(x,y)) divides the graph board in to x columns and Y rows.

Answer 1

[russhh]

There is an example of that kind of plot at http://docs.ggplot2.org/current/geom_boxplot.html You'd need to manipulate your matrix from wide (as you've given at postimg.org) to long format before using the code there:

library(reshape) # wide to long
library(ggplot2) # boxplots
library(magrittr) # %>%
set.seed(1) # 

# simplified version of your data.frame
dfr.wide <- data.frame(
  gene.id = as.character(1:10),
  category = sample(c('known', 'unknown'), size = 10, replace = TRUE),
  tissue.1 = rnorm(10),
  tissue.2 = rnorm(10),
  tissue.3 = rnorm(10),
  tissue.4 = rnorm(10)
  )

dfr.wide %>% head

  gene.id category   tissue.1    tissue.2    tissue.3   tissue.4
1       1    known -0.8204684 -0.04493361 -0.05612874 -0.4149946
2       2    known  0.4874291 -0.01619026 -0.15579551 -0.3942900
3       3  unknown  0.7383247  0.94383621 -1.47075238 -0.0593134
4       4  unknown  0.5757814  0.82122120 -0.47815006  1.1000254
5       5    known -0.3053884  0.59390132  0.41794156  0.7631757
6       6  unknown  1.5117812  0.91897737  1.35867955 -0.1645236

# want:
# tissue   category gene expression
# tissue.1 unknown  '1'  1.90 ...
# .
# .

dfr.long <- melt(
  dfr.wide,
  id = c('gene.id', 'category'),
  variable_name = 'tissue'
  )

dfr.long %>% head
  gene.id category   tissue      value
1       1    known tissue.1 -0.8204684
2       2    known tissue.1  0.4874291
3       3  unknown tissue.1  0.7383247
4       4  unknown tissue.1  0.5757814
5       5    known tissue.1 -0.3053884
6       6  unknown tissue.1  1.5117812

# plot it:
ggplot(
  data = dfr.long,
  aes(x = tissue, y = value)
  ) + geom_boxplot(aes(col = category))

End:

Plotting code:

[Bioiris]

Thank you! it's work but there is the probleme of resolution i can't differentiate between boxplots

i need a method without ggplot

[russhh]

I'd suggest you experiment with ggplot for a while before you disregard it: play with the aesthetics/themes, reduce your dataset etc. The problem may be one of design rather than implementation

[nwon]

I second ggplots and look at violin plots within it link to geom_violin plots and you can colour by some factor or group.

You can then use the facet function to further subclassify your groups. Link to facet grid

Answer 2

[nadimintikrish]

You can certainly use par(mfrow=c(x,y)) where x= rows and y=columns. Run the box plots the x*y array and finally set it to par(mfrow=c(1,1)). par(mfrow(x,y)) divides the graph board in to x columns and Y rows.

[Antonio R. Franco]

You can reset the graphical setting with a single dev.off() command

Answer 3

[Antonio R. Franco]

Take a look to this ggplot2 tutorial that includes the code you can reuse for your own data Don't forget to revise the Aesthetics section to change the aspect of your graphics