how to improve PCA visualization using ggplot?

I have used this piece of script to draw the PCA based on eigenvalue and eigenvector

percentage <- round((eigenval/(sum(eigenval))*100), 2)
percentage <- as.matrix(percentage)
percentage <- paste0(names(percentage), " (", percentage, "%)")
Names <- c ("mn27hd", "mdkk987", "mnsdnu83", "sjednu83", "bjeo972s")
pop.colour <- c("blue", "red", "green", "orange", "brown")
ggplot(eigenvec, aes(x=PC1, y=PC2, colour=pop.colour, label=Names)) +
    geom_point(size=3) + geom_text(aes(label=Names), hjust=0, vjust=-0.5) + xlab(percentage[1]) + ylab(percentage[2])

enter image description here

how can I:

change the xlab and ylab names to show as: PC1 (36.61%), PC2 (33.7%)?
move only the pink label slightly left-side cause when I do it generally then the two left labels go outside the image or too close to the edge?
change the legend name and categories from the colour to the population name?

eigenvalue r eigenvector pca ggplot

• 3,802 views • link

written 4.7 years ago by mthm •

1 answer

I like to use labs instead of xlab and ylab and add the information to the data.frame directly. Do you only have 4 data point and you are certain that the name are corresponding to the correct PC value? If so do

percentage <- round((eigenval/(sum(eigenval))*100), 2)
percentage <- as.matrix(percentage)
percentage <- paste0(names(percentage), " (", percentage, "%)")
eigenvec$Pop <- c ("mn27hd", "mdkk987", "mnsdnu83", "sjednu83", "bjeo972s")

pop.colour <- c("blue", "red", "green", "orange", "brown")
ggplot(eigenvec, aes(x=PC1, y=PC2, colour=Pop, label=Pop)) +
    geom_point(size=3) + geom_text(hjust=-0.1, vjust=-0.5) +
    scale_color_manual(values = pop.colour) + 
    labs(x=paste0("PC1 (",percentage[1],")"), y= paste0("PC2 (",percentage[2], ")"), color = "Population")

I don't know have mock data to play with so there might be bug to the code. You can modify hjust in geom_text to adjust the horizontal shift

• 0 views • link

written 4.7 years ago by Sam

Thanks the legend worked. the PC1 and PC2 axes didn't! it says:

Error in "PC1 (" + percentage[1] : 
  non-numeric argument to binary operator

that is how my percentage looks like

> percentage
[1] " (36.61%)" " (33.7%)"  " (19.75%)" " (18.03%)" " (-0.1%)"

I can only adjust the hjust=0.4 without throwing the two green and blue labels out of the graph! which is still not enough for the purple label to be fully in! I was thinking if it is possible to change the dimensions of the graph so that labels fit in?

• 0 views • link

written 4.7 years ago by mthm •

My bad, should have used , instead of +. Have now fixed that.

If you don't mind installing an additional package, I found ggrepel really useful in this scenario. You can find examples here

• 0 views • link

written 4.7 years ago by Sam

Great, fixed the label with this and worked fine (x=paste0("PC1" ,percentage[1]), y= paste0("PC2" ,percentage[2])

• 0 views • link

written 4.7 years ago by mthm •