The example is split into 2 Parts:
Part 1 must be completed first to create a file,
SNPs_cleaned.csv, that has been completely prepared for
analysis.
Now in Part 2, you will analyze the data with PCA. The steps here will be:
scale())prcomp())screeplot())vegan::scores())In the code below all code is provided. Your tasks will be to do 4 things:
Load the vcfR package with library()
library(vcfR) # KEY
##
## ***** *** vcfR *** *****
## This is vcfR 1.13.0
## browseVignettes('vcfR') # Documentation
## citation('vcfR') # Citation
## ***** ***** ***** *****
library(vegan)
## Loading required package: permute
## Loading required package: lattice
## This is vegan 2.6-4
library(ggplot2)
library(ggpubr)
Set the working directory
setwd("~/Downloads/COMPBIO")
Load the data
SNPs_cleaned <- read.csv(file = "SNPs_cleaned.csv")
warning("If this didn't work, its may be because you didn't set your working directory.")
## Warning: If this didn't work, its may be because you didn't set your working
## directory.
Here, we need to center the data before we perform PCA. To center the
data we loaded, SNPs_cleaned.csv, we use the
scale() function, which makes the mean equal to 1 and the
standard deviation equal to 0.
SNPs_scaled <- scale(SNPs_cleaned)
Here, we use the base R function prcomp() on our scaled
data stored in SNPs_scaled to run the PCA analysis. We then
save the results to the object pca_scaled.
pca_scaled <- prcomp(SNPs_scaled)
Here, we use the function screeplot() on our PCA data
stored in pca_scaled to make a scree plot, which shows us
the variance between the multiple PCs. We labelled the y-axis “Relative
importance” with the argument ylab and gave the plot a main
title called “SNP Scree Plot” with the argument main.
screeplot(pca_scaled,
ylab = "Relative importance",
main = "SNP Scree Plot")
The scree plot that prints as a result shows us that PC1 is the most important for analysis. This is determined due to the large drop between PC1 and PC2, as the PCs before the drop are the most important. As the researchers, we can now decide to perform the rest of our analysis using PC1 and PC2, since the other PCs will not give much useful information.
In the series of codes, we explain the amount of variation
demonstrated by the different PCs shown in the scree plot. First, we
save the summary statistics of the data to the object
summary_out_scaled. Then, we create a function that will
extract the information about the explained variation that we want to
analyze. The function contains an argument that allows the researcher to
chose how many PCs to plot. The default is 2, but in the next chunk of
code, we run the function with 10 PCs, since that is how many are
included in the scree plot. Lastly, we created a new scree plot that
uses the y-axis as the percent variation captured by each PC using the
function barplot().
summary_out_scaled <- summary(pca_scaled)
PCA_variation <- function(pca_summary, PCs = 2){
var_explained <- pca_summary$importance[2,1:PCs]*100
var_explained <- round(var_explained,1)
return(var_explained)
}
var_out <- PCA_variation(summary_out_scaled,PCs = 10)
N_columns <- ncol(SNPs_scaled)
barplot(var_out,
main = "Percent variation Scree plot",
ylab = "Percent variation explained")
abline(h = 100/N_columns*100, col = 2, lwd = 2)
The line that is added to the plot shows where the height of all the PCs should be if they were all equally important. As we can see, PC1 is significantly above this line, and the rest of the PCs are significantly below this line. This further shows that PC1 is the most important PC for analysis.
Using the function biplot(), we create the biplot that
demonstrates the relationship between the features of the SNP data
against the two PC axes.
biplot(pca_scaled)
As we can see from the biplot, the information is hard to interpret and decipher. This is due to there being lots of different data points that cloud the plot. From this biplot, we cannot determine any noticeable relationships.
Here, we use the function vegan::scores() to extract the
PCA scores from the PCA data we have stored in the object
pca_scaled. We then saved this data to the object
pca_scores. These scores can later be used to run PCA using
the vegan package.
pca_scores <- vegan::scores(pca_scaled)
Here, we create a vector containing all of the population codes of
the the corresponding samples of SNP data using the function
c() and save the vector to an object called
pop_id.
pop_id <- c("Nel","Nel","Nel","Nel","Nel","Nel","Nel","Nel",
"Nel", "Nel", "Nel", "Nel", "Nel", "Nel", "Nel", "Alt",
"Alt", "Alt", "Alt", "Alt", "Alt", "Alt", "Alt", "Alt",
"Alt", "Alt", "Alt", "Alt", "Alt", "Alt", "Sub", "Sub",
"Sub", "Sub", "Sub", "Sub", "Sub", "Sub", "Sub", "Sub",
"Sub", "Cau", "Cau", "Cau", "Cau", "Cau", "Cau", "Cau",
"Cau", "Cau", "Cau", "Cau", "Cau", "Div", "Div", "Div",
"Div", "Div", "Div", "Div", "Div", "Div", "Div", "Div",
"Div", "Div", "Div", "Div")
Here, we combine the population codes with the PCA scores in a data
frame called pca_scores2 using the function
data.frame(). We need the data in this structure to run PCA
on it.
pca_scores2 <- data.frame(pop_id,
pca_scores)
Using the ggpubr package, we use the function
ggscatter() to create a scatterplot of the PCA scores along
the PC1 and PC2 axes. We labelled the X and Y axes appropriately using
the xlab and ylab arguments and made the shape
and color of the data points vary based on their respective population
codes using the arguments color and shape.
ggpubr::ggscatter(data = pca_scores2,
y = "PC2",
x = "PC1",
color = "pop_id",
shape = "pop_id",
xlab = "PC1 (19.9% variation)",
ylab = "PC2 (2.3% variation)",
main = "SNP PCA Scatterplot")
The scatterplot shows the Cau and Div population codes most correlated with PC1, as they are farther along the x-axis. The Alt and Nel population codes seem to have slight correlations with PC2, and the Sub population seems to have negative correlations with both PC1 and PC2.