#Task 3: PCA Analysis of RNA-seq Data
This task performs Principal Component Analysis (PCA) on the RNA-seq gene expression matrix to explore sample clustering patterns and identify variance structure in the data.
We conduct PCA on both: - The z-score normalized expression matrix by samples (to identify clustering of samples). - The transposed matrix by genes (to examine gene-level variation across samples).
Each PCA analysis generates PDF plots for the top principal components (PC1, PC2, and PC3), which can be used for QC or further interpretation.
# Load z-score normalized expression matrix if not in environment
GE_z <- read.csv("RNA_TMM_log2_zscore.csv", row.names = 1)
# -------------------------
# PCA by samples (columns of GE_z)
# -------------------------
# Perform PCA on z-score normalized gene expression matrix (genes × samples)
# Each column is a sample → this PCA reflects sample clustering
pc <- prcomp(GE_z, scale. = TRUE)
# Save PCA plots of samples to a PDF file
pdf("PCA_plots_samples.pdf")
# Plot PC1 vs PC2
plot(pc$x[,1], pc$x[,2], col="blue", pch=16, main="PC1 vs PC2")
# Plot PC2 vs PC3
plot(pc$x[,2], pc$x[,3], col="green", pch=16, main="PC2 vs PC3")
# Plot PC1 vs PC3
plot(pc$x[,1], pc$x[,3], col="red", pch=16, main="PC1 vs PC3")
# Close PDF device
dev.off()## png
## 2# -------------------------
# PCA by genes (transpose of GE_z)
# -------------------------
# Transpose expression matrix → rows now represent samples, columns represent genes
GE_t <- t(GE_z)
# Remove genes with more than 20% missing values
GE_t <- GE_t[, colMeans(is.na(GE_t)) <= 0.2]
# Remove genes with zero standard deviation (constant values across samples)
GE_t <- GE_t[, apply(GE_t, 2, function(x) sd(x, na.rm = TRUE)) > 0]
# Replace remaining NAs (if any) with the column-wise mean
GE_t[is.na(GE_t)] <- apply(GE_t, 2, function(x) mean(x, na.rm = TRUE))
# Perform PCA on this filtered and imputed gene matrix
pc <- prcomp(GE_t, scale. = TRUE)
# Save PCA plot of genes to a PDF
pdf("PCA_plots_genes.pdf")
plot(pc$x[,1], pc$x[,2], col="blue", pch=16, main="PC1 vs PC2 (Genes)")
dev.off()## png
## 2