R语言中实现按染色体分组着色的PCA可视化求助

Hey there! Let's walk through how to adjust your PCA visualization to color genes by their chromosome. I'll break this down into simple, actionable steps using the tools you're already using (factoextra's fviz_pca_ind), plus some data prep to tie chromosome info to your genes.

Step 1: Prepare Your Data #

First, you need two core datasets:

• A gene expression matrix: Rows = genes, columns = samples (if your data is flipped, we'll fix that later)
• A metadata table: A data frame with at least two columns: one for gene IDs (matching exactly to your expression matrix row names) and one for chromosome assignments.

Here's example sample data you can use to test (replace with your real data):

# Load required packages first
library(FactoMineR)
library(factoextra)
library(dplyr)
library(RColorBrewer) # For better color palettes (optional)

# Create sample gene expression data (50 genes, 10 samples)
set.seed(123)
gene_expr <- matrix(rnorm(50*10), 
                    nrow = 50, 
                    dimnames = list(paste0("Gene", 1:50), paste0("Sample", 1:10)))

# Create sample chromosome metadata (map each gene to chr1-chr5)
gene_chr <- data.frame(
  Gene = paste0("Gene", 1:50),
  Chromosome = sample(paste0("chr", 1:5), 50, replace = TRUE)
)

Step 2: Run PCA #

We'll use PCA() from FactoMineR. Make sure your genes are rows (these are the "individuals" in PCA terms) and samples are columns (the "variables"). If your data has samples as rows instead, transpose it with t(gene_expr).

# Run PCA (scale.unit = TRUE standardizes gene expression values—critical for meaningful PCA!)
res.pca <- PCA(gene_expr, scale.unit = TRUE, graph = FALSE)

Step 3: Merge Chromosome Info with PCA Results #

We need to link each gene's chromosome to its PCA coordinates so we can use this for coloring in the plot:

# Extract PCA individual coordinates and add chromosome data
pca_gene_data <- as.data.frame(res.pca$ind$coord) %>%
  mutate(
    Gene = rownames(.), # Pull gene IDs from row names
    Chromosome = gene_chr$Chromosome[match(Gene, gene_chr$Gene)] # Match each gene to its chromosome
  )

Step 4: Visualize PCA with Chromosome Coloring #

Now modify your fviz_pca_ind call to color by chromosome instead of cos2. We'll add some polish to make the plot clear and readable:

# Choose a color palette (adjust based on how many chromosomes you have)
# Using RColorBrewer's qualitative palette for better distinction
chr_palette <- brewer.pal(n = length(unique(pca_gene_data$Chromosome)), name = "Set3")

# Generate the final plot
fviz_pca_ind(res.pca,
             col.ind = pca_gene_data$Chromosome, # Color points by chromosome
             palette = chr_palette, # Use our custom color set
             repel = TRUE, # Prevent label overlap (your original setting!)
             legend.title = "Chromosome", # Label the legend clearly
             title = "PCA of Genes Colored by Chromosome", # Add a descriptive title
             geom = c("point", "text"), # Show both points and gene labels (remove "text" if too crowded)
             point.size = 3 # Make points easier to spot
)

Key Notes & Troubleshooting #

• Matching Gene IDs: Double-check that gene IDs in your expression matrix and metadata are identical (case-sensitive!). If they don't match, match() will return NA for those genes.
• Scaling: Never skip scale.unit = TRUE—gene expression values vary wildly, and scaling ensures every gene contributes equally to the PCA.
• Too Many Chromosomes: If you have dozens of chromosomes, skip the gene labels (remove "text" from geom) or use a smaller point size to avoid clutter.
• Custom Colors: If you don't want to use RColorBrewer, define your own palette with hex codes, like palette = c("#FF0000", "#00FF00", "#0000FF").

内容的提问来源于stack exchange，提问作者FeliciaC