TCGA Breast Cancer Gene Expression Analysis Pipeline

This repository contains an R-based analysis pipeline for investigating differential gene expression in HER2/ERBB2+ breast cancer using TCGA RNA-seq data.

Prerequisites

Required R Packages

# Install BiocManager if not already installed
install.packages("BiocManager")

# Install required packages
BiocManager::install(c(
  "DESeq2",            # For differential expression analysis
  "clusterProfiler",   # For pathway analysis
  "pheatmap",          # For heatmap visualization
  "ComplexHeatmap",    # For advanced heatmap visualization
  "glmnet",           # For LASSO regression
  "org.Hs.eg.db"      # For gene annotation
))

# Additional CRAN packages
install.packages(c(
  "survival",        # For survival analysis
  "survminer",       # For survival visualization
  "ggplot2",         # For plotting
  "reshape2"         # For data manipulation
))

Data Requirements

The pipeline expects three input files from the TCGA breast cancer dataset:

1. RNA-seq data (data_mrna_seq_v2_rsem.txt)
2. Clinical data (data_clinical_patient.txt)
3. Copy Number Alteration (CNA) data (data_cna.txt)

Pipeline Components

1. Data Loading and Preprocessing

# Load data files
rna_seq <- read.delim("data_mrna_seq_v2_rsem.txt", sep="\t", header=TRUE)
clinical <- read.delim("data_clinical_patient.txt", sep="\t", header=TRUE)
cna <- read.delim("data_cna.txt", sep="\t", header=TRUE)

Key preprocessing steps:

• ID standardization across datasets
• Removal of metadata rows from clinical data
• Handling of missing values
• Sample matching across datasets

2. ERBB2 Status Classification

The pipeline classifies samples based on ERBB2 amplification status:

• Amplified: CNA > 0
• Not Amplified: CNA ≤ 0

3. Differential Expression Analysis

Using DESeq2 for:

• Data normalization
• Differential expression testing
• Variance stabilizing transformation (VST)

4. Visualization Components

PCA Plot

# Generate PCA plot
pca_data <- plotPCA(vst, intgroup="ERBB2_Status", returnData=TRUE)
ggplot(pca_data, aes(PC1, PC2, color=ERBB2_Status)) +
  geom_point(size=3) +
  geom_density2d()

Heatmap Generation

# Generate heatmap of top DE genes
pheatmap(mat, 
         annotation_col=metadata_factors,
         scale="row",
         show_rownames=TRUE)

5. Pathway Analysis

Using clusterProfiler for:

• GO enrichment analysis
• GSEA analysis
• Pathway visualization

6. Survival Analysis

Implements LASSO-regularized Cox regression:

• Patient stratification
• Survival curve generation
• Risk score calculation

Function Descriptions

ID Standardization

standardize_ids_strict <- function(ids) {
  ids <- toupper(ids)                      
  ids <- gsub("[^A-Z0-9.]", ".", ids)     
  ids <- gsub("\\.\\d+$", "", ids)        
  ids <- gsub("\\.+", ".", ids)           
  ids <- sub("\\.$", "", ids)             
  return(ids)
}

Data Filtering

• Removes low count genes (< 10 counts)
• Handles missing values in survival data
• Matches samples across datasets

Output Files

The pipeline generates:

1. Differential expression results
2. PCA plots
3. Heatmaps
4. Pathway enrichment results
5. Survival analysis plots
6. Risk stratification results

Validation

The pipeline includes validation steps:

• Known ERBB2+ signature genes verification
• Data quality checks
• Sample matching verification
• Survival data completeness checks

Usage Example

# Load required libraries
source("required_libraries.R")

# Run analysis pipeline
source("main_analysis.R")

# Generate visualizations
source("visualization.R")

# Perform survival analysis
source("survival_analysis.R")

Notes

• Ensure all input files are in the correct format
• Monitor memory usage with large datasets
• Consider using parallel processing for large-scale analyses
• Verify sample IDs match across all input files

Troubleshooting

Common issues and solutions:

1. Sample ID mismatches: Use the standardize_ids_strict function
2. Memory issues: Filter low-count genes early
3. Missing survival data: Check completeness of clinical data
4. Zero-variance genes: Remove before LASSO regression

Contributing

Feel free to submit issues and enhancement requests!