PLS-DA Analysis for MS/MS IDA Data

Complete implementation of Partial Least Squares Discriminant Analysis (PLS-DA) for tandem mass spectrometry data analysis.

---

Overview

This repository provides a comprehensive, production-ready implementation of PLS-DA specifically designed for MS/MS (Mass Spectrometry) IDA (Information Dependent Acquisition) data analysis. Perfect for metabolomics, proteomics, and biomarker discovery workflows.

Key Features

• Complete PLS-DA implementation using NIPALS algorithm
• Zero dependencies - uses only base R
• Cross-validation for optimal model selection
• Variable Importance (VIP) scores for biomarker discovery
• 7 publication-ready visualizations
• 4 detailed CSV exports for further analysis
• Comprehensive documentation with examples
• Example dataset included for immediate testing

---

Quick Start

Installation

# Clone the repository
git clone https://github.com/[your-username]/PLSDA-MSMS-Analysis.git
cd PLSDA-MSMS-Analysis

Run the Example

# In R console
source("plsda_analysis.R")

# Or from command line
Rscript plsda_analysis.R

That's it! The script will:

1. Generate example MS/MS data (150 samples, 50 features, 3 classes)
2. Perform PLS-DA analysis with cross-validation
3. Create 7 plots and 4 data files
4. Display comprehensive results summary

Runtime: ~5-10 seconds for example data

---

Example Output

The analysis automatically generates:

Visualizations (PNG)

1. Scores Plot - Sample clustering and group separation
2. Test Predictions - Model validation on held-out data
3. VIP Scores - Top 20 most important features
4. Loadings Plot - Feature contribution to separation
5. CV Results - Model optimization curve
6. Confusion Matrix - Classification performance heatmap
7. Variance Explained - Component importance

Data Files (CSV)

1. Predictions - Sample classifications with scores
2. VIP Scores - All features ranked by importance
3. Model Summary - Performance metrics (accuracy, R²X, R²Y)
4. Loadings - Feature contributions to each component

---

Documentation

For Beginners

• QUICK_START.md - Get running in 5 minutes
• Sections 1-5 of README_FULL.md - Basic concepts and usage

For Advanced Users

• README_FULL.md - Complete 2000+ line documentation including:
  • Statistical theory and mathematical background
  • Using your own data (step-by-step guide)
  • Parameter tuning and optimization
  • Troubleshooting common issues
  • Best practices for publication
  • Advanced topics (permutation tests, bootstrap, etc.)

---

Use Cases

Perfect for:

• Metabolomics - Identify metabolite biomarkers between groups
• Proteomics - Discover differentially expressed proteins
• Lipidomics - Classify samples based on lipid profiles
• Biomarker Discovery - Rank features by discriminative power
• Quality Control - Validate analytical methods
• Method Development - Optimize sample preparation protocols

---

Why This Implementation?

Advantages of PLS-DA for MS/MS Data

Feature	Benefit
Handles high dimensionality	Works with 1000s of m/z features
Manages multicollinearity	Correlated features (common in MS data)
Supervised classification	Uses group labels for maximum separation
Interpretable results	VIP scores, loadings, and scores plots
Robust to noise	Dimensionality reduction filters noise
No external dependencies	Pure R implementation

Why Not Just Use PCA?

• PCA is unsupervised (ignores group labels)
• PLS-DA maximizes separation between known groups
• PLS-DA provides biomarker rankings (VIP scores)
• PLS-DA is designed for classification tasks

---

Requirements

• R version ≥ 4.0.0
• No additional packages required!
• Memory: Minimum 4GB RAM (8GB recommended)
• Storage: ~10MB for outputs

---

Methodology

This implementation uses:

• NIPALS algorithm for PLS component extraction
• Stratified train-test split (70-30 default)
• K-fold cross-validation for component optimization
• VIP scores for feature importance ranking
• Dummy matrix encoding for multi-class problems

See README_FULL.md Section 13 for complete statistical background.

---

Using Your Own Data

Your CSV should have this structure:

SampleID,Class,Batch,mz_200.00,mz_250.00,mz_300.00,...
S001,Control,1,1234.5,2345.6,3456.7,...
S002,Treatment,1,5678.9,6789.0,7890.1,...

See README_FULL.md Section 9 for detailed integration guide.

Quick integration:

# Load your data
my_data <- read.csv("your_msms_data.csv")

# Extract features and labels
feature_cols <- grep("^mz_", colnames(my_data), value = TRUE)
features <- as.matrix(my_data[, feature_cols])
labels <- factor(my_data$Class)

# Continue with script from Section 2 (Preprocessing)

---

Contributing

Contributions are welcome! Areas for improvement:

• Additional preprocessing methods
• Support for more data formats
• Integration with pathway analysis tools
• Additional visualization options
• Performance optimizations

Please open an issue or submit a pull request.

---

Citation

If you use this code in your research, please cite:

This repository:

AyebBlk. (2025). PLS-DA Analysis for MS/MS IDA Data. 
GitHub: https://github.com/AyehBlk/PLSDA-MSMS-Analysis

Original PLS-DA method:

Barker, M., & Rayens, W. (2003). Partial least squares for discrimination. 
Journal of Chemometrics, 17(3), 166-173.

NIPALS algorithm:

Wold, S., Sjöström, M., & Eriksson, L. (2001). PLS-regression: 
a basic tool of chemometrics. Chemometrics and Intelligent Laboratory 
Systems, 58(2), 109-130.

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

You are free to:

• Use for academic research
• Use for commercial projects
• Modify and distribute
• Include in your own projects

Just include the license and attribution!

---

Troubleshooting

Common Issues

Problem: Low accuracy (<70%)

• Check if classes are actually separable (biological question)
• Ensure sufficient samples (minimum 20 per class recommended)
• Try different preprocessing methods

Problem: Error "Singular matrix"

• Features are too highly correlated
• Remove features with correlation >0.95

Problem: All VIP scores <1

• May need fewer components
• Check if preprocessing is appropriate
• Verify classes are actually different

See README_FULL.md Section 12 for complete troubleshooting guide.

---

Support

• Check the comprehensive documentation
• Read the FAQ
• Open an issue for bugs
• Start a discussion for questions

---

🌟 Star History

If you find this useful, please consider giving it a star! ⭐

---

🔗 Related Resources

• MetaboAnalyst - Web-based metabolomics analysis
• KEGG - Pathway database
• mixOmics R package - Extended multivariate methods
• ropls Bioconductor - Alternative PLS-DA implementation

---

👤 Author

Ayeh Bolouki

• GitHub: @AyehBlk
• Role: Computational Biologist / Bioinformatician

---

Project Status

Active Development - Maintained and open to contributions

Current Version: 1.0 Last Updated: October 2025

---

Made with ❤️ - Let's make free science for everybody around the world.

_{If this helped your research, consider citing it in your publications!}