cELISA Analysis Repository

A comprehensive R analysis repository for competitive ELISA (cELISA) data. This repository contains statistical analysis scripts and visualization tools for investigating antibody functional activity and correlations.

Repository Structure

cELISA-analysis/
├── R/                           # Analysis scripts
│   ├── setup.R                  # Package management and setup
│   ├── generate_all_figures.R   # Main script to generate all figures
│   ├── multivariate_feature_selection.R # Automatic feature selection analysis
│   ├── pairwise_assay_analysis.R # Pairwise combination analysis
│   ├── tra_correlation_analysis.R # TRA correlation between timepoints
│   ├── odpr_deltaod_correlations.R # Functional vs level correlations
│   ├── tra_deltaod_correlations.R # TRA vs antibody level correlations
│   ├── pairwise_correlation_matrix.R # Simplified correlation matrices
│   ├── univariate_deltaod_analysis.R # Individual assay performance
│   └── roc_analysis.R           # ROC curve analysis for TRA prediction
├── data/                        # Input data files
├── figures/                     # Output plots and figures
├── outputs/                     # Analysis results and statistics
├── cELISA-analysis.Rproj       # RStudio project file
└── README.md                   # This file

Requirements

R Environment

• R version: 4.0 or higher
• RStudio: Recommended for interactive use

R Packages

The analysis automatically installs required packages:

CRAN Packages:

• readxl - Excel file reading
• ggplot2 - Data visualization
• ggpubr - Publication-ready plots
• dplyr - Data manipulation
• tidyr - Data tidying
• corrplot - Correlation matrices
• circlize - Color mapping
• cowplot - Plot arrangements
• pROC - ROC curve analysis
• openxlsx - Excel file writing
• GGally - Pairwise correlation plots
• reshape2 - Data reshaping
• caret - Machine learning and classification
• lme4 - Linear mixed-effects models

Bioconductor Packages (optional):

• ComplexHeatmap - Advanced heatmaps

Quick Start

1. Clone Repository

git clone <repository-url>
cd cELISA-analysis

2. Open in RStudio

Double-click cELISA-analysis.Rproj to open the project in RStudio.

3. Install Dependencies

source("R/setup.R")

4. Prepare Data

Place your Excel data file as data/Compete.xlsx. The file should contain:

• StudyDay columns (252, 560 for Post-Dose 3 and 4)
• TRA values
• Antibody level measurements (FABlog10, eELISAEUlog10, IgG1/3/4DeltaOD, c1qDeltaOD, etc.)
• Subject demographics (Gender, Age) if available

5. Run Analysis

# Generate all figures and tables (main workflow)
source("R/generate_all_figures_tables.R")

# Or run specific analyses:
source("R/multivariate_feature_selection.R")    # Automatic feature selection
source("R/pairwise_assay_analysis.R")           # Pairwise combination analysis  
source("R/tra_correlation_analysis.R")          # TRA timepoint correlations
source("R/roc_analysis.R")                      # ROC curve analysis
source("R/univariate_deltaod_analysis.R")       # Individual assay performance

Analysis Workflow

1. Setup and Package Management (setup.R)

• Centralized package management: Install and load all required packages
• CRAN mirror configuration: Ensures reliable package installation
• Bioconductor support: Optional advanced packages like ComplexHeatmap
• Environment verification: Checks successful package loading

2. Core Statistical Analysis

A. Multivariate Feature Selection (multivariate_feature_selection.R)

• Automatic feature selection: Stepwise selection using AIC criterion
• Bootstrap validation: 50 iterations with 70/30 train-test splits
• Class balancing: Upsampling for TRA>80% prediction
• Performance metrics: AIC, Accuracy, F1, PPV, NPV, AUC with confidence intervals
• Dual timepoint analysis: StudyDay 252 and 560 comparison

B. Pairwise Assay Analysis (pairwise_assay_analysis.R)

• Combination analysis: All 15 pairwise combinations of 6 assays
• Statistical approach: Logistic regression with bootstrap sampling
• Comprehensive metrics: Performance evaluation across all combinations
• 3-significant figure formatting: Standardized result presentation

C. Individual Assay Performance (univariate_deltaod_analysis.R)

• Single assay evaluation: Performance of each assay individually
• Bootstrap sampling: Robust performance estimation
• Comprehensive metrics: Complete performance characterization
• Excel export: Formatted results for publication

3. Correlation Analysis

A. TRA Timepoint Correlations (tra_correlation_analysis.R)

• Longitudinal analysis: TRA correlations between Post-Dose 3 and 4
• Matched subject analysis: Paired timepoint comparisons
• Statistical significance: Spearman correlations with p-values

B. Functional vs Level Correlations (odpr_deltaod_correlations.R)

• ODPR-DeltaOD relationships: Functional activity vs antibody levels
• Multiple assay types: IgG1, IgG3, IgG4, C1q analysis
• Dual timepoint analysis: Both StudyDay 252 and 560
• Publication-ready plots: High-resolution correlation plots

C. TRA-Antibody Correlations (tra_deltaod_correlations.R)

• TRA prediction analysis: Correlation with antibody levels
• Threshold visualization: TRA>80% cutoff analysis
• Multiple predictors: All antibody assays vs TRA

4. Visualization and Matrices (pairwise_correlation_matrix.R)

• Simplified correlation matrices: Using GGally for clean visualization
• Functional vs level separation: Clear variable grouping
• Statistical annotations: Correlation coefficients and significance

Project Context

This repository contains R scripts for statistical analysis of competitive ELISA (cELISA) data from a clinical study investigating antibody responses.

Study Overview

• Analysis focus: IgG subclass and C1q binding data from post-vaccination timepoints
• Data source: Clinical study data from vaccination trials
• Statistical approach: Correlation analysis, ROC analysis and predictive modeling
• Output: Publication-ready figures and statistical results

Code Architecture

• Modular design: Individual R scripts for specific analyses and figures
• Standardized workflows: Consistent data processing across all analyses
• Visualization: ggplot2-based plots with publication-quality themes
• Statistical methods: Bootstrap validation, feature selection, ROC analysis
• Reproducibility: Comprehensive documentation and error handling

Development Principles

• Consistent variable naming conventions across all scripts
• Proper error handling for missing data and edge cases
• Publication-quality plots with standardized formatting
• Complete documentation of statistical methods and assumptions
• Reproducible analysis workflows with version control

Key Features

Statistical Methods

• Statistical approaches: Logistic regression with bootstrap validation
• Stepwise feature selection: AIC-based automatic model selection
• Cross-validation: 70/30 train-test splits with 50 bootstrap iterations
• Class balancing: Upsampling techniques for imbalanced outcomes
• Performance metrics: Comprehensive evaluation (AIC, Accuracy, F1, PPV, NPV, AUC, Sensitivity, Specificity)
• Correlation analysis: Spearman correlations robust to outliers
• ROC analysis: Predictive performance assessment for TRA>80% outcomes

Visualization Features

• Publication-ready plots: High-resolution outputs (300 DPI)
• Consistent styling: Standardized themes and colors
• Statistical annotations: P-values and correlation coefficients
• Flexible functions: Reusable plotting components
• Multiple formats: TIFF, PNG output options

Reproducibility

• R project structure: Self-contained analysis environment
• Automated setup: Package installation and management
• Clear documentation: Comprehensive commenting
• Version control ready: Git integration

Output Files

Figures

• figures/tra_correlation_*.png - TRA correlations between timepoints
• figures/*_vs_*.png - Scatter plots with correlation analysis
• figures/pairwise_correlation_*.png - Correlation matrices
• figures/roc_*.pdf - ROC curves for TRA>80% prediction

Statistical Results

• outputs/multivariate_feature_selection_summary.csv - Feature selection results
• outputs/Pairwise_Day252_sig3.xlsx - Pairwise combination analysis results
• outputs/TRA_correlation_summary.csv - TRA correlation analysis
• outputs/univariate_performance_*.xlsx - Individual assay performance metrics
• outputs/ROC_Summary_*.xlsx - ROC analysis results with AUC confidence intervals

Usage Examples

Run Complete Analysis Pipeline

# 1. Setup environment (run once per session)
source("R/setup.R")

# 2. Run all analyses
source("R/multivariate_feature_selection.R")  # Feature selection
source("R/pairwise_assay_analysis.R")         # Pairwise combinations
source("R/tra_correlation_analysis.R")        # TRA correlations
source("R/univariate_deltaod_analysis.R")     # Individual assays
source("R/roc_analysis.R")                    # ROC analysis

Run Specific Analysis Components

# Feature selection for optimal assay combinations
source("R/multivariate_feature_selection.R")

# Evaluate all pairwise assay combinations
source("R/pairwise_assay_analysis.R")

# Correlation between timepoints
source("R/tra_correlation_analysis.R")

Troubleshooting

Package Installation Issues

If Bioconductor packages fail to install:

# Install BiocManager first
if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")
    
BiocManager::install("ComplexHeatmap")

Data Loading Problems

• Ensure Excel file is named exactly Compete.xlsx
• Check that data contains StudyDay 252 and 560 values
• Verify required columns: StudyDay, TRA, FABlog10, eELISAEUlog10, IgG1DeltaOD, IgG3DeltaOD, IgG4DeltaOD, c1qDeltaOD
• Ensure numeric columns contain only numbers or NA

Analysis Issues

• Insufficient data: Some analyses require minimum sample sizes for bootstrap validation
• Model convergence: Logistic regression may fail with sparse data - check class balance
• Memory usage: Large bootstrap iterations may require more RAM

Memory Issues

For large datasets:

• Process subsets of data
• Use gc() to free memory
• Consider data.table for efficiency

License

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

Citation

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

[To be added]
cELISA Analysis Repository. GitHub: [https://github.com/niaid/cELISA-StatisticalAnalysis]

Support

For questions or issues:

• Open a GitHub issue for code-related problems
• Check R documentation for package-specific help

Primary authors: Yuyan Yi, Jingwen Gu

Contact: [email protected]

---

Note: This repository contains analysis code only. Actual clinical data is not included due to privacy considerations. Users must provide their own data following the specified format.