DROMA.R: Drug Omics Association Analysis Extension for DROMA.Set

Overview

DROMA.R is an R package that provides advanced analysis functions for drug-omics associations using DromaSet and MultiDromaSet objects from the DROMA.Set package. It supports meta-analysis of drug-omics associations across multiple datasets, comprehensive visualization tools, and batch processing of features. This package extends DROMA.Set with statistical analysis capabilities for biomarker discovery in precision medicine. All data loading functions now apply z-score normalization by default for improved analysis consistency.

CTRDB Support: Includes specialized functions for analyzing Clinical Trial Database (CTRDB) data, enabling patient-level drug response analysis and cross-drug signature stratification.

Core Design Principles

1. Modular Architecture: Functions organized into data loading, pairing, meta-analysis, and visualization modules
2. Z-score Normalization: All continuous data normalized by default for cross-dataset comparability
3. Statistical Rigor: Appropriate methods for continuous (Spearman) and discrete (Wilcoxon + Cliff's Delta) data
4. Flexible Analysis: Supports both single dataset and multi-dataset meta-analysis workflows

It is a part of DROMA project. Visit the official DROMA website for comprehensive documentation and interactive examples.

Features

• 🔗 DROMA.Set Integration: Works seamlessly with DromaSet and MultiDromaSet objects
• 📊 Meta-analysis: Advanced statistical analysis across multiple datasets
• 🏥 Clinical Trial Database (CTRDB): Specialized functions for clinical drug response analysis
• 🎨 Comprehensive Visualization: Forest plots, volcano plots, comparison plots with consistent theming
• ⚡ Batch Processing: Efficient analysis of multiple features simultaneously
• 🧮 Multiple Statistical Methods: Spearman correlation, Wilcoxon tests, Cliff's Delta effect sizes
• 🚀 Performance Optimization: Parallel processing support for large datasets
• 🔄 Z-score Normalization: All data loading functions apply z-score normalization by default

Installation

Prerequisites

First, install the DROMA.Set package:

# Install DROMA.Set (replace with actual installation method)
# devtools::install_github("mugpeng/DROMA_Set")

Install DROMA.R

# Install devtools if not already installed
if (!requireNamespace("devtools", quietly = TRUE)) {
  install.packages("devtools")
}

# Install DROMA.R from GitHub
devtools::install_github("mugpeng/DROMA_R")

Quick Start

1. Load Required Packages

library(DROMA.Set)  # For data management
library(DROMA.R)    # For analysis functions

2. Create DromaSet Objects

# Connect to DROMA database
connectDROMADatabase("path/to/your/droma.sqlite")

# Create a single DromaSet for one project
gCSI <- createDromaSetFromDatabase("gCSI", "path/to/droma.sqlite")

# Create a MultiDromaSet for multiple projects
multi_set <- createMultiDromaSetFromDatabase(
    project_names = c("gCSI", "CCLE"),
    db_path = "path/to/droma.sqlite"
)

4. Analyze Drug-Omics Associations

# Single project analysis
result <- analyzeDrugOmicPair(
  gCSI,
  select_omics_type = "mRNA",
  select_omics = "ABCB1",
  select_drugs = "Paclitaxel",
  data_type = "all",
  tumor_type = "all"
)

# Multi-project meta-analysis
multi_result <- analyzeDrugOmicPair(
  multi_set,
  select_omics_type = "mRNA",
  select_omics = "ABCB1",
  select_drugs = "Paclitaxel",
  overlap_only = FALSE
)

# Display results
print(result$meta)
plot(result$plot)

5. Batch Feature Analysis

# Find genes associated with drug response
batch_results <- batchFindSignificantFeatures(
  multi_set,
  feature1_type = "drug",
  feature1_name = "Paclitaxel",
  feature2_type = "mRNA",
  overlap_only = FALSE
)

# Create volcano plot
volcano_plot <- plotMetaVolcano(batch_results, es_t = 0.3, P_t = 0.05)
print(volcano_plot)

Core Functions by Module

🔧 Data Loading Module (Z-score Normalized by Default)

• loadMolecularProfiles(): Load molecular profiles with automatic z-score normalization
• loadTreatmentResponse(): Load drug response data with automatic z-score normalization
• loadMultiProjectMolecularProfiles(): Load multi-project molecular profiles
• loadMultiProjectTreatmentResponse(): Load multi-project drug response data
• applyZscoreNormalization(): Apply z-score normalization to existing data
• isZscoreNormalized(): Check if data has been z-score normalized

🔗 Data Pairing Module

• pairDrugOmic(): Pair continuous drug and omics data
• pairDiscreteDrugOmic(): Pair discrete omics with drug data
• pairContinuousFeatures(): General function for pairing continuous features
• pairDiscreteFeatures(): General function for pairing discrete and continuous features

📊 Meta-Analysis Module

• metaCalcConCon(): Meta-analysis for continuous vs continuous data (Spearman correlation)
• metaCalcConDis(): Meta-analysis for continuous vs discrete data (Wilcoxon + Cliff's Delta)
• analyzeContinuousDrugOmic(): Wrapper for continuous drug-omics analysis
• analyzeDiscreteDrugOmic(): Wrapper for discrete drug-omics analysis

🎨 Visualization Module

• createForestPlot(): Forest plots for meta-analysis results
• plotMetaVolcano(): Volcano plots for batch analysis results
• plotContinuousDrugOmic(): Scatter plots with correlation statistics
• plotDiscreteDrugOmic(): Box plots for group comparisons

🚀 High-Level Analysis Functions

• analyzeDrugOmicPair(): Complete workflow for single drug-omics pair analysis
• batchFindSignificantFeatures(): Batch screening of multiple features
• processDrugData(): Process drug sensitivity data with normalization
• getDrugSensitivityData(): Combined processing and annotation
• analyzeStratifiedDrugOmic(): Stratified analysis by another drug's response
• createStatisticalDashboard(): Interactive dashboard for statistical results
• generateStatisticalPlots(): Generate comprehensive statistical overview plots

🏥 Clinical Trial Database (CTRDB) Module

• analyzeClinicalDrugResponse(): Analyze clinical drug response with omics data
• analyzeStratifiedCTRDB(): Stratified analysis across different drugs
• analyzeClinicalMeta(): Meta-analysis across clinical datasets
• getClinicalSummary(): Summary statistics for clinical analysis
• getStratifiedCTRDBSummary(): Summary for stratified CTRDB analysis

Note: getPatientExpressionData() has been moved to DROMA.Set package (CTRDB_SQLManager.R) for better separation of database operations and analysis functions.

🛠 Utility Functions

• bright_palette_26: Pre-defined color palette for visualizations
• formatTime(): Format processing time outputs
• estimateTimeRemaining(): Estimate batch processing time
• formatDrugTable(): Format drug sensitivity data for display
• annotateDrugData(): Add clinical annotations to drug data
• loadFeatureData(): Load feature data for batch analysis
• filterFeatureData(): Filter features by minimum sample size

Examples

Example 1: Loading Data with Z-score Normalization

# Load required packages
library(DROMA.Set)
library(DROMA.R)

# Create DromaSet
gCSI <- createDromaSetFromDatabase("gCSI", "path/to/droma.sqlite")

# Load mRNA data with z-score normalization (default)
mrna_normalized <- loadMolecularProfiles(
  gCSI,
  molecular_type = "mRNA",
  features = "ABCB1"
)

# Load without normalization
mrna_raw <- loadMolecularProfiles(
  gCSI,
  molecular_type = "mRNA",
  features = "ABCB1",
  zscore = FALSE
)

# Check if data is normalized
cat("Normalized:", isZscoreNormalized(mrna_normalized))
cat("Raw:", isZscoreNormalized(mrna_raw))

# Load drug data with normalization
drug_normalized <- loadTreatmentResponse(
  gCSI,
  drugs = "Paclitaxel"
)

Example 2: Multi-Project Normalized Loading

# Create MultiDromaSet
multi_set <- createMultiDromaSetFromDatabase(c("gCSI", "CCLE"))

# Load normalized data across projects
multi_mrna <- loadMultiProjectMolecularProfiles(
  multi_set,
  molecular_type = "mRNA",
  features = "ABCB1",
  overlap_only = FALSE
)

# Load normalized drug data across projects
multi_drugs <- loadMultiProjectTreatmentResponse(
  multi_set,
  drugs = "Paclitaxel",
  overlap_only = FALSE
)

# Check normalization status for each project
for (project in names(multi_mrna)) {
  cat(project, "normalized:", isZscoreNormalized(multi_mrna[[project]]), "\n")
}

Example 3: Single Project Analysis

# Load required packages
library(DROMA.Set)
library(DROMA.R)

# Create DromaSet
gCSI <- createDromaSetFromDatabase("gCSI", "path/to/droma.sqlite")

# Analyze Paclitaxel vs ABCB1 expression
result <- analyzeDrugOmicPair(
  gCSI,
  select_omics_type = "mRNA",
  select_omics = "ABCB1",
  select_drugs = "Paclitaxel"
)

# View results
print(result$meta)
plot(result$plot)

Example 4: Multi-Project Meta-Analysis

# Create MultiDromaSet
multi_set <- createMultiDromaSetFromDatabase(c("gCSI", "CCLE"))

# Analyze across projects with overlapping samples
result <- analyzeDrugOmicPair(
  multi_set,
  select_omics_type = "mutation_gene",
  select_omics = "TP53",
  select_drugs = "Cisplatin",
  overlap_only = FALSE
)

# Create forest plot
createForestPlot(result$meta)

Example 5: Batch Analysis

# Find all genes associated with Paclitaxel response
batch_results <- batchFindSignificantFeatures(
  multi_set,
  feature1_type = "drug",
  feature1_name = "Paclitaxel",
  feature2_type = "mRNA",
  cores = 4  # Use parallel processing
)

# Sort by significance
sorted_results <- batch_results[order(batch_results$p_value), ]
print(head(sorted_results, 10))

# Create volcano plot
volcano_plot <- plotMetaVolcano(batch_results)
print(volcano_plot)

Example 6: Stratified Analysis

# Analyze drug response stratified by another drug's sensitivity
# This helps identify context-dependent biomarkers

stratified_result <- analyzeStratifiedDrugOmic(
  dromaset_object = multi_set,
  stratification_drug = "Cisplatin",     # Drug for stratification
  select_omics_type = "mRNA",            # Omics data type
  select_omics = "ERCC1",                # Target gene
  select_drugs = "Bortezomib",           # Drug to analyze
  stratify_by = "response_median",       # Stratification method
  tumor_type = "all"
)

# View stratified results
print(stratified_result$statistics)
print(stratified_result$comparison)

Example 7: Drug Sensitivity Data Processing

# Process drug data with annotations
drug_data <- getDrugSensitivityData(
  dromaset_object = gCSI,
  drug_name = "Paclitaxel",
  data_type = "all",
  tumor_type = "all",
  db_path = "path/to/droma.sqlite"  # For loading annotations
)

# View processed data
head(drug_data)

# Create sensitivity rank plot
formatDrugTable(drug_data, drug_name = "Paclitaxel")

Example 8: Clinical Trial Database (CTRDB) Analysis

# Connect to CTRDB database
con <- connectCTRDatabase("path/to/ctrdb.sqlite")

# Analyze clinical drug response
result <- analyzeClinicalDrugResponse(
  select_omics = "EGFR",
  select_drugs = "Erlotinib",
  data_type = "all",
  tumor_type = "all",
  connection = con,
  meta_enabled = TRUE
)

# View individual patient plots
print(result$plot)

# View meta-analysis forest plot
print(result$forest_plot)

# Get summary statistics
summary <- getClinicalSummary(result)
print(summary)

Example 9: Stratified CTRDB Analysis

# Stratified analysis: Drug B signature applied to Drug A
result <- analyzeStratifiedCTRDB(
  drug_b_name = "Cisplatin",      # Signature generation
  drug_a_name = "Paclitaxel",     # Signature application
  select_omics = "EGFR",          # Feature to analyze
  connection = con,
  top_n_genes = 100,
  data_type = "all",
  tumor_type = "all"
)

# View signature genes
print(result$signature_genes)

# View forest plot with meta-analyzed correlations
print(result$correlation_results$forest_plot)

# View combined scatter plots
print(result$correlation_results$combined_scatter_plot)

# Get comprehensive summary
summary <- getStratifiedCTRDBSummary(result)
print(summary)

Data Types Supported

Molecular Profiles

• mRNA: Gene expression data
• cnv: Copy number variation data
• mutation_gene: Gene-level mutation data
• mutation_site: Site-specific mutation data
• fusion: Gene fusion data
• meth: DNA methylation data
• proteinrppa: Reverse-phase protein array data
• proteinms: Mass spectrometry proteomics data

Treatment Response

• drug: Drug sensitivity/response data

Statistical Methods

Continuous vs Continuous Analysis

When analyzing relationships between two continuous variables (e.g., gene expression vs drug response):

• Spearman Correlation: Non-parametric rank correlation coefficient
• Fisher's Z-transformation: Converts correlation coefficients to normally distributed values for meta-analysis
• Random-Effects Model: Combines results across multiple studies accounting for heterogeneity

Continuous vs Discrete Analysis

When comparing continuous values between discrete groups (e.g., mutated vs wild-type):

• Wilcoxon Rank-Sum Test: Non-parametric test for group differences
• Cliff's Delta: Effect size measurement for group differences (range: -1 to 1)
  • |d| < 0.147: Negligible effect
  • 0.147 ≤ |d| < 0.33: Small effect
  • 0.33 ≤ |d| < 0.474: Medium effect
  • |d| ≥ 0.474: Large effect

Meta-Analysis Approach

1. Within-study analysis: Calculate effect sizes and p-values for each dataset
2. Quality control: Filter results based on sample size and data quality
3. Cross-study combination: Use random-effects meta-analysis to combine results
4. Heterogeneity assessment: Evaluate variability between studies

Performance Tips

1. For large-scale batch analyses, use cores > 1 to enable parallel processing
2. Consider setting overlap_only = TRUE with MultiDromaSet when you need comparable sample analyses across all datasets
3. Use the db_path parameter with annotateDrugData() or getDrugSensitivityData() to automatically load sample annotations from the database when needed
4. For visualization of large datasets, consider filtering to specific tumor types or data types

Contributing

Contributions to DROMA.R are welcome! Please feel free to submit issues or pull requests on GitHub.

License

This project is licensed under the MPL-2 License - see the LICENSE file for details.

Citation

If you use DROMA.R in your research, please cite:

Li, S., Peng, Y., Chen, M. et al. Facilitating integrative and personalized oncology omics analysis with UCSCXenaShiny. Commun Biol 7, 1200 (2024). https://doi.org/10.1038/s42003-024-06891-2

Contact

For questions and feedback, please contact Peng Yu Zhong at [email protected].

---

DROMA.R - Advanced drug-omics association analysis powered by DROMA.Set 🧬💊📊