R4RA Molecular Analysis

This repository contains scripts to support the publication: 'Deep Molecular Pathology Profiling of Synovial Tissue Biopsies in the R4RA Randomised Clinical Trial Identifies Predictive Signatures of Response/Resistance to Biologic Therapies in Rheumatoid Arthritis'

Repository Structure

.
├── 1_R4RA_power_calculation
├── 2_R4RA_PCA
├── 3_Baseline_clustering_M3C
├── 4_DEG_crossover
├── 5_Subset_Module_Scores
├── 6_WTA_GeoMx_QC_Normalization_DEG
├── 7_R4RA_longitudinal_pathway_analysis
└── 8_CDAI50_response_prediction
    ├── 1_Data_Exploration
    ├── 2_Machine_Learning_binary
    └── 3_Plotting_and_summary_scripts

• 1_R4RA_power_calculation: statistical power calculations to ensure that RNA-Seq gene expression studies were conducted with enough sequencing depth and sample size.
• 2_R4RA_PCA (Supplementary Figure 1): Principal component analysis and plots
• 3_Baseline_clustering_M3C (Figure 2a & b): heatmaps using montecarlo concensus clustering with the M3C package
• 4_R4RA_crossover_analysis (Figures 2b-g, 3a-e):
  • differential expression comparing response for each treatment (DESeq2_crossover_TOC_RTX.R)
  • modular differential expression using qusage (modular_analysis_withcovariates.R, load_mods_qmod.R, qmod_AS.R)
  • 3-way differential gene expression analysis on baseline synovial biopsies of patients that switched drug over time. Based on their response to different treatments, patients were classified as Pro-RTX, Pro-TOC, and Refractory (R4RA_crossover_analysis.R).
• 5_Subset_Module_Scores (Figure 3h): cell subset module scores
• 6_WTA_GeoMx_QC_Normalization_DEG (Figure 4b, d, e): Digital Spatial Profiling (DSP) data analysis using DESeq2 for preprocessing and DEG analysis.
• 7_R4RA_longitudinal_pathway_analysis (Figure 5g-i): longitudinal differential gene expression analysis using mixed-effects model. Genes that showed expression change over time were enriched via pathway analysis.
• 8_CDAI50_response_prediction (Figure 6): analysis to predict CDAI50% response to rituximab and tocilizumab using clinical variables and gene expression.

Additional software packages

For the purposes of this publication (Figure 5a-f) an R package, glmmSeq, was designed to model gene expression with a general linear mixed model (glmm). This is publicly available on CRAN. The source code can be found here.

Data exploration website

The supporting website is available at: https://r4ra.hpc.qmul.ac.uk/