Skip to content

R_PACKAGE_MANAGEMENT.md

Latest commit

 

History

History
138 lines (94 loc) · 4.61 KB

R_PACKAGE_MANAGEMENT.md

File metadata and controls

138 lines (94 loc) · 4.61 KB

R Package Management - Implementation Summary

Overview

All R packages are now pre-installed in the Docker image, eliminating runtime installations that were causing slowdowns and redundant package installs during API calls.

Package Management Strategy

Using BiocManager for Version Compatibility

Instead of manually specifying exact versions for every package, we're using BiocManager as the package manager, which provides several key benefits:

  1. Automatic Compatibility: By setting Bioconductor version 3.21, BiocManager automatically ensures all Bioconductor packages are compatible with each other
  2. Dependency Resolution: BiocManager handles CRAN packages and resolves dependencies that work well with the Bioconductor ecosystem
  3. Less Version Conflicts: No need to manually track and update dozens of specific version numbers
  4. Easier Maintenance: Future updates are simpler - just update the Bioconductor version and rebuild

Dockerfile Changes

Old Approach (Problematic)

  • Used remotes::install_version() for specific versions of each package
  • Required manual version management for 20+ packages
  • Prone to version conflicts between CRAN and Bioconductor packages
  • More complex and harder to maintain

New Approach (Recommended)

# Set Bioconductor version 3.21 - this manages compatibility
BiocManager::install(version = '3.21', ask = FALSE, update = TRUE)

# Install Bioconductor packages - version managed by BiocManager
BiocManager::install(c('WGCNA', 'DESeq2', 'limma', 'biomaRt', 'sva',
                       'STRINGdb', 'apeglm', 'impute'), ask = FALSE)

# Install CRAN packages - BiocManager ensures Bioconductor compatibility
BiocManager::install(c('tidyverse', 'Rtsne', 'umap', 'ggplot2', 'readr',
                       'ape', 'mice', 'dplyr', 'gplots', 'ggVennDiagram',
                       'pheatmap', 'RColorBrewer', 'stringr', 'here', 'lme4'),
                     ask = FALSE)

R Scripts Modified

All runtime package installation code has been removed from the following files:

Core Scripts

  • api/code/req_packages.R - Now just loads pre-installed packages
  • api/code/micro_functions.R - Replaced install_and_load() with direct library() calls
  • api/code/Microarray_Updated_Workflow.R - Simplified package loading

Venn Diagram Scripts

  • api/code/1_init_venn_m.R - Removed installation function
  • api/code/3_plot_venn_m.R - Removed installation function
  • api/code/4_wide_frame_venn_m.R - Removed installation checks

Other Analysis Scripts

  • api/code/5_Up_Down_m.R - Removed ggplot2 installation check
  • api/code/2_plot_heatmap.R - Removed pheatmap/RColorBrewer installation
  • api/code/1_Biomart_init_m.R - Removed biomaRt installation function

STRING Workflow

  • api/string/1_setup_string_env.R - Removed install_and_load() function

Configuration

  • api/code/.Rprofile - Added safeguards to prevent runtime installations

Benefits

  1. Performance: No more waiting for package installations during API calls
  2. Reliability: Packages are tested and verified during Docker build, not at runtime
  3. Reproducibility: Docker image contains exact package versions
  4. Version Compatibility: BiocManager ensures all packages work together
  5. Simpler Maintenance: Easier to update packages in the future

Expected Package Versions (Bioconductor 3.21)

When using Bioconductor 3.21 with R 4.4+, you'll get approximately:

  • WGCNA: ~1.73
  • DESeq2: ~1.44.0
  • limma: ~3.60.0+
  • biomaRt: ~2.60.0+
  • sva: ~3.52.0
  • STRINGdb: ~2.16.0+
  • tidyverse: ~2.0.0
  • And compatible versions of all other packages

Note: Exact versions may vary slightly but BiocManager ensures compatibility

Docker Build Command

To rebuild the Docker image with the new package setup:

docker build -t plagl1-server .
docker run --name plgl-server -p 8000:8000 -d plagl1-server:latest

Or use the command from the Dockerfile:

docker rm -f plgl-server && docker build -t wonderful_rubin . && docker run --name plgl-server -p 8000:8000 -d wonderful_rubin:latest

Testing

After building the new Docker image, test that all packages load correctly:

# In R console within container
library(WGCNA)
library(DESeq2)
library(limma)
library(biomaRt)
library(sva)
library(STRINGdb)
library(tidyverse)
# ... etc

All packages should load without any installation prompts or errors.

Future Updates

To update R packages in the future:

  1. Change the Bioconductor version in Dockerfile (e.g., from '3.21' to '3.22')
  2. Rebuild the Docker image
  3. Test the application

BiocManager will automatically pull compatible versions of all packages.