justold/immunopipe-rpkgs repository overview

⁠immunopipe-rpkgs

Intermediate Docker image for Immunopipe containing custom R packages built from source.

⁠Overview

immunopipe-rpkgs is the second layer in Immunopipe's multi-stage Docker build system. This image extends immunopipe-base by adding custom R packages that are not available in standard conda/CRAN repositories, providing specialized visualization and analysis capabilities for single-cell genomics.

⁠Purpose

This intermediate image serves as:

• Layer 2 in a 3-tier build strategy (base → rpkgs → immunopipe)
• Custom R package layer containing GitHub-sourced packages built from source
• Build optimization layer that separates slow-changing custom packages from frequently updated pipeline code

⁠Contents

This image includes all dependencies from immunopipe-base plus the following custom R packages:

⁠Custom R Packages

• gglogger - Enhanced logging utilities for ggplot2-based workflows
• enrichit - Gene set enrichment analysis utilities
• plotthis - Comprehensive visualization toolkit for single-cell data
• scplotter - Specialized single-cell plotting functions
• biopipen.utils - Utility functions for the biopipen ecosystem

All packages are installed from source and compiled during the build process to ensure compatibility with the base environment.

⁠Build Process

Built using a multi-stage approach:

1. Builder stage: Installs packages from environment_rpkgs.yml using micromamba
2. Final stage: Copies only compiled R libraries from builder to immunopipe-base, avoiding metadata bloat
3. Cleanup: Removes conda package cache and build artifacts

Automated builds: Triggered via GitHub Actions when:

• environment_rpkgs.yml changes
• Dockerfile.rpkgs changes
• Base dependencies change (environment_base.yml or Dockerfile.base)

All triggers require changes on the dev branch.

⁠Usage

⁠As a Base for immunopipe

FROM justold/immunopipe-rpkgs:latest

COPY --chown=$MAMBA_USER:$MAMBA_USER . /immunopipe

# Install Python pipeline code with Poetry
RUN python -m poetry install --no-cache -v --all-extras

⁠For Development/Testing

docker pull justold/immunopipe-rpkgs:latest
docker run -it justold/immunopipe-rpkgs:latest R

# Test custom packages in R
> library(plotthis)
> library(scplotter)

⁠Build Strategy Rationale

Why separate custom R packages?

1. Build time optimization: R packages from source can take 30-60 minutes to compile; separating them from frequently-changed Python code reduces rebuild times
2. Layer caching: Custom R packages change infrequently compared to pipeline code
3. Size efficiency: Multi-stage build avoids including conda metadata and build tools in final layer

⁠Image Size

Adds ~1-2GB to the base image size (total ~6-10GB).

⁠Maintenance

• Source: immunopipe/docker/environment_rpkgs.yml⁠
• Dockerfile: immunopipe/Dockerfile.rpkgs⁠
• Update frequency: Changes when custom R packages need version updates or new packages are added

⁠Package Sources

Custom R packages are typically sourced from:

• pwwang's GitHub repositories (gglogger, plotthis, scplotter)
• Bioconductor development versions
• Specialized bioinformatics tools

⁠Related Images

• justold/immunopipe-base (parent): Base system and conda dependencies
• justold/immunopipe (child): Complete pipeline with Python code

⁠License

See the Immunopipe repository⁠ for license information.

⁠Support

For issues or questions:

• GitHub Issues: https://github.com/pwwang/immunopipe/issues⁠
• Documentation: https://pwwang.github.io/immunopipe/⁠