AutoRel (Auto-Relevant)

AutoRel is a tool designed to identify biologically relevant genes from RNA-Seq differential expression analysis, going beyond simple statistical significance.

🚀 Quick Start (CLI)

1. Install as an R Package (Directly from GitHub):

   # In R:
   devtools::install_github("jochotecoa/AutoRel")

2. Run via CLI:

   Rscript run_AutoRel.R --counts counts.csv --results res.csv --metadata meta.csv --contrast Group --control Control --report

3. Verify installation (Optional):

   Rscript test_installation.R

🛠️ Developer Usage

You can use AutoRel directly inside your R scripts:

library(AutoRel)

results <- run_prioritization(
  norm_counts = my_counts,
  res = my_deseq2_results,
  coldata = my_metadata,
  contrast_group = "Group",
  control_level = "Control",
  model_path = system.file("extdata", "autorrel.rds", package="AutoRel"),
  output_path = "results/"
)

📦 Model Versioning

This tool uses GitHub Releases to manage model versions. This ensures that the code version you use is always paired with the correct trained model.

For Maintainers: How to Release a New Model

To update the model hosted on GitHub:

1. Tag the commit: git tag v0.1.0
2. Push the tag: git push origin v0.1.0
3. Create a Release on GitHub and upload the autorrel.rds file as a binary asset.

The install_AutoRel.R script will then automatically pull the model from that release.

🐳 Docker Usage

Docker allows you to run AutoRel without worrying about R dependencies.

1. Build the image:

   docker build -t autorell .

2. Run the analysis: (Mount your local data folder to /app/data in the container)

   docker run -v /path/to/your/data:/app/data autorell --counts data/counts.csv --results data/res.csv --metadata data/meta.csv --contrast Group --control Control

CLI Arguments

Argument	Description
-c, --counts	Path to normalized counts CSV (gene names in 1st column)
-r, --results	Path to DESeq2 results CSV (gene names in 1st column)
-m, --metadata	Path to coldata/metadata CSV (sample names in 1st column)
-g, --contrast	Column name in metadata used for grouping
-l, --control	(Optional) Name of the control level
-o, --output	(Optional) Output directory (default: output/autorrel_results)
--report	(Optional) Generate an automated HTML analysis report

📊 Visual Reporting

When using the --report flag, AutoRel generates a professional HTML report containing:

• Summary findings: Total genes prioritized as relevant.
• Interactive Tables: Filterable list of relevant genes.
• Relevance vs. Significance Plot: A volcano-style visualization highlighting prioritized genes.
• Top Genes Heatmap: Scaled expression profiles of the top 20 prioritized genes.

Project Structure

• run_AutoRel.R: Main Command Line Interface.
• install_AutoRel.R: Script to install dependencies and download the model.
• scripts/:
  • autorrel/: Core pipeline and model logic.
  • analysis/: Pipeline and model comparison scripts.
  • case_studies/: Specific applications (APAP, Colorectal Cancer, etc.).
• models/: Storage for the trained autorrel.rds model.
• data/: Input datasets and example files.
• output/: Results, figures, and model exports.

Requirements

• R (>= 4.0)
• DESeq2
• Tidyverse, Caret, Magrittr, Optparse

---

Developed by Juan Ochoteco Asensio