Creating R Research Projects

This skill helps create and manage a complete R-based research analysis workflow. It is designed for scientific computing, statistical modeling, bioinformatics, and data visualization tasks.

Use this skill when the user wants to:

• Analyze datasets using R
• Perform statistical tests or modeling
• Run bioinformatics or omics analysis in R
• Generate plots, figures, or reports
• Create a reproducible R project structure
• Install and manage R package dependencies

---

What This Skill Does

When activated, this skill will:

1. Create a structured R project

   • data/ for raw and processed data
   • scripts/ for analysis code
   • results/ for outputs
   • reports/ for R Markdown or Quarto reports

2. Set up environment

   • Initialize .Rproj (if using RStudio)
   • Create renv environment for reproducibility
   • Install required CRAN/Bioconductor packages

3. Generate analysis scripts

   • Data loading and cleaning
   • Statistical analysis or modeling
   • Visualization with ggplot2
   • Save outputs (CSV, plots, model summaries)

4. Create a report

   • R Markdown / Quarto document
   • Includes methods, results, and figures
   • Render to HTML or PDF

---

Example User Requests That Should Trigger This Skill

• "Use R to analyze this CSV and generate plots"
• "Run differential expression analysis in R"
• "Create a statistical report for this dataset"
• "Build an R project for microbiome analysis"
• "Fit a regression model in R and summarize results"

---

Example Workflow

User: Analyze this gene expression dataset and produce figures.

Skill actions:

• Create project structure
• Install tidyverse, DESeq2, ggplot2
• Write analysis script
• Generate PCA plot and volcano plot
• Produce an HTML report

---

Tools & Packages Commonly Used

Purpose	R Packages
Data wrangling	tidyverse, data.table
Visualization	ggplot2, patchwork
Statistics	stats, lme4, survival
Bioinformatics	Bioconductor packages (DESeq2, edgeR, limma)
Reporting	rmarkdown, quarto
Reproducibility	renv

---

Notes

• Prefer reproducible workflows (renv, scripted analysis)
• Avoid interactive-only steps unless requested
• All outputs should be saved to files, not just printed to console