GenAI Tools for Coding & Research Workflows — A Practical 8-Step Process

Welcome to this repository containing a practical textbook on integrating Large Language Models (LLMs) into coding and research workflows. This guide is geared toward biostatisticians, bioinformaticians, and data scientists, providing hands-on steps and examples to enhance productivity and maintain rigorous standards when employing GenAI tools.

Repository Overview

• Chapters

  1. Chapter 1: Laying the Foundation — Plan & Context
  2. Chapter 2: Knowledge is Power — Do Your Research
  3. Chapter 3: The AI Assistant — Prompt LLM to Generate Code
  4. Chapter 4: Critical Eye — Review & Understand the Code
     • Introduces Git-based version control practices for managing LLM-generated code iterations.
  5. Chapter 5: Making it Better — Refine Code & Add Features
  6. Chapter 6: The Iterative Journey — Iterate Until Satisfied
  7. Chapter 7: Standardization — Refactor to Lab Template
     • Explains using the R Code Refactoring Prompt and an R Code Template to enforce lab-wide coding standards.
  8. Chapter 8: Sharing Your Work — Generate Documentation
     • Demonstrates how to leverage the GitHub Repository Documentation Guidelines for creating robust project docs.
  • Conclusions
    • A concise, optimized conclusion summarizing the eight-step process and emphasizing the importance of version control, collaboration, and reproducibility.

• Docs

  • Appendix A: Recommended LLM Tools and Platforms
  • Appendix B: Prompt Library for Biostatistics, Bioinformatics, and Data Science
  • Appendix C: Further Reading and Resources
  • Appendix D: Project Plans and Code Templates
    • Directs you to the individual plan and template files under docs/templates.
  • References
    • Hyperlinked citations used throughout the textbook.

• Templates (in docs/templates/)

  • R Code Templates & Prompts
    • R_CodeTemplate.R: A production-quality script template featuring metadata, library management, CLI parsing, and more.
    • R_CodeRefactoringPromptExample.md: Guidance for prompting an LLM to refactor your existing code according to lab standards.
  • GitHub Documentation Guide
    • GitHubRepoDocumentationGuidelines.md: A prompt and template for creating comprehensive README.md files, environment setup instructions, and usage docs.
  • Project Plans
    • ProjectPlan_InitialDraft.md
    • ProjectPlan_Improved.md
    • ProjectPlan_Advanced.md
  • Coding Prompt
    • CodingPromptForRScript.md: An example prompt for generating production-ready R scripts with data.table and optparse.

How to Use This Repository

1. Follow the 8-Step Process
   Read the chapters in order to learn how to plan, prompt, refine, and document your AI-assisted code effectively.

2. Leverage the Appendices

   • Check out the recommended LLM tools, curated prompts, further reading, and code templates to accelerate your workflow.

3. Explore the Templates

   • Use the R templates, refactoring prompts, and documentation guidelines in docs/templates/ to standardize your code and streamline collaboration.

4. Apply Version Control

   • As described in Chapter 4, store each LLM-generated or refined code iteration in Git (e.g., GitHub) to maintain a clear history, review diffs, and revert if needed.

5. Stay Tuned for References

   • All references are hyperlinked and listed in docs/References.md.

Data Simulation Script

If you'd like to test your workflows or the 8-step process on example data, we provide a script that simulates EHR and demographic records with realistic data quality issues:

• Script: scripts/simulate_ehr_data.R
• Purpose: Generates BMI, height, weight, and demographic data for a specified number of individuals, optionally introducing missing or implausible values.

Example Usage

Rscript scripts/simulate_ehr_data.R \
  --output_ehr "./data/raw/ehr_bmi_simulated_data.tsv" \
  --output_ehr_dict "./data/raw/data_dictionary.txt" \
  --output_demo "./data/raw/demographics_simulated_data.tsv" \
  --output_demo_dict "./data/raw/demographics_data_dictionary.txt" \
  --seed 123 \
  --n_individuals 1000

Explanation of Arguments:

• --output_ehr: Where to save the simulated EHR (BMI) data (TSV).
• --output_ehr_dict: Where to write the EHR data dictionary (TXT).
• --output_demo: Where to save the demographics data (TSV).
• --output_demo_dict: Where to write the demographics data dictionary (TXT).
• --seed: Random seed for reproducibility.
• --n_individuals: Number of unique individuals to simulate (default=1000).

By running this script, you can quickly produce synthetic data for testing code-refinement prompts, checking your data-cleaning pipelines, or practicing the entire 8-step process from ingestion to documentation.

Contributing

• We welcome pull requests for refinements, corrections, or extensions.
• Please open issues for any questions, or join discussions to keep this textbook accurate and helpful.

License

This repository is licensed under the GNU General Public License v3.0.

---

Happy learning and coding with GenAI tools!