Skip to content
Knowledge Base

Recommendations for FAIR Data Analysis

The following practices support FAIR (Findable, Accessible, Interoperable, Reusable) data analysis. They apply regardless of workflow format or tooling, and help ensure that your work can be replicated, reproduced, or reused – by yourself or others, in the future or in new contexts.

Follow the KISS principle – “Keep It Single-Step”.

Section titled “Follow the KISS principle – “Keep It Single-Step”.”
Section titled Follow the KISS principle – “Keep It Single-Step”.
  • Each script should ideally perform only one distinct task.
  • Avoid mixing unrelated processes in a single script.
  • This improves:
    • Reusability across projects
    • Understandability of the script’s purpose
    • Clarity around its inputs and outputs- A focused script is easier to reuse, understand, and annotate, especially regarding the inputs it consumes and the outputs it creates

Separate data analysis logic from data

Section titled “Separate data analysis logic from data”
Section titled Separate data analysis logic from data
  • Avoid hard-coded paths or filenames.
  • Parameterize inputs, outputs, and configuration as arguments.
  • Benefits:
    • Easier reuse across datasets and platforms
    • Smoother integration into larger workflows or pipelines

Collect information about dependencies

Section titled “Collect information about dependencies”
Section titled Collect information about dependencies
  • Record the script language (e.g. Python, R, Bash)
  • List all external packages, libraries, and software tools used
  • Specify versions where applicable
  • List runtime environments and hardware requirements (if relevant)
  • Include useful metadata:

Uncouple processes from the execution environment

Section titled “Uncouple processes from the execution environment”
Section titled Uncouple processes from the execution environment
  • Ensure that your analysis can be run anywhere, not just on your machine.
  • Use containerization (e.g., Docker, Singularity) to bundle:
    • Dependencies
    • Correct versions
    • Required system tools
  • Use existing community-maintained containers where possible, e.g.
  • This makes your analysis portable and reproducible across different systems.
Section titled “Resources and recommended reads”
Section titled Resources and recommended reads

This guide is adapted and includes recommendations from: