FAIR Data Principles — Findable, Accessible, Interoperable, Reusable

Overview

Guidelines for making scientific data FAIR: Findable, Accessible, Interoperable, and Reusable.

Findable

• Assign globally unique persistent identifiers (DOIs) to datasets
• Rich metadata describing the dataset (title, authors, description, keywords, dates)
• Metadata registered in searchable resources (DataCite, re3data, FAIRsharing)
• Data indexed in domain-specific repositories

Accessible

• Data retrievable by identifier using standardized protocol (HTTP, FTP)
• Metadata accessible even if data is restricted
• Authentication/authorization where necessary, clearly documented
• Long-term preservation plan (minimum 10 years for funded research)

Interoperable

• Use formal, shared vocabularies (ontologies: GO, ChEBI, EFO, MeSH)
• Standard file formats (CSV, JSON, HDF5, NetCDF — not proprietary)
• Include references to related datasets and publications
• Machine-readable metadata (JSON-LD, Dublin Core, schema.org)

Reusable

• Clear data usage license (CC-BY, CC0 recommended for scientific data)
• Detailed provenance (how data was collected, processed, quality controlled)
• Meet community standards (MIAME for microarrays, MINSEQE for sequencing)
• Version control for datasets that evolve

Recommended Repositories

Domain	Repository
General	Zenodo, Figshare, Dryad
Genomics	GEO, SRA, ENA
Proteomics	PRIDE, MassIVE
Structures	PDB, EMDB
Clinical	ClinicalTrials.gov, YODA
Chemistry	ChEMBL, PubChem
Materials	NOMAD, Materials Cloud