UniProt
UniProt (the Universal Protein Resource) is the reference catalogue of what we know about proteins across every branch of life. Each entry pairs an amino-acid sequence with structured annotations: what the protein does, where it sits in the cell, the pathways it takes part in, the diseases linked to its variants, post-translational modifications, and cross-references out to hundreds of other biological databases. It splits into two tiers you should treat differently. Swiss-Prot is the manually reviewed, expert-curated core, high in confidence but smaller; TrEMBL is the automatically annotated bulk, which is where the figure of hundreds of millions of sequences comes from. For anything a model will present as fact, prefer the reviewed set.
You have several ways in. The REST API lets you fetch single entries or run queries and return results as JSON, XML, FASTA, or RDF, which is handy for enriching a pipeline on demand. For building a retrieval index you will more often want the bulk downloads: the full Swiss-Prot flat file is a manageable few gigabytes, whereas the complete TrEMBL set is very large, so filter by organism or by the proteomes you care about rather than ingesting everything. There is also a SPARQL endpoint if you want to query it as linked data.
For RAG in the life sciences, UniProt gives you a trustworthy backbone of protein facts to ground answers and cut down on invented gene names or functions. It suits drug-discovery assistants, literature tools, and any system that needs to resolve a protein identifier to a description. A practical tip: the annotation fields are structured, so you often get better retrieval by turning each entry into a short natural-language summary before embedding, rather than indexing the raw record.
The data is released under CC BY 4.0, so you are free to reuse it, including in commercial products, as long as you credit UniProt. Alongside it, PubChem and ChEMBL cover the small-molecule side, and pairing UniProt targets with those bioactivity databases is a common combination.
Related sources
AlphaFold Protein Structure Database
EMBL-EBI and Google DeepMind's open database of AI-predicted protein structures, covering over 200 million sequences across almost all of UniProt. Each model ships with per-residue confidence and predicted aligned error scores, and is reachable by website, API, FTP and Google Cloud bulk access. Released under CC BY 4.0 and designed to pair with UniProt and the experimental structures in the PDB.
ClinicalTrials.gov
A US National Institutes of Health database of privately and publicly funded clinical studies. It holds structured records on more than 450,000 studies, including protocols, conditions, interventions, outcomes, and results, all in the public domain.
DrugBank
A freely accessible database that combines detailed drug data with drug target information, covering more than 15,000 drug entries. It links chemistry, pharmacology, and biology in one place, which makes it a strong knowledge source for medical and pharmaceutical RAG.
MIMIC-III / MIMIC-IV
Freely accessible critical care databases with de-identified health data from more than 40,000 intensive care patients at Beth Israel Deaconess Medical Center. They include vital signs, lab results, medications, and clinical notes. Access needs credentialing and a data-use agreement.