GLKB API

A researcher investigating 1,500 candidate genes cannot read 15,000 abstracts, but automating with a general LLM trades reliability for speed. GLKB addresses both: send a list of genes, variants, or drug targets and get back cited answers, grounded in 263M+ biomedical terms and live PubMed.

263M+

Biomedical Terms

14.6M+

Relationships

Why GLKB

Scale

Process hundreds of genes, variants, or drug targets in parallel. 1,500 genes in ~30 minutes.

Reliability

Every claim linked to a PubMed ID. Gaps labeled "none found," never fabricated.

Structured Output

Define a schema in your prompt. Unsupported fields are left empty, not guessed.

Reproducibility

Same Prompt + Same List -> Consistent, parseable output every time.

What you can build

Gene Prioritization

1,500 candidate genes from a screen or GWAS

Literature score + evidence + PMIDs per gene

DEG Annotation

800 differentially expressed genes from RNA-seq

Function, disease association, stress direction per gene

Drug Target Dossiers

50 candidate kinase targets

Mechanism, human genetics, known modulators, clinical stage

GWAS Locus Interpretation

200 fine-mapped locus genes

Functional evidence, model organism phenotype, priority tier

Variant Interpretation

Gene:variant strings from exome sequencing

Domain context, pathogenic variant literature, ClinVar context

Proteomics Annotation

400 proteins from a mass-spec experiment

Known function, interaction partners, perturbation evidence

Knowledge base

The GLKB knowledge graph is built from the full PubMed corpus and encodes gene-disease associations, pathway memberships, co-occurrence statistics, and ontology hierarchies across 263 million biomedical terms and 14.6 million relationships.