Biological Knowledge Graphs

AstraZeneca and QOMPLX Inc. are among the named corporate participants in biological knowledge graphs. AstraZeneca-affiliated research examines graph topology and applications to drug repurposing, while QOMPLX patents federated graph systems for privacy-preserving analysis of biological data. Genentech and Novartis also appear among the patent holders in this field.

FACTS identifies 14 startups in the field, including QOMPLX Inc. and Causality Biomodels. QOMPLX patents federated computational graph systems for privacy-preserving genomic analysis across institutions. Causality Biomodels works in biocuration and knowledge extraction, with activity spanning text mining, biological database graph analysis and semantic frameworks for public-health intelligence.

Chinese Academy of Sciences CAS has 231 documents in the recent window, and the University of Chinese Academy of Sciences has 128. CAS-affiliated researchers have developed a rice knowledge graph integrated with an LLM, PertKGE for compound–protein interaction analysis and a multimodal graph combining single-cell transcriptomics with literature evidence. Other active groups include the University of Sichuan and China Agricultural University.

The field has $90,734,198 recorded across 81 grants since 2020. These figures capture grants associated with biological knowledge graphs, not necessarily funding dedicated exclusively to graph technology. The data also lists 30 funding rounds, but the individual rounds include broader life-sciences and infrastructure activity, so they should not be treated as a clean measure of investment in this field.

QOMPLX Inc. is the leading named patent holder, with 10 patents in the data, followed by Burzynski Stanislaw R with 9. Genentech has 4, while Novartis and University Pompeu Fabra each have 3. These counts identify holders associated with the topic; they do not show the scope, status or commercial use of each patent.

Patent activity declined 64.7% in the latest 12-month comparison, while publication activity grew 14.7%. The contrast suggests research is expanding faster than patenting in the measured period. Patent counts alone do not establish commercial readiness, and biological knowledge graph work may also be protected through software, datasets, trade secrets or methods that are not captured by these counts.

Recent work includes KG-Microbe, published in September 2026, for modular microbiome and microbial-science graphs, and CAREPath, published in October 2026, for mechanism-aware drug-repurposing paths. AResKGLM applies graph-grounded language models to antimicrobial-resistance reasoning. Other recent research investigates language-model-generated graph consistency, multimodal graph completion and text-to-Cypher question answering.

The five approaches identified are ontology-based semantic graphs, property graph models, literature-mined graphs, curated database integration and multi-omics graph integration. Ontologies emphasize shared meaning; property graphs offer flexible structures; literature mining scales extraction from papers; curated integration prioritizes reviewed sources; and multi-omics integration connects distinct biological data types. Each approach trades flexibility or scale against consistency, curation effort and data quality.

The United States leads with 25,903 documents, followed by China with 17,998 and the United Kingdom with 9,989. Germany has 8,865 and India has 7,658. These figures reflect documents associated with the topic, rather than a direct measure of research quality, investment or commercial deployment in each country.

The listed applications include pharmacology and safety, drug discovery and repurposing, precision medicine, clinical decision support, and disease and gene discovery. Pharmacology and safety has 4,479 documents, while drug discovery and repurposing has 997. Graphs can connect drugs, genes, diseases and other evidence to help prioritize candidates, explore mechanisms and support clinical reasoning.