Cross-Medicine Knowledge Graph for Type 2 Diabetes Drug Discovery

Zekun Zhou published a cross-medicine knowledge graph (CMKG) for Type 2 Diabetes Mellitus (T2DM) drug discovery on figshare in May 2026. The graph contains 15 entity types (245,235 entities) and 52 relation types (7,155,373 triples), integrating data from sources like Hetionet, SymMap, TCMBank, STRING, and TTD. It was constructed to enable interpretable prioritization of drug candidates using graph embedding models and a unified path scoring framework.

Use Cases

• Link prediction for potential drug-disease relationships based on the integrated biomedical knowledge graph.
• Interpretable path scoring to explain drug candidate mechanisms based on the described hybrid confidence calibration.
• Benchmarking graph embedding models (TransE, DistMult, ComplEx, RotatE) on multi-source biomedical data.
• Prioritizing traditional Chinese medicine herbs or modern compounds for T2DM based on the unified scoring system.

Strengths

• Contains 7,155,373 triples connecting 245,235 entities, providing a substantial network for analysis.
• Integrates data from five distinct biomedical sources (Hetionet, SymMap, TCMBank, STRING, TTD), likely increasing coverage.
• The underlying study reports performance metrics for the ComplEx model (MRR = 0.213 ± 0.004, Hits@10 = 0.418 ± 0.003).
• Focuses on 709 core Type 2 Diabetes Mellitus genes, providing a specific disease context.

Limitations

• The dataset is very small (9.5 KB), suggesting it may only contain a list of triple types or high-level metadata, not the full knowledge graph.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for graph learning tasks.

Provenance

Source

Integrated from Hetionet, SymMap, TCMBank, STRING, and TTD.

Collection Method

Entity alignment and relation consolidation using Jaccard and overlap-based fusion strategies.

Freshness

Last updated 2026-05-13 17:33:18; freshness should be verified.