MechRepoNet with DrugCentral Indications Knowledge Graph

Mechanistic Repositioning Network with Indications (MIND) is a knowledge graph incorporating two biomedical resources: MechRepoNet and DrugCentral. MechRepoNet is a knowledge graph comprising of 18 biomedical resources that reflects and expands on important drug mechanism relationships identified from a curated biomedical drug mechanism dataset. MechRepoNet consists of 9,652,116 edges, 250,035 nodes, 9 node types and 22 relations. DrugCentral is a curated, open-access, online resource that integrates structure, bioactivity, regulatory, pharmacologic actions and approved drug indications from the FDA and other regulatory agencies. As of October 2021, DrugCentral has 11,292 indications, consisting of 2,494 unique compounds and 1,459 unique diseases.

MIND was created by mapping DrugCentral edges to existing MechRepoNet nodes through the Unified Medical Language System (UMLS) and Medical Subject Headings (MeSH). A total of 5,379 indication edges made up of 1,308 unique compounds and 1,030 unique diseases were mapped to MechRepoNet. DrugCentral edges incorporated in MIND are labeled have the relationship indication. The indication edge is differentiated from the treat edge in that the latter represents a weaker link between a drug and a disease as some treat edges are not approved by a regulatory body in contrast to the former. In MIND, when both indication and treat edges exist, indication superseded and replaced treat edges. The MIND training set has a total of 9,651,040 edges. 

The indications from DrugCentral were divided into a train set (80%; 4303 triples) and a test/valid set (20%; 1076 triples; test.txt/valid.txt). The DrugCentral train set was merged into MechRepoNet to form the train.txt file.