Path count (PC) measures how many path instances match a specified metapath between defined source and target nodes. Each matching path contributes 1, so PC reports raw prevalence without adjusting for graph connectivity. section 4.3
Use path count as a metapath-based feature in a graph-analysis pipeline:
- Define source and target node types, then choose a metapath describing the permitted node and relationship sequence. section 4.3
- Extract matching path instances and count them for each source–target pair, including pairs without a direct connection. section 4.3 section 10.2.2
- Interpret raw counts cautiously: highly connected intermediate nodes can inflate PC without indicating a stronger or more meaningful relationship. section 10.2.2
- Use degree-weighted path count when hub bias matters; it weights paths inversely by intermediate-node degrees and emphasizes more specific pathways. section 10.2.2
Drug–disease feature
For metformin and type 2 diabetes, count paths matching CbGaD: Compound–binds–Gene–associates–Disease. The resulting PC becomes one feature for that compound–disease pair; the same query can be adapted to other metapaths and pairs. section 10.2.2
