Supp. Fig. 1 Comparison of feature retrieval mechanisms in Neo4j (A) and relational databases (B).

(A) In Neo4j, accessing all features (properties) of a target node requires first locating the node and subsequently fetching its associated property records. Because properties are stored as a linked list, a high volume of features necessitates the sequential traversal of multiple property blocks. This feature retrieval overhead is repeated for all neighboring nodes in the subgraph. (B) Conversely, relational databases typically store entity features contiguously. As long as the row does not exceed the database page size, all features of a target node can be retrieved in a single read operation. While fetching neighboring nodes still requires database joins, the contiguous storage of features eliminates the need for property block traversals which significantly reduces the required query time for retrieval.