AcuGRL: Knowledge-Guided Heterogeneous Graph Representation Learning for Exploring regularity in acupuncture records with auxiliary information

We propose AcuGRL, a graph representation learning-based framework that models the relationships between acupoints and disease phenotypes as a heterogeneous graph. This framework incorporates a domain knowledge-guided scheme to capture both the structural and semantic features of the network, generating effective embeddings for downstream tasks. Additionally, we integrate micro-level genetic targets with macro-level disease phenotypes to further enhance network connectivity and provide richer contextual information.