graph-level anomaly detection has gained significant attention as it finds
many applications in various domains, such as cancer diagnosis and enzyme
prediction. However, existing methods fail to capture the underlying properties
of graph anomalies, resulting in unexplainable framework