Table 1 Annotation prediction performance using various branches of data from the GO database. GO branches used includes biological process, cell component, molecular function, and gene ontology. Performance was measured using Pairwise F-score (Fp), BCubed F-score (Fb), and Normalized Mutual Information (NMI). The CliXo, GCN-V, and GCN-V + E models served as baseline measurements and the proposed model, MPGNN-HiLander, resulted in the highest accuracy across all metrics and datasets

From: A graph neural network approach for hierarchical mapping of breast cancer protein communities