Fig. 4From: A knowledge graph approach to predict and interpret disease-causing gene interactionsMetapath influence on rule confidence Metapaths significantly associated with higher rule confidence are shown. Rule confidence distributions are compared for each metapath, considering both its presence and absence as a rule condition. The significance of the difference is determined using a one-tailed Wilcoxon ranksum test with Bonferroni correction (adjusted p-value \(\le 0.01\)). Metapaths are ranked based on their effect size, measured by the rank-biserial correlation. A Rules mined from paths including Phenotype (incl. Pheno). B Rules mined from paths excluding Phenotype (excl. Pheno)Back to article page