Phenologs at knowledge-graph scale: a mouse-to-slime-mold catalog of candidate genes for thousands of human diseases

A Monarch Knowledge Graph phenologs analysis identifies 61,243 phenologs linking 3,752 human diseases to 9,700 model-organism phenotypes across seven species, using a 5% FDR.

• Why it matters: Bridging model-organism gene–phenotype annotations to human disease is difficult, and phenologs provide a chance-overlap framework to nominate candidate disease genes.
• What they did: The study computes phenologs from Monarch using PANTHER protein families as orthogroups and estimates a 5% false discovery rate from 1,000 randomized networks. It ranks nominated candidate genes with a hypergeometric scheme and compares it to a naive Bayes scheme.
• The result: Both ranking methods recover held-out genes at similar rates, but only the hypergeometric score enables cross-disease comparison; top 100 candidates per disease recover 9 of 73 added associations, 23.5× random expectation.