This document discusses using a knowledge graph and machine learning techniques to generate hypotheses for clinical therapies for NGLY1 deficiency, a rare genetic disease. It describes biocuration efforts to capture the pathophysiology and phenotypic spectrum of NGLY1 deficiency and add this information to a knowledge graph. The document proposes using the updated knowledge graph with an edge prediction algorithm called Rephetio to identify potential drug repurposing candidates for treating NGLY1 deficiency.