Identifying the oligogenic causes of rare diseases remains a challenge, notwithstanding the advancements made in the last decade. While a variety of predictive and ranking approaches have been proposed, their precision remains limited, as it remains difficult to know which features may be most relevant for the design of new predictors. We hypothesize that structured biological information, which provides an integration of relevant biological networks and ontologies in a single heterogeneous knowledge graph, can make a difference as it allows for learning a relevant genetic representation through KGE methods. An exhaustive benchmarking is performed wherein we assess the performance of various state-of-the-art embedding models for the task of identifying potentially pathogenic gene pairs. The results obtained show that these KGEs provide highly accurate predictions, leading to an AUC PR of up to 0.93, representing a significant advancement over previous approaches. We show nonetheless that care needs to be taken in the cross-validation when using embeddings, as data leakage between folds will reveal overly optimistic results. The further evaluation of the methods on a holdout set and on a group of new male infertility cases show that three Translational Distance models (TransE, MurE, RotatE) and two Semantic Matching models (DistMult, QuatE) provide better results. The analysis is concluded by comparing all known gene combinations for these top-ranking models, examining their similarities and differences. Overall, KGEs provide a predictive advancement but new steps will need to be taken to generate explanations as to why the pairs are relevant for oligogenic diseases.

Co-authors: Barbara Gravel, Alexandre Renaux, Ann Nowé, Maris Laan, Tom Lenaerts