Multi-domain knowledge graph embeddings for gene-disease association prediction

Abstract

Background: Predicting gene-disease associations typically requires exploring diverse sources of information as well as sophisticated computational approaches. Knowledge graph embeddings can help tackle these challenges by creating representations of genes and diseases based on the scientific knowledge described in ontologies, which can then be explored by machine learning algorithms. However, state-of-the-art knowledge graph embeddings are produced over a single ontology or multiple but disconnected ones, ignoring the impact that considering multiple interconnected domains can have on complex tasks such as gene-disease association prediction.

Results: We propose a novel approach to predict gene-disease associations using rich semantic representations based on knowledge graph embeddings over multiple ontologies linked by logical definitions and compound ontology mappings. The experiments showed that considering richer knowledge graphs significantly improves gene-disease prediction and that different knowledge graph embeddings methods benefit more from distinct types of semantic richness.

Conclusions: This work demonstrated the potential for knowledge graph embeddings across multiple and interconnected biomedical ontologies to support gene-disease prediction. It also paved the way for considering other ontologies or tackling other tasks where multiple perspectives over the data can be beneficial. All software and data are freely available.

Keywords: Gene-disease association prediction; Knowledge graph; Knowledge graph embeddings; Machine learning; Ontologies.

Fig. 1

Example of a direct relationship between hearing loss and the EPS8LA Gene

Fig. 2

Overview of the methodology with four basic steps: 1) build the knowledge graph with ontologies and annotations; 2) create embeddings to represent each gene and disease; 3) produce a final vector of the pairs in the dataset; 4) gene-disease association prediction

Fig. 3

Example of a logical definition of the class Human Phenotype ontology class for “Hearing impairment” (HP:0000365): ’Hearing impairment’ EquivalentTo ’has part’ some (’decreased rate’ and (’inheres in’ some ’sensory perception of sound’) and (’has modifier’ some ’abnormal’))

Fig. 4

Example of a Logical definition simplified with a more direct relation between two classes. The HP term for “Hearing impairment” (HP:0000365) is related to a restriction that involves the GO term “Sensory perception of sound” (GO:0007605)

Fig. 5

ROC curves and AUC values obtained for different vector operators with RF classifier for the HP-simple + LD + GO

Fig. 6

Recall-Precision diagram including f-measure values as height-lines. The diagram uses all knowledge graphs for OPA2Vec and RDF2Vec with RF using a 70-30 split

Fig. 7

Computational time for each embedding method with two knowledge graphs where the smallest size corresponds to removing the main branch of the human phenotype ontology (Phenotypic abnormality)