Evaluation of knowledge graph embedding approaches for drug-drug interaction prediction in realistic settings
Published in BMC Bioinformatics, 2009
Recommended citation: Celebi, R., Uyar, H., Yasar, E. et al. Evaluation of knowledge graph embedding approaches for drug-drug interaction prediction in realistic settings. BMC Bioinformatics 20, 726 (2019). https://doi.org/10.1186/s12859-019-3284-5 https://bmcbioinformatics.biomedcentral.com/track/pdf/10.1186/s12859-019-3284-5.pdf
We aimed to present realistic evaluation settings to predict drug-drug interactions (DDIs) using knowledge graph embeddings. Methods for prediction of DDIs have tendency to report high accuracy but yet there has been little impact on translational research due to systematic biases induced by networked/paired data.
In the study, we propose a simple disjoint cross-validation scheme to evaluate drug-drug interaction predictions for the scenarios where the drugs have no known DDIs. Our contribution can be summarized as follows : i) comparison of different knowledge graph embedding approaches on DDI prediction task ii) evaluation of different knowledge graphs as background knowledge for feature learning iii) testing DDI prediction task for the cold-start scenarios.
Recommended citation: Celebi, R., Uyar, H., Yasar, E. et al. Evaluation of knowledge graph embedding approaches for drug-drug interaction prediction in realistic settings. BMC Bioinformatics 20, 726 (2019). https://doi.org/10.1186/s12859-019-3284-5