Towards Incorporating Entity-specific Knowledge Graph Information in Predicting Drug-Drug Interactions

TL;DR

This paper introduces BERTKG-DDI, a method that enhances drug-drug interaction prediction by integrating entity-specific knowledge graph embeddings with BioBERT-based relation classification, achieving a 4.1% F1-score improvement.

Contribution

It presents a novel approach combining knowledge graph embeddings with language model embeddings for improved drug interaction prediction.

Findings

BERTKG-DDI outperforms baseline models by 4.1% in macro F1-score.

Incorporating KG embeddings improves relation classification accuracy.

The method demonstrates the benefit of integrating domain knowledge with language models.

Abstract

Off-the-shelf biomedical embeddings obtained from the recently released various pre-trained language models (such as BERT, XLNET) have demonstrated state-of-the-art results (in terms of accuracy) for the various natural language understanding tasks (NLU) in the biomedical domain. Relation Classification (RC) falls into one of the most critical tasks. In this paper, we explore how to incorporate domain knowledge of the biomedical entities (such as drug, disease, genes), obtained from Knowledge Graph (KG) Embeddings, for predicting Drug-Drug Interaction from textual corpus. We propose a new method, BERTKG-DDI, to combine drug embeddings obtained from its interaction with other biomedical entities along with domain-specific BioBERT embedding-based RC architecture. Experiments conducted on the DDIExtraction 2013 corpus clearly indicate that this strategy improves other baselines architectures by 4.1% macro F1-score.