JMIR Publications

ABSTRACT

Background:

Extracting the interactions between chemicals and proteins from the biomedical literature is important for many biomedical tasks such as drug discovery, medicine precision and knowledge graph construction. Several computational methods have been proposed for automatic chemical-protein interaction (CPI) extraction. However, most of the proposed models could not effectively learn semantic and syntactic information from complex sentences in biomedical texts.

Objective:

To relieve the problem, we propose a method to effectively encode syntactic information from long text for CPI extraction.

Methods:

Due to the ability of capturing syntactic information from dependency graphs, graph convolutional networks (GCNs) have recently drawn increasing attention in natural language processing. To investigate the performance of the GCN on CPI extraction, this paper proposes a novel model based on the GCN. The model can effectively capture sequential information and long-range syntactic relations between words by using dependency structure of input sentences.

Results:

We evaluated our model on the ChemProt corpus released by BioCreative VI and it achieves an F-score of 65.17%, which is 1.07% higher than that of the state-of-the-art system.

Conclusions:

Our model can obtain more information from dependency graph than previous proposed models. Experimental results suggest that it is competitive to the state-of-the-art methods and significantly outperforms other methods on the ChemProt corpus, which is the benchmark dataset for CPI extraction.