SumGNN: multi-typed drug interaction prediction via efficient knowledge graph summarization.

Lucas M Glass,Chao Zhang,Yue Yu,Jimeng Sun,Cao Xiao,Kexin Huang

doi:10.1093/bioinformatics/btab207

Abstract

Thanks to the increasing availability of drug-drug interactions (DDI) datasets and large biomedical knowledge graphs (KGs), accurate detection of adverse DDI using machine learning models becomes possible. However, it remains largely an open problem how to effectively utilize large and noisy biomedical KG for DDI detection. Due to its sheer size and amount of noise in KGs, it is often less beneficial to directly integrate KGs with other smaller but higher quality data (e.g. experimental data). Most of existing approaches ignore KGs altogether. Some tries to directly integrate KGs with other data via graph neural networks with limited success. Furthermore most previous works focus on binary DDI prediction whereas the multi-typed DDI pharmacological effect prediction is more meaningful but harder task. To fill the gaps, we propose a new method SumGNN: knowledge summarization graph neural network, which is enabled by a subgraph extraction module that can efficiently anchor on relevant subgraphs from a KG, a self-attention based subgraph summarization scheme to generate reasoning path within the subgraph, and a multi-channel knowledge and data integration module that utilizes massive external biomedical knowledge for significantly improved multi-typed DDI predictions. SumGNN outperforms the best baseline by up to 5.54%, and performance gain is particularly significant in low data relation types. In addition, SumGNN provides interpretable prediction via the generated reasoning paths for each prediction. The code is available in Supplementary Material. Supplementary data are available at Bioinformatics online.

Highlights

Adverse drug-drug interactions (DDI) are modifications of the effect of a drug when administered with another drug, which is a common and dangerous scenario for patients with complicated conditions
Existing deep learning models are often trained only based on the DDI dataset at hand, ignoring the large biomedical knowledge graph (Ioannidis et al, 2020; Himmelstein and Baranzini, 2015) which can benefit the DDI predictions since DDI is driven by complicated biomedical mechanism
We find that SumGNN achieves the best performance in DDI prediction on two datasets, accurately predicting the correct DDI pharmacological effect consistently

Summary

Introduction

Adverse drug-drug interactions (DDI) are modifications of the effect of a drug when administered with another drug, which is a common and dangerous scenario for patients with complicated conditions. Undetected adverse DDIs have become serious health threats and caused nearly 74, 000 emergency room visits and 195, 000 hospitalizations each year in the United States alone (Percha and Altman, 2013) To mitigate these risks and costs, accurate prediction of DDIs becomes a clinically important task. Knowledge Graph: Over the years, large knowledge graph (KG) such as (Rotmensch et al, 2017), Hetionet (Himmelstein and Baranzini, 2015) and DRKG (Ioannidis et al, 2020) have been constructed from literature mining and database integration These KGs are large and noisy: out of their tens of thousands of nodes with millions of edges, only a small subgraph is relevant to a prediction target. The subgraph formulation allows noise reduction by anchoring on relevant information and is highly scalable since the message passing receptive field is significantly decreased

Methods

Results

Conclusion