Paper Info
Title
Ninimhmda: Neural Integration Of Neighborhood Information On A Multiplex Heterogeneous Network For Multiple Types Of Human Microbe-Disease Association
Abstract
Motivation: Many computational methods have been recently proposed to identify differentially abundant microbes related to a single disease; however, few studies have focused on large-scale microbe-disease association prediction using existing experimentally verified associations. This area has critical meanings. For example, it can help to rank and select potential candidate microbes for different diseases at-scale for downstream lab validation experiments and it utilizes existing evidence instead of the microbiome abundance data which usually costs money and time to generate.Results: We construct a multiplex heterogeneous network (MHEN) using human microbe-disease association database, Disbiome and other prior biological databases, and define the large-scale human microbe-disease association prediction as link prediction problems on MHEN. We develop an end-to-end graph convolutional neural network-based mining model NinimHMDA which can not only integrate different prior biological knowledge but also predict different types of microbe-disease associations (e.g. a microbe may be reduced or elevated under the impact of a disease) using one-time model training. To the best of our knowledge, this is the first method that targets on predicting different association types between microbes and diseases. Results from large-scale cross validation and case studies show that our model is highly competitive compared to other commonly used approaches.Availabilityand implementation: The codes are available at Github https://github.com/yuanjing-ma/NinimHMDA.Contact: yuanjingma2020@u.northwestern.edu or hongmei@northwestern.eduSupplementary information: Supplementary data are available at Bioinformatics online.
Year
DOI
Venue
2021
10.1093/bioinformatics/btaa1080
BIOINFORMATICS
DocType
Volume
Issue
Journal
36
24
ISSN
Citations 
PageRank 
1367-4803
1
0.37
References 
Authors
0
2
Name
Order
Citations
PageRank
Yuanjing Ma110.71
Hongmei Jiang292.74