A Bipartite Network Module-Based Project to Predict Pathogen-Host Association.

Jie Li,Shiming Wang,Yadong Wang,Zhuo Chen

doi:10.3389/fgene.2019.01357

Abstract

Pathogen–host interactions play an important role in understanding the mechanism by which a pathogen can infect its host. Some approaches for predicting pathogen–host association have been developed, but prediction accuracy is still low. In this paper, we propose a bipartite network module-based approach to improve prediction accuracy. First, a bipartite network with pathogens and hosts is constructed. Next, pathogens and hosts are divided into different modules respectively. Then, modular information on the pathogens and hosts is added into a bipartite network projection model and the association scores between pathogens and hosts are calculated. Finally, leave-one-out cross-validation is used to estimate the performance of the proposed method. Experimental results show that the proposed method performs better in predicting pathogen–host association than other methods, and some potential pathogen–host associations with higher prediction scores are also confirmed by the results of biological experiments in the publically available literature.

Highlights

Pathogen–host interactions (PHIs) play a crucial role in understanding the mechanisms of infections and identifying potential targets for infection therapeutics
Each known pathogen–host interaction is chosen as a test data set in turn, the remaining known interactions are chosen as the training set, and the pathogen– host association score in the training set is calculated using BNMP
To clarify the influence of the balance parameter x, area under the ROC curve (AUROC) and area under the PR curve (AUPR) values were calculated with different values of x, as shown in Figures 2A and B

Summary

Introduction

Pathogen–host interactions (PHIs) play a crucial role in understanding the mechanisms of infections and identifying potential targets for infection therapeutics. Various biological experimental or computing methods have been developed to test and predict the interactions between pathogens and hosts. It is time-consuming and laborious to test PHIs through biological experimentation and costs a lot of money. Computing methods such as biological reasoning and machine learning are considered as another important approach for predicting PHIs. Three main approaches can be used to predict PHIs: biological reasoning homology-based, structure-based, and domain/motif interaction-based (Nourani et al, 2015).

Methods

Results

Conclusion