Context-specific biological network construction

Abstract

Biological network shows various compositions of biological relations responds to various biomedical contexts such as disease and drug. Therefore, construction of context-specific biological network is very important problem to discover underlying biological principles respect to biomedical contexts. In recent studies, the approach is emerged that construct context-specific network by collecting biological relations related to context of interest from massive public biological database and literatures. Therefore, finding contexts related to context of interest is important problem of the research. However, there were limitations in previous researches that they only find limited range of contexts and they only considered pathological feature from ontology. In this research, two strategies were employed to overcome these limitations. Firstly, continuous similarity for all context term pairs were measured and biological relations which have context with higher similarity than minimum similarity threshold were included in context-specific network. Secondly, molecular biological feature from similarity between context-associated protein-protein interaction networks was employed and integrated to pathological similarity from ontology. Performance of proposed method was evaluated by F-score comparison and ROC curve and showed higher performance than previous methods. Context-specific biological network from this research will promote the research for underlying biological principle of biomedical contexts.