Relevance search for predicting lncRNA–protein interactions based on heterogeneous network

• Protein–protein information is introduced to help in solving lncRNA–protein interaction prediction problem.
• The heterogeneous lncRNA–protein network is constructed.
• An algorithm which can measure relevance between heterogeneous objects is used to predict lncRNA–protein interaction.

lncRNA plays important roles in many biological and pathological processes. lncRNA–protein interaction is the most common way of lncRNA performing their functions. Thus, predicting lncRNA–protein interaction is very significant to understand the nature of lncRNA. Earlier methods to predict RNA–protein interaction always adopt classification algorithms using features extracted from RNA and protein themselves. But their performance is not good enough in lncRNA–protein interaction prediction because of lncRNA’s low conservation. In this paper, we try to use information implicit in the topologies of biological network associated with lncRNA to solve this problem. Firstly we construct a heterogeneous lncRNA–protein network which incorporate protein interaction information. We use an algorithm called HeteSim which can evaluate relevance between heterogeneous objects to perform relevance search in the lncRNA and protein heterogeneous network. The relevance search results are used to help the identification of lncRNA–protein interactions.