LncRNA-miRNA Interaction Prediction Based On Information Completion

Long noncoding RNAs(lnc RNAs)are a type of transcribed RNA molecules with a length of more than 200 nucleotides that do not encode proteins.Lnc RNAs are one of the crucial components in the gene regulatory network,and play central roles in many biological processes,and are also associated with some complex and difficult diseases.However,currently the specific functions of most lnc RNAs are vague for researchers.In general,lnc RNAs interact with other biological macromolecules when they exert their biological functions.Recently,some studies have found that lnc RNA can act as a decoy or sponge to regulate the behavior of mi RNAs.At the same time,mi RNAs often appear in the molecular mechanism of lnc RNAs.Therefore,actively exploring the lnc RNA-mi RNA interaction can greatly improve the understanding of lnc RNA function.Traditional wet laboratory methods can identify whether an lnc RNA-mi RNA pair interacts or not,but they are time-consuming,labor-intensive and expensive,which makes research progress slow.With the increasing accumulation of related data,although machine learning technology has been introduced into the prediction of lnc RNA-mi RNA interaction,only a few models have been published,and there is still room for improvements,for example: the similarity measurement method used is not effective,requires too many additional features as auxiliary information,and cannot predict lnc RNA or mi RNA without any interaction records.In order to solve above problems,this paper proposed a model called sequencederived neighborhood weighted propagation method with information completion(SNWPM-IC)to predict the interactions between lnc RNAs and mi RNAs.Its main innovations are neighborhood weighted similarity and information completion strategies.The general processes of SNWPM-IC are as follows: first,SNWPM-IC fully exploits lnc RNA sequences,mi RNA sequences and known interactions to calculate lnc RNA-lnc RNA similarities and mi RNA-mi RNA similarities by using neighborhood weighted similarity;then SNWPM-IC integrates multiple lnc RNA-lnc RNA similarities and multiple mi RNA-mi RNA similarities respectively by using information completion strategies,and constructs the integrated lnc RNA similarity-based graph and the integrated mi RNA similarity-based graph;finally,the label propagation process is executed on each of the above two graphs to score lnc RNA-mi RNA pairs,and SNWPM-IC adopts the linear combination of their outputs as final predictions.A series of experimental results demonstrate that the SNWPM-IC model can predict lnc RNA-mi RNA interaction more accurately than other state-of-the-art methods.