Research On MiRNA-disease Association Prediction Based On Heterogeneous Networks

MiRNA is a kind of non-coding RNA composed of 20-25 nucleotides.In the early days,because of its undetected function,miRNAs were once called "dark matter".With the deepening of research,a large number of experimental data and cases show that miRNAs are closely related to diseases and play an important role in many biological processes necessary for human beings.Therefore,miRNAs are often used as biomarkers to assist doctors in medical diagnosis.Exploring the associations between miRNAs and diseases is of great significance in the field of science and the development of human medical and health undertakings,so it has gradually become an important research topic for researchers.At present,although the association data between miRNAs and diseases can be obtained by traditional biological experimental methods,the experimental cycle of these methods is often very long and requires a lot of human and material resources.As a result,the number of association data obtained is extremely limited.In recent years,under the guidance of emerging bioinformatics,researchers have proposed a large number of more accurate and effective methods to predict the miRNA-diseases associations compared with traditional experimental methods,many of which have achieved remarkable results.In this paper,three methods for predicting the relationship between miRNAs and diseases are proposed by using the functional similarity network of miRNAs,the semantic similarity network of diseases and the association network of miRNAs and diseases.The prediction method based on KATZ model is KATZMDA,the method based on sparse nearest neighbor is SNMDA,and the method based on local linear coding is LLCMDA.Considering that the miRNA functional similarity and the disease semantic similarity are sparse,all the three methods are to reconstruct the miRNAs similarity and disease similarity.KATZMDA model integrated reconstructed similarity information with existing information to construct heterogeneous networks,and then used KATZ algorithm to predict correlation scores.SNMDA and LLCMDA reconstructed similarity information using sparse representation and LLC algorithm respectively,and then the label propagation algorithm was applied to get the relationship between miRNAs and diseases.In the experimental part,we used leave-one-out cross-validation and K-fold cross-validation to test the validity of the proposed methods.Besides,we carried out case study to further verify the effectiveness of our methods.The experimental results show that the three methods proposed in this paper can effectively predict the association between miRNAs and diseases.