Differential Network-based LncRNA Biomarker Discovery

Long noncoding RNA(Lnc RNA) accounts for a large part of the human genome and has complex structures as well as diverse functions. As a consequence, lnc RNA has become a focus of bioinformatics. There have been plenty of researches showed that lnc RNA involves in many biological processes and has complex relationships with numerous diseases. Biomarker has been used in biomedicine for a long time and plays a great deal of important role in clinical medicine such as disease diagnosis, treatment and so on. So far, many biomarkers are coding genes. Studies about lnc RNA biomarkers are rare.Most of the studies about lnc RNA biomarkers find lnc RNA biomarkers through methods of differential expression or multiplicity using lnc RNA expression data and confirm lnc RNA biomarkers in samples by Real-Time PCR. The number of lnc RNA biomarkers they find are small. The procedure of confirmation is complex and expensive. Because the methods of both fold change and different expression only consider the differences on the lnc RNA expression profiles, this paper constructs a differential network using lnc RNA's expression profiles. The differential network considers not only the differences on the lnc RNA expression, but also the function similarity between lnc RNAs. Then we extract the biomarkers of long noncoding RNA from the differential network.To verify the effectiveness of differential network, we use the lnc RNA biomarkers extracted from the differential network to class the disease subtypes and conduct some comparative experiments. We apply the method of different expression on the same data set to extract lnc RNA biomarkers as well as class disease subtypes. In the terms of the classification accuracy, the method of differential network is superior to the method of different expression. Also, the number of lnc RNA biomarkers extracted by the method of differential network is less than the method of different expression, indicating a lower medical cost. In order to verify the generality of the method of differential network, we conduct the same experiment procedure on the data set of coding gene and get the same conclusion. Then we make a cross-over analysis. The results about the classification accuracy obtained from the lnc RNA's expression profiles and the coding gene expression profiles using the method of differential network are equally matched. But the result about the number of coding gene biomarkers is less than that of lnc RNA biomarkers. We think it is due to the less samples of the lnc RNA expression profiles.In summary, the method of differential network is superior to the method of different expression on the problem of extracting lnc RNA biomarkers or coding gene biomarkers to class disease subtypes. But the cross-over analysis about the method of differential network shows that, there is no advantages for the lnc RNA biomarkers in terms of the number because of the less samples of lnc RNA expression profiles. With the progress of science and technology, the cost of sequencing will be gradually decreased and the samples of lnc RNA expression profile will be more and more, which can compensate for the defect. Then lnc RNA biomarkers will have a wider application space and help people solve more problems.