WLDAP: A computational model of weighted lncRNA-disease associations prediction

Abstract

Increasing evidence has demonstrated that long non-coding RNAs (lncRNAs) play essential roles in various human complex diseases. Compared with protein-coding genes, the associations between diseases and lncRNAs are still not well studied. Hence, inferring disease-associated lncRNAs on a genome-wide scale has become an urgent matter. However, known associations are still being studied in small quantities because experimental verification requires a large amount of human and material resources. To solve this problem, we proposed a method called the weight matrix of lncRNA-disease associations prediction (WLDAP) by combining various biological information. Firstly, this method incorporated information about lncRNA-disease associations. Secondly, the Gaussian interaction profile kernel was used to calculate the similarity between diseases and lncRNAs. Finally, the weighting model was used to obtain the similarity score between diseases and potential lncRNAs. After leave-one-out cross-validation (LOOCV) was applied, the AUC value of WLDAP reached 92.07%, the AUC value obtained by 5-fold cross-validation was 93.7%, indicating a relatively good prediction performance is relatively good. In addition, the top five candidates successfully predicted the rankings of colorectal cancer, lung cancer, and breast cancer. Most of these predictions are confirmed by different relevant databases and various literature, providing the value of WLDAP in demonstrating potential lncRNA-disease associations. This method will help people further explore complex human diseases at the molecular level, providing a high-quality basis for the diagnosis, prognosis, prevention, and treatment of diseases.