IIRWR: Internal Inclined Random Walk With Restart for LncRNA-Disease Association Prediction

Abstract

Experimental studies have demonstrated that long-non-coding RNAs (lncRNAs) are closely related to human disease. However, due to the complexity of diseases and high costs of bio-experiments, associations between diseases and lncRNAs are still unclear. Hence, it is essential to establish effective computational models to predict the potential relationships between diseases and lncRNAs. In this paper, different from traditional prediction models based on random walk with restart (RWR), a novel prediction model based on internal inclined random walk with restart (IIRWR) has been established to infer potential lncRNA-disease associations and compared to the state-of-the-art RWR-based prediction models. One major novelty of our IIRWR-based prediction model is the introduction of the concept of disease clique, which makes the process of the random walk to possess an internal tendency. The other major novelty of our model lies in the addition of the weights of disease linkages to the traveling network, which guarantees our model can achieve excellent prediction performance while the number of known lncRNA-disease associations is limited. The simulation results show that our model can achieve reliable AUCs of 0.8080, 0.8363, and 0.8745 under the frameworks of five-fold cross-validation (CV), ten-fold CV, and leave-one-out cross validation (LOOCV), respectively. Moreover, in case studies of cervical cancer and leukemia, the experimental results show that eight and ten out of the top ten predicted lncRNAs can be confirmed by related literature, which demonstrates that our method is effective in predicting novel diseases associated lncRNAs.