RWLPAP: Random Walk for lncRNA-protein Associations Prediction

Background: During recent years, a lot of experimental studies have shown that lncRNA-binding proteins play a key role in many biomedical processes. Therefore, it is important to predict the potential lncRNA-protein associations in biomedical researches. Objective: To predict the associations between lncRNAs and proteins more reliably and more efficiently. Method: Considering the limitations of previous computational methods, we introduce a predictive model called Random Walk for lncRNA-Protein Associations Prediction (RWLPAP). It belongs to semi-supervised learning algorithms, and thus RWLPAP successfully avoids the difficulty of extracting negative data sets and features. Results: By the leave-one-out cross validation, we compare RWLPAP with previous methods and conclude that RWLPAP has an AUC of 0.88, which is significantly higher than other three models. It suggests that RWLPAP is more reliable and effective in predicting the interactions between lncRNAs and proteins. Conclusion: In the case study, according to the rank of predictive scores, we can find that the scores of some lncRNA-protein associations are highly ranking by our method when is compared with other three methods. It indicates that our method is very effective and comprehensive. Therefore, we can expect that RWLPAP will be a useful bioinformatic tool in the future.