LncRNA-Disease Association Prediction Model Applying Distance-based Data Labeling

Jaein Kim, Seung-Won Yoon, In-Woo Hwang, Kyu-Chul Lee

http://doi.org/10.5626/JOK.2023.50.5.420

lncRNAs are noncoding RNAs of 200 or more nucleotides. For a long time, non-coding RNA has been considered unimportant because it cannot directly produce proteins, but recent studies have reported that non-coding RNA plays a role in regulating protein expression. Abnormal expression of lncRNAs causes various diseases and predicting the associations between lncRNAs and diseases would help diagnose diseases in the early stages or prevent diseases. However, research that predicts the correlation of biological data is time-consuming and costly if it is conducted as a direct experiment. Therefore, it is important to overcome these challenges using computational methods. Therefore, in this study, we propose a lncRNA-disease association prediction model based on Long Short-Term Memory (LSTM). In addition, since negative samples were randomly generated in previous studies, there is uncertainty in the data. So this study also proposes a distance-based data labeling method that solves this uncertainty. Our model achieved the highest AUC (0.97) through the data labeling method and classification model presented in this study.