GCLNSTDA: Predicting tsRNA-Disease Association Based on Contrastive Learning and Negative Sampling

Published: 01 Jan 2024, Last Modified: 06 Feb 2025BCB 2024EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: An increasing number of studies have shown that tsRNAs are closely associated with diseases. Using computational methods to predict potential tsRNA-disease associations can effectively reduce the amount of human and material resources consumed. We propose a computational framework (GCLNSTDA) to predict tsRNA-disease associations based on contrastive learning and negative sampling methods. Firstly, we reconstruct the tsRNA-disease association by the truncated singular values. Then, the features of tsRNA and disease are learned by using contrastive learning and graph neural network based on the reconstructed and the original tsRNA-disease association. Finally, the multilayer perceptron is used to calculate the association prediction scores. In addition, we select high-quality negative samples by Bayesian negative sampling method to further improve the model performance. We conduct five-fold cross-validation and denovo experiments on a manually collected tsRNA-disease association dataset, the experimental results show that GCLNSTDA outperforms the other six compared methods. We also perform a case study on lung cancer and the experimental results show that GCLNSTDA is an effective tool for predicting potential associations between tsRNA and disease.
Loading