gShapeLnoc: A Graph Network Incorporating Shapelet Embedding Model for LncRNA Subcellular Localization

Abstract

The subcellular localization of Long non-coding RNAs (LncRNAs) is a pivotal research area with profound implications for understanding underlying molecular mechanisms, involvement in pathological processes, and regulation of gene expression. Traditional machine learning based methods often rely on k-mer frequencies for classification, ignoring the global features of LncRNAs. More recent methods based on deep learning have utilized sequence and graph models for LncRNA classification. However, while these methods could improve their combination of LncRNA features, they still possess limitations, for example, ignoring the fact that mutations could occur in LncRNAs. Simultaneously, it employs the Shapelet model to extract local features of the most representative k-mer among different LncRNA classes. Furthermore, gShapeLnoc combines global and local feature representations for predicting the subcellular localization of LncRNAs. We have evaluated the performance of the gShapeLnoc algorithm on a real dataset, and the results demonstrate that it outperforms existing state-of-the-art methods in terms of accuracy.