lncRNA localization and feature interpretability analysis

Abstract

Subcellular localization is crucial for understanding the functions and regulatory mechanisms of biomolecules. Long non-coding RNAs (lncRNAs) have diverse roles in cellular processes, and their localization within specific subcellular compartments provides insights into their biological functions and implications in health and disease. The nucleolus and nucleoplasm are key hubs for RNA metabolism and cellular regulation. We developed a model, LncDNN, for identifying the localization of lncRNAs in the nucleolus and nucleoplasm. LncDNN uses three different encoding schemes and employs Shapley Additive Explanations for feature analysis and selection. The results show that LncDNN is more accurate than other models. Additionally, an interpretable analysis of the features influencing the model was conducted. LncDNN is applicable for identifying the localization of lncRNA in the nucleolus and nucleoplasm, aiding in the understanding and in-depth study of related biological processes and functions.

Keywords

• MT: Bioinformatics
• lncRNAs
• subcellular localization
• nucleolus
• nucleoplasm
• machine learning