LncTracker: A Unified Multi-Channel Framework for Multi-Label lncRNA Localization.

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• Long non-coding RNAs (lncRNAs) are crucial regulators of biological processes.
• lncRNA subcellular localization dictates their function and is vital for understanding cellular regulation and disease.
• Existing prediction methods often overlook structural information and struggle with multi-label classification.

Purpose of the Study:

• To develop an efficient deep learning framework, LncTracker, for multi-label prediction of lncRNA subcellular localization.
• To integrate both sequence and secondary structure information for improved prediction accuracy.
• To provide a user-friendly web server for LncTracker accessibility.

Main Methods:

• Developed LncTracker, a multi-channel deep learning framework.
• Integrated primary sequence and secondary structure information, converting structures into attributed graphs.
• Utilized joint sequence-structure representations for predicting localization across seven compartments.

Main Results:

• LncTracker demonstrated superior performance compared to state-of-the-art methods, especially for imbalanced datasets.
• The framework effectively identified sequence motifs and key sub-structures contributing to localization.
• Achieved robust and accurate multi-label prediction of lncRNA subcellular localization.

Conclusions:

• LncTracker offers a powerful and accurate approach for predicting lncRNA subcellular localization by integrating sequence and structural data.
• The tool enhances the study of lncRNA functions and their roles in disease mechanisms.
• The publicly available web server promotes broader research and application of lncRNA localization prediction.