Prediction of lncRNA functions using deep neural networks based on multiple networks

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long non-coding RNAs (lncRNAs) play crucial roles in organismal physiology, yet their functions remain largely uncharacterized.
• The exponential growth of lncRNA data in biological databases necessitates advanced computational methods for functional annotation.
• Understanding lncRNA function is critical for advancing biological research and therapeutic development.

Purpose of the Study:

• To develop a novel computational method for predicting the functions of long non-coding RNAs (lncRNAs).
• To effectively utilize the increasing volume of lncRNA-related data in biological databases.
• To improve the accuracy and efficiency of lncRNA functional annotation.

Main Methods:

• Proposed DNGRGO, a computational method based on global heterogeneous networks for lncRNA function prediction.
• Calculated similarities among proteins, miRNAs, and lncRNAs.
• Annotated lncRNA functions by identifying similar protein-coding genes with known Gene Ontology (GO) annotations.

Main Results:

• DNGRGO demonstrated superior predictive performance compared to existing methods, achieving maximum F-measure and coverage.
• Manual annotation of GO terms to lncRNAs validated the method's effectiveness.
• The integration of miRNA data significantly enhanced DNGRGO's predictive accuracy.

Conclusions:

• DNGRGO effectively annotates lncRNAs by capturing low-dimensional features within heterogeneous networks.
• The method provides a valuable tool for exploring the functions of uncharacterized lncRNAs.
• Integrating miRNA data is a promising strategy for improving lncRNA functional prediction models.