Predicting binary, discrete and continued lncRNA-disease associations via a unified framework based on graph...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long non-coding RNAs (lncRNAs) are increasingly recognized for their roles in human diseases.
• Identifying lncRNA-disease associations (LDA) is crucial but challenging due to high costs of in vivo methods.
• Existing computational approaches often lack the ability to predict novel associations or quantify the intensity of LDA.

Purpose of the Study:

• To develop a novel computational framework for predicting lncRNA-disease associations.
• To address limitations of existing methods by enabling predictions for unassociated lncRNAs and diseases.
• To uncover nuanced biological relationships beyond binary associations, including the intensity of LDA.

Main Methods:

• Proposed a graph regression-based unified framework (GRUF).
• Enabled prediction of associations for lncRNAs and diseases with no prior known links.
• Developed a method to predict continuous LDA intensity, not just binary associations.

Main Results:

• GRUF demonstrated superior performance compared to state-of-the-art methods, achieving a 5%-16% improvement in AUC.
• The framework provides a confidence score for predicted LDAs, correlating with the number of involved RNA-Binding Proteins.
• Three of the top five novel LDA predictions by GRUF were indirectly validated by existing literature and biological facts.

Conclusions:

• GRUF effectively predicts novel lncRNA-disease associations, including for previously uncharacterized molecules.
• The framework offers a more comprehensive understanding of LDA by quantifying association strength and pathological implications.
• GRUF advances computational approaches for disease-related lncRNA discovery and mechanistic insights.