PanGIA: A universal framework for identifying association between ncRNAs and diseases

Area of Science:

• Biomedical Research
• Computational Biology
• Genomics

Background:

• Noncoding RNAs (ncRNAs) play crucial roles in biological functions and human diseases.
• Predicting ncRNA-disease associations is vital for biomedical research.
• Existing computational methods often focus on single ncRNA types, neglecting crucial interactions.

Purpose of the Study:

• To develop a computational framework, PanGIA, for simultaneously predicting associations between multiple ncRNA types and diseases.
• To address the limitation of type-specific prediction methods by incorporating cross-type ncRNA interactions.
• To enhance the accuracy and scope of ncRNA-disease association predictions.

Main Methods:

• Proposed PanGIA (Pan-ncRNA Graph-Interaction Attention network), a novel computational framework.
• Designed PanGIA to integrate multiple ncRNA types: microRNAs (miRNAs), long noncoding RNAs (lncRNAs), circular RNAs (circRNAs), and PIWI-interacting RNAs (piRNAs).
• Utilized graph-interaction attention mechanisms to capture competitive and cooperative interactions among ncRNAs.

Main Results:

• PanGIA demonstrated superior performance compared to type-specific state-of-the-art methods in both individual and comprehensive predictions.
• The framework showed robustness when nodes or ncRNA types were removed, validating the importance of cross-type information.
• Ablation studies confirmed the benefits of integrating information from multiple ncRNA types.

Conclusions:

• PanGIA offers significant advantages in predicting disease associations across diverse ncRNA types.
• Case studies validated the model's predictions, with high-confidence associations supported by literature evidence.
• PanGIA provides a new paradigm for exploring disease-associated ncRNAs with strong biological interpretability and practical application potential.