Predicting human disease-associated circRNAs based on locality-constrained linear coding.

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Circular RNAs (circRNAs) are endogenous non-coding RNAs increasingly linked to human disease pathogenesis.
• Experimental identification of circRNA-disease associations is costly and time-consuming.
• Novel computational methods are crucial for efficient prediction of these associations.

Purpose of the Study:

• To develop and evaluate a computational method for predicting circRNA-disease associations.
• To integrate multiple data sources including known associations and semantic similarity networks.
• To provide a reliable tool for understanding disease mechanisms and facilitating diagnosis and treatment.

Main Methods:

• Developed the LLCDC computational method integrating known circRNA-disease associations and semantic similarity networks.
• Utilized Locality-Constrained Linear Coding (LLC) to reconstruct circRNA and disease similarity networks.
• Applied label propagation to predict circRNA-disease associations and evaluated using 5-fold cross-validation.

Main Results:

• The LLCDC method achieved a high Area Under the Curve (AUC) of 0.9177, outperforming three other methods.
• Case studies on gastric cancer, breast cancer, and papillary thyroid carcinoma validated the method's reliability.
• The study successfully predicted novel circRNA-disease associations.

Conclusions:

• LLCDC offers an effective computational approach for predicting circRNA-disease associations.
• This method can significantly contribute to understanding disease pathogenesis and developing targeted therapies.
• The findings highlight the potential of computational methods in circRNA research for disease management.