circR2Disease

circR2Disease predicts associations between circular RNAs (circRNAs) and diseases by leveraging a curated database of experimentally supported circRNA–disease associations and an improved collaborative filtering recommendation algorithm (ICFCDA).

Key Features:

• Curated database: A circR2Disease database of experimentally supported circRNA–disease associations with an emphasis on disease-related and cancer-associated entries.
• ICFCDA algorithm: An Improved Collaboration Filtering for circRNA–Disease Associations (ICFCDA) recommendation system used to predict potential circRNA–disease links.
• Data integration: Extraction of multiple data types from various databases to support model construction and prediction.
• Similarity networks: Construction of circRNA similarity networks and disease similarity networks as the basis for prediction.
• Cold-start handling: Use of computational approaches within ICFCDA to address the cold-start problem in traditional experimental methods.
• Validation metrics: Performance evaluation using leave-one-out cross-validation achieving an area under the curve (AUC) of 0.946.
• Case study corroboration: Case studies on common diseases validated by additional databases.
• Biological context: Retains circRNA-specific attributes such as closed-loop structures without 5'-3' polarity or polyadenylated tails, and enhanced stability and conservation relative to linear RNAs.

Scientific Applications:

• Investigation of disease-related circRNAs: Facilitation of studies into circRNAs associated with various diseases.
• Cancer research: Support for examining roles of circRNAs in cancer contexts.
• Biomarker and therapeutic target identification: Identification of potential circRNA biomarkers and therapeutic targets for disease.

Methodology:

ICFCDA extracts multiple data types from various databases to construct circRNA similarity networks and disease similarity networks and then applies an improved collaborative filtering recommendation algorithm to predict new circRNA–disease associations, with performance assessed by leave-one-out cross-validation (AUC = 0.946) and validated by case studies against additional databases.