GCNCDA

GCNCDA predicts potential associations between circular RNAs (circRNAs) and diseases to identify molecular links relevant to disease pathogenesis and biomarker and therapeutic target discovery.

Key Features:

• Graph Convolutional Network (FastGCN): Uses the Fast Graph Convolutional Networks (FastGCN) algorithm to extract high-level features from large-scale circRNA-disease networks.
• Unified Descriptor Formation: Constructs a unified descriptor by integrating disease semantic similarity and Gaussian Interaction Profile (GIP) kernel similarities for diseases and circRNAs based on known circRNA-disease associations.
• Feature Extraction and Classification: Applies FastGCN for feature extraction and the Forest by Penalizing Attributes (Forest PA) classifier to predict circRNA-disease associations.
• Validation and Performance: Reported results include 91.2% accuracy, 92.78% sensitivity, and 90.90% AUC on the circR2Disease benchmark dataset and demonstrated competitive performance against other classifiers and feature-extraction methods.

Scientific Applications:

• Disease Pathogenesis Exploration: Identification of novel circRNA-disease associations to aid elucidation of molecular mechanisms underlying complex diseases.
• Diagnostic and Therapeutic Improvements: Prioritization of candidate circRNAs that can inform development of diagnostic markers and therapeutic targets.
• Case Study Validation: Top predicted circRNAs were validated via literature and databases for diseases including breast cancer, glioma, and colorectal cancer.
• Comparative Benchmarking: Enables comparison with various classifier models, feature-extraction methods, and other state-of-the-art approaches.

Methodology:

Data integration combining disease semantic similarity with Gaussian Interaction Profile (GIP) kernel similarities for diseases and circRNAs; feature extraction using Fast Graph Convolutional Networks (FastGCN); prediction using the Forest by Penalizing Attributes (Forest PA) classifier; validation via 5-fold cross-validation.