CircCNN

CircCNN predicts pre-mRNA back-splicing sites to identify sequence motifs underlying circular RNA (circRNA) biogenesis and inform studies of gene expression regulation.

Key Features:

• Convolutional Neural Network (CNN) Architecture: Employs a convolutional neural network with batch normalization to predict pre-mRNA back-splicing sites.
• Position Probability Matrix (PPM) Features: Extracts Position Probability Matrix (PPM) features and converts them into motifs for downstream analysis.
• Motif and Short-Sequence Analysis: Analyzes motif distribution and special short sequences that are crucial for pre-mRNA back-splicing.
• Experimental Validation: Demonstrates superior performance over baseline models through validation on three distinct datasets.

Scientific Applications:

• Gene Expression Regulation: Facilitates elucidation of circRNA regulatory roles in gene expression by identifying back-splicing sites and associated motifs.
• Disease Research: Identifies candidate targets related to complex malignant diseases for further therapeutic and mechanistic studies.

Methodology:

Uses a convolutional neural network with batch normalization; extracts Position Probability Matrix (PPM) features and converts them into motifs; performs motif distribution and short-sequence analysis; validated on three datasets showing improved performance over baseline models.