G4detector

G4detector predicts G-quadruplex (G4) formation in guanine-rich DNA sequences to enable genome-wide identification of G4 structures associated with chromatin architecture, gene regulation, genomic instability, genetic diseases, and cancer progression.

Key Features:

• Multi-kernel CNN: Uses a multi-kernel convolutional neural network (CNN) to detect sequence patterns predictive of G4 formation.
• RNA secondary structure integration: Incorporates RNA secondary structure data alongside sequence information to improve prediction accuracy.
• G4-seq benchmarking: Trained and benchmarked on high-throughput G4-seq datasets derived from multiple species' genomes.
• Genome-wide detection: Performs genome-wide identification of potential G4s across entire genomes.
• Cross-species extrapolation: Exhibits extrapolative capability enabling human-trained models to be applied to non-human species' genomes.
• Comparative performance: Demonstrates superior performance compared to existing methods in benchmark comparisons for G4 prediction.

Scientific Applications:

• Genome-wide G4 prediction: Predicts potential G4 formation across whole genomes and novel DNA sequences as a computational alternative to experimental G4-seq.
• Functional interpretation: Provides predictions to support investigation of G4 roles in chromatin architecture, gene regulation, genomic instability, genetic diseases, and cancer progression.
• Comparative genomics: Enables cross-species analysis of G4 distribution and conservation using models trained on different species' G4-seq data.

Methodology:

Multi-kernel convolutional neural network (CNN) that incorporates RNA secondary structure data; trained and benchmarked on high-throughput G4-seq datasets from multiple species' genomes.