G4Boost

G4Boost predicts the formation and thermodynamic stability of guanine-rich nucleic acid sequences that can form G-quadruplexes (G4s) in DNA and RNA.

Key Features:

• Decision Tree-Based Prediction: G4Boost employs a decision tree machine learning algorithm to assess potential G-quadruplex formation from nucleotide sequence data.
• Sequence Intrinsic Feature Utilization: It leverages intrinsic nucleotide sequence features, including composition and estimated structural topologies, to predict folding probability and thermodynamic stability of G4 motifs.
• High Predictive Accuracy: The method predicts quadruplex folding state with >93% accuracy and an F1-score of 0.96, and predicts folding energy with RMSE of 4.28 and R² of 0.95.
• Cross-Species Application: G4Boost has been validated on experimentally determined G4 structures across species, including plants and humans.
• Comparative Performance: The tool outperforms existing machine-learning-based predictors DeepG4, Quadron, and G4RNA Screener in both accuracy and F1-score.

Scientific Applications:

• Gene Regulation Studies: Provides predictions of G4 formation and stability to inform investigations of gene regulation mechanisms.
• Nucleic Acid Structural Analysis: Supports analysis of G-quadruplex motifs in DNA and RNA for studies of nucleic acid secondary structure.
• Cross-Species Comparative Studies: Enables comparison of experimentally determined G4 stability across species such as plants and humans.

Methodology:

G4Boost analyzes nucleotide sequence intrinsic features (including composition and estimated structural topologies) with a decision tree algorithm to identify potential G4-forming regions and predict folding probability and thermodynamic stability.