G4Beacon: An In Vivo G4 Prediction Method Using Chromatin and Sequence Information

Area of Science:

• Genomics
• Epigenetics
• Bioinformatics

Background:

• G-quadruplex (G4) structures are epigenetic regulatory elements forming in guanine-rich DNA regions.
• Predicting in vivo G4 formation within living cells presents significant challenges.
• Understanding G4 structures is crucial for epigenetic regulation studies.

Purpose of the Study:

• To develop a robust machine learning method for predicting G-quadruplex formation in vivo.
• To enhance the accuracy and reliability of G4 structure prediction across diverse cell types.
• To provide an accessible tool for identifying G4s in various cellular contexts.

Main Methods:

• Utilized the Gradient-Boosting Decision Tree (GBDT) algorithm for predictive modeling.
• Integrated ATAC-seq data with sequences of in vitro G4s.
• Trained and validated the G4Beacon model on diverse cell line data.

Main Results:

• G4Beacon demonstrated high accuracy in predicting in vitro G4 formation ability in different cell types.
• The model performed excellently even with highly skewed test datasets.
• G4Beacon accurately identified in vivo G4s in new cell lines, irrespective of experimental platform.

Conclusions:

• G4Beacon is an accurate, reliable, and user-friendly method for predicting in vivo G-quadruplex structures.
• The tool facilitates precise G4 identification across various cell lines and experimental conditions.
• G4Beacon advances the study of G4s as epigenetic regulatory elements.