Prediction of strand-specific and cell-type-specific G-quadruplexes based on high-resolution CUT&Tag data

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• G-quadruplex (G4) structures are non-classical DNA motifs found throughout the genome.
• G4s play roles in various biological processes and are enriched in functional genomic regions in a cell-type-specific manner.
• Accurate prediction of G4s is crucial, but experimental methods are laborious; computational approaches are needed.

Purpose of the Study:

• To develop a novel computational method for predicting G-quadruplex (G4) structures.
• To leverage high-resolution G4 CUT&Tag sequencing data for improved model training.
• To analyze sequence features and transcription factor binding motifs associated with G4 formation.

Main Methods:

• Constructed a new dataset utilizing G4 CUT&Tag sequencing data.
• Developed and applied an XGBoost machine learning model for G4 prediction.
• Performed sequence analysis to identify factors influencing G4 formation.

Main Results:

• The XGBoost model demonstrated strong performance in predicting G4s within and across different cell types.
• Sequence analysis revealed that flanking sequences, particularly GC content, significantly impact G4 structure formation.
• Identified G4 motifs, including those for known transcription factors like SP2 and BPC, suggesting their regulatory role.

Conclusions:

• The developed computational method provides an accurate approach for G4 prediction using G4 CUT&Tag data.
• Flanking sequences and specific transcription factors are key determinants of G4 structure formation in the genome.
• This work enhances our understanding of G4s' role in genomic regulation.