m6Aboost

m6Aboost predicts genuine N6-methyladenosine (m6A) sites in miCLIP2 data using a machine learning model trained on experimental and RNA sequence features, including characteristics of m6A deposition learned from Mettl3 knockout cells.

Key Features:

• miCLIP2 data processing: Functions assign read counts and extract features from miCLIP2 data stored in GRanges objects to run the m6Aboost model.
• DRACH-independent prediction: m6Aboost predicts m6A sites without filtering for DRACH motifs or requiring Mettl3 depletion.
• Cross-species applicability: The model identifies high-confidence m6A sites in both murine and human cell lines.

Scientific Applications:

• m6A site detection: m6Aboost enables deep and accurate detection of m6A RNA modifications at single-nucleotide resolution from miCLIP2 data.
• False positive reduction: The computational pipeline addresses inherent false positives in antibody-based m6A detection methods.
• Non-canonical m6A site identification: The model detects m6A sites outside of DRACH motifs based on learned deposition characteristics.

Methodology:

m6Aboost was trained as a machine learning model using experimental and RNA sequence features, with analyses calibrated using Mettl3 knockout cells to learn characteristics of m6A deposition.