BoostMEC

BoostMEC predicts CRISPR-Cas9 cleavage efficiency from single guide RNA (sgRNA) sequence and thermodynamic features to inform sgRNA design and selection.

Key Features:

• Modeling algorithm: Uses gradient boosting techniques implemented with LightGBM to model sgRNA cleavage efficiency.
• Input features: Integrates position-specific sequence composition, global sequence properties, and thermodynamic characteristics derived from sgRNAs.
• Feature engineering: Employs both direct and derived sequence features as predictors of cleavage efficiency.
• Machine learning approach: Processes features through conventional machine learning techniques rather than deep learning architectures.
• Interpretability: Provides clearer insights into feature contributions compared with deep learning–based prediction approaches.
• Benchmarking: Evaluated against 10 popular models across 13 external datasets to assess comparative performance.

Scientific Applications:

• sgRNA design and selection: Guides selection and optimization of sgRNAs for CRISPR-Cas9 experiments by predicting cleavage efficiency.
• CRISPR-Cas9 outcome prediction: Predicts editing efficiency to inform experimental planning for genome editing.
• Comparative model evaluation: Serves as an interpretable alternative for comparing feature contributions among CRISPR prediction methods.

Methodology:

Integrates direct and derived sgRNA sequence features (position-specific sequence composition, global sequence properties, thermodynamic characteristics) and models cleavage efficiency using gradient boosting (LightGBM) and conventional machine learning techniques; performance was benchmarked against 10 popular models across 13 external datasets.