CGD

CGD provides on-target scoring and design evaluation for guide RNAs across CRISPRi, CRISPRa, Cas9, Cas9NG, and Cas12a systems using Elastic Net Logistic Regression (ENLOR) to improve gRNA efficacy prediction.

Key Features:

• Supported CRISPR systems: Supports CRISPR interference (CRISPRi), CRISPR activation (CRISPRa), Cas9, non-canonical Cas9 (Cas9NG), and Cas12a.
• On-target scoring algorithm: Implements Elastic Net Logistic Regression (ENLOR) as the core on-target scoring method.
• Specialized models: Provides system-specific models CGDi, CGDa, CGD9, CGD9NG, and CGD12a trained on public datasets for each CRISPR system.
• Training data: Models are trained using public datasets specific to each CRISPR modality.
• Benchmarking and comparative performance: Benchmarked against ElasticNet Linear Regression (ENLR), Random Forest and Boruta (RFB), and Xgboost with built-in feature selection, showing improved prediction on independent test datasets.
• Implementation: Implemented in Python.
• Generalization focus: Designed to reduce bias and improve generalization in gRNA efficacy prediction across datasets.

Scientific Applications:

• Genome engineering in mammalian cells: Supports guide RNA design and efficacy prediction for genome engineering experiments in mammalian somatic cells.
• CRISPRi/CRISPRa and nuclease editing workflows: Applicable to CRISPRi, CRISPRa, and nuclease-based editing workflows using Cas9, Cas9NG, and Cas12a.
• Target-site assessment: Enables assessment of potential target sites across different CRISPR systems to inform experimental design.

Methodology:

CGD uses Elastic Net Logistic Regression (ENLOR) and system-specific models (CGDi, CGDa, CGD9, CGD9NG, CGD12a) trained on public datasets, with benchmarking against ENLR, Random Forest and Boruta (RFB), and Xgboost evaluated on independent test datasets.