DeepHF

DeepHF predicts guide RNA (gRNA) activity for high-fidelity SpCas9 variants to optimize CRISPR-Cas9 genome editing.

Key Features:

• High-Fidelity Cas9 Variants: Targets wild-type SpCas9, eSpCas9(1.1), and SpCas9-HF1 for variant-specific gRNA activity prediction.
• Deep Learning Integration: Uses a deep learning model combining Recurrent Neural Networks (RNNs) with biological features to predict gRNA activity and reportedly outperforms other gRNA design models.
• Extensive Dataset Utilization: Trained and validated on indel rates from a genome-scale screen comprising indel rates for over 50,000 gRNAs across approximately 20,000 genes.
• Feature Evaluation: Evaluates the contribution of 1,031 features to gRNA activity predictions.

Scientific Applications:

• Genome editing optimization: Improves selection of gRNAs for gene knockout, insertion, and correction using predicted activity scores.
• Off-target risk reduction: Enables selection of high-activity gRNAs for high-fidelity SpCas9 variants to minimize off-target effects in precision and therapeutic applications.
• Functional genomics and screening: Supports large-scale functional genomics studies by providing activity predictions across thousands of gRNAs.

Methodology:

Genome-scale screens collected indel rates for >50,000 gRNAs across ~20,000 genes, and Recurrent Neural Network (RNN)-based deep learning models integrating 1,031 biological features were developed to predict gRNA activity.