DeepCas13

DeepCas13 predicts CRISPR-Cas13d guide RNA on-target and off-target effects to improve guide selection for targeting protein-coding and non-coding RNAs.

Key Features:

• Deep Learning Model: Employs a deep learning model that predicts guide RNA on-target activity using sequence and secondary-structure features.
• On-Target Prediction: Predicts guide efficiency by analyzing sequence data and structural information for both coding and non-coding targets.
• Off-Target Viability Effects: Identifies off-target viability impacts that correlate with on-target efficiencies, particularly for guides targeting non-essential genes.
• Normalization and Control: Emphasizes selection of appropriate negative control guides to reduce false positives in proliferation screens.
• Application to lncRNAs: Applied to guide RNAs targeting long non-coding RNAs (lncRNAs) to identify guides that influence cell viability and proliferation across cell lines.
• Validation: Prediction performance has been validated with secondary CRISPR-Cas13d screens and quantitative RT-PCR experiments.

Scientific Applications:

• Guide selection for CRISPR-Cas13d experiments: Inform selection of guides with higher predicted on-target activity and lower off-target viability effects.
• Functional screening of non-coding RNAs: Support identification of lncRNA-targeting guides that affect cell viability and proliferation.
• Reduction of off-target cellular consequences: Help minimize unintended viability effects and false positives in proliferation screens.

Methodology:

Uses a deep learning model that analyzes guide RNA sequence and predicted secondary-structure features to predict on-target activity and to identify correlations between predicted activities and off-target viability effects, with normalization considerations based on negative control guides.