CASowary

CASowary predicts the efficacy of single guide RNAs (sgRNAs) for CRISPR Cas13 and other transcript-targeting systems to enable selection of effective RNA-targeting guides based on sequence features and target availability.

Key Features:

• Implementation: Implemented in Python 3.6.8 as a machine learning predictor for sgRNA efficacy.
• Model architecture: Uses a Decision Tree architecture to classify sgRNA efficacy into four categories based on expected target transcript knockdown.
• Feature set: Utilizes 112 distinct features encompassing sequence features and target availability.
• Training data: Trained on publicly available RNA knockdown data from Cas13 experiments targeting 555 sgRNAs in HEK293 cells.
• Target accessibility: Incorporates transcriptome-wide protein occupancy information to account for target availability.
• Cell-line integration: Integrates POP-seq protein occupancy maps generated in HeLa cells and publicly available protein–RNA interaction data from HEK293 cells for cell-line-specific predictions.
• Biological considerations: Accounts for transcriptome complexity, including sequence redundancy, three-dimensional RNA structure, and RNA Binding Protein (RBP) interactions.
• Performance: Reports a noise-normalized accuracy exceeding 70% on the training dataset.
• Experimental validation: Predictions were experimentally confirmed using an independent RNA-targeting system, CIRTS.
• Scale: Capable of generating sgRNA predictions across entire transcriptomes.

Scientific Applications:

• sgRNA design for Cas13: Selection and prioritization of sgRNAs for CRISPR Cas13-mediated RNA targeting and depletion.
• Cell-line-specific guide selection: Design of sgRNAs tailored to cell lines using POP-seq occupancy maps (HeLa) and HEK293 protein–RNA interaction data.
• Accounting for RBPs and structure: Prioritization of guides that consider RBP interactions and target accessibility influenced by RNA structure.
• Transcriptome-wide screening: Rapid deployment of sgRNA prediction across whole transcriptomes for large-scale studies.
• Therapeutic target development: Support for identifying RNA-targeting guides relevant to therapeutic interventions targeting RNA regulatory processes.
• Experimental planning: Informing experimental validation strategies using independent RNA-targeting systems such as CIRTS.

Methodology:

Trained on publicly available Cas13 RNA knockdown data (555 sgRNAs in HEK293) combined with transcriptome-wide protein occupancy information, using a Decision Tree classifier built on 112 features to assign sgRNAs to four efficacy categories and integrating POP-seq HeLa maps and HEK293 protein–RNA interaction data for cell-line-specific predictions.