Deep learning platform for predicting CRISPR-Cas13 on-target and off-target activity

This technology is a deep-learning guide-design platform that identifies CRISPR-Cas13 guide RNAs capable of achieving desired levels of gene knockdown for more precise RNA targeting and gene regulation.

Unmet Need: Reliable design of Cas13 guides with predictable activity and specificity

Current CRISPR-Cas13 guide design approaches are primarily optimized to identify highly active, perfectly matched guide RNAs. However, they have limited ability to predict the effects of guide-target mismatches, assess potential off-target activity, or select guides that produce intermediate levels of RNA knockdown. These limitations make it difficult to precisely control transcript abundance when maximal gene suppression is not desired.

The Technology: Predictive Cas13 guide design for tunable gene silencing and off-target assessment

This technology is a deep-learning framework to predict CRISPR-Cas13 guide RNA activity based on guide-target sequence and RNA-context features. By modeling the effects of sequence mismatches and other target-specific characteristics, the platform can support selection of guides with high activity, assessment of potential off-target activity, or predetermined intermediate levels of RNA knockdown. This enables guide selection based on the desired degree of gene suppression rather than maximizing activity alone.

The technology has been validated across multiple Cas13d screens, target genes, and cell lines, including experimental demonstration of tunable gene knockdown using mismatch-containing guides.

Applications:

• CRISPR-Cas13 guide RNA design for RNA-targeting therapeutics
• Functional genomics and target validation
• Gene expression and dosage studies
• Synthetic biology and cell engineering
• Development of gene-specific knockdown research tools

Advantages:

• Enables tunable gene expression beyond maximal knockdown
• Supports prediction of mismatch-dependent and potential off-target activity
• Enables guide selection for predetermined levels of RNA suppression
• Accounts for RNA-context features beyond sequence complementarity alone
• Supports both highly active and partially active guide selection

Lead Inventor:

David A. Knowles, Ph.D.

Patent Information:

Patent Pending

Related Publications:

• Schertzer MD, Stirn A, Isaev K, et al. “Cas13d-mediated isoform-specific RNA knockdown with a unified computational and experimental toolbox.” Nature Communications. 2025 Jul 29; 16: 6948.

• Wessels HH, Stirn A, Méndez-Mancilla A, Kim EJ, Hart SK, Knowles DA, Sanjana NE. “Prediction of on-target and off-target activity of CRISPR-Cas13d guide RNAs using deep learning.” Nature Biotechnology. 2024 Apr; 42(4): 628-637.

Tech Ventures Reference:

• IR CU27053

• Licensing Contact: Joan Martinez