Machine learning-based prediction models to guide the selection of Cas9 variants for efficient gene editing

Area of Science:

• Molecular Biology
• Genomics
• Biotechnology

Background:

• CRISPR-Cas9 technology is expanding with new variants offering improved efficiency, specificity, and alternative PAM recognition.
• Selecting optimal Cas9 variants and guide RNAs (gRNAs) is essential for high-fidelity genome editing.
• Systematic quantification and predictive modeling of gRNA performance are needed.

Purpose of the Study:

• To systematically compare the activity and specificity of gRNAs across four SpCas9 variants.
• To develop machine learning models for predicting gRNA efficiency and specificity.
• To provide accessible tools for researchers selecting Cas9 variants and gRNAs.

Main Methods:

• Utilized synthetic gRNA-target paired libraries.
• Employed next-generation sequencing to assess editing outcomes.
• Developed and validated machine learning models for prediction.

Main Results:

• Nucleotide composition in the PAM-distal region significantly impacts HiFi Cas9 and LZ3 Cas9 editing efficiency.
• Machine learning models were successfully developed to predict gRNA efficiency and specificity for the studied variants.
• Performance variations were observed among the four SpCas9 variants.

Conclusions:

• The study provides a framework for evaluating and predicting gRNA performance with different Cas9 variants.
• Machine learning models can aid in the rational design and selection of gRNAs for efficient genome editing.
• Accessible prediction tools are crucial for advancing CRISPR applications in diverse research fields.