Machine learning-coupled combinatorial mutagenesis enables resource-efficient engineering of CRISPR-Cas9 genome...

Area of Science:

• Molecular Biology
• Biotechnology
• Bioinformatics

Background:

• The Cas9 protein is crucial for genome editing, utilizing numerous amino acid residues for DNA binding.
• Optimizing Cas9 activity requires screening vast numbers of combinatorial variants, posing a significant challenge for experimental methods.

Purpose of the Study:

• To develop and validate a machine learning-coupled approach for efficient Cas9 engineering.
• To reduce the experimental burden in identifying high-performing Cas9 variants.

Main Methods:

• Generation and cross-validation of ten in silico and experimental datasets for Cas9 combinatorial mutagenesis.
• Application of machine learning algorithms to analyze mutagenesis libraries.
• Structure-guided engineering for variant identification.

Main Results:

• The machine learning-coupled approach reduced experimental screening by up to 95%.
• Top-performing Cas9 variants were enriched approximately 7.5-fold compared to a null model.
• The N888R/A889Q variant demonstrated increased editing activity in KKH-SaCas9 and its derived base editor.

Conclusions:

• A resource-efficient workflow for high-throughput genome editor engineering was validated.
• Machine learning significantly enhances the efficiency of Cas9 engineering.
• The identified variants and workflow are applicable to improving genome editing tools in human cells.