Machine learning-based prediction of the activity and specificity of Cas9 variants in gene editing

Machine learning-based prediction of the activity and specificity of Cas9 variants in gene editing Overall design: This study aims to compare the activity and specificity of SpCas9 variants by deep learning at genome-scale. In this study, we have focused on the development of a system for evaluating the activity and specificity of SpCas9 variants by using >180,000 guide RNAs (gRNAs) covering ~20,000 protein-coding genes and ~10,000 non-coding genes in synthetic constructs with a high-throughput manner. With the help of deep learning algorithms in the field of artificial intelligence, eight prediction models with the best generalization performance now are constructed: AIdit_ON_HiFi, AIdit_ON_NG, AIdit_ON_Sniper, AIdit_ON_LZ3, AIdit_OFF_HiFi, AIdit_OFF_NG, AIdit_OFF_Sniper and AIdit_OFF_LZ3. Moreover, we found a novel Cas9 variant SuperFi Cas9 has a very low activity compared with other variants in vivo. Finally, we confirm that this study will greatly facilitate CRISPR-based genome editing.