Advancing CRISPR/Cas gene editing with machine learning

Abstract

Clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (Cas) system is a powerful tool for gene editing. Recent advancement and adaptation of machine learning (ML) approaches in gene editing field have benefited both the users and developers of the CRISPR/Cas toolset. Editing outcomes of given single guide RNAs (sgRNA) can be predicted by ML models, lowering the experimental burden in optimising sgRNA designs for specific gene editing tasks. ML models can also predict protein structures and provide a directed evolution framework, facilitating the engineering process of better gene editing tools. Nonetheless, the current gene editing-related ML models can sometimes suffer from confirmational biases due to the selection of training datasets, limiting the scope of usage. Efforts should be made in building better models and expanding the use of ML in other aspects of CRISPR/Cas gene editing.