Applications and Future Perspectives of Artificial Intelligence in gRNA Design

Abstract

The rapid development of CRISPR gene editing has brought revolutionary breakthroughs to the biomedical area, yet the design of gRNA is still facing problems as unstable editing efficiency and the risk of off-target. The introduction of Artificial Intelligence (AI) technology has guided this issue to a brighter future. Studies have shown that deep learning models can predict editing efficiency and assess off-target risk precisely through analyzing information as the sequence features of gRNA, while the introduction of tools like Reinforcement Learning (RL) further enlarges the space of gRNA design. There is still a deficiency in recent studies. This paper analyzes the application of AI in gRNA design, including editing efficiency and off-target risk assessment. The workflow, performance, and limitations of each model are concluded, and the conclusion of AI having a great devotion to gRNA design and the tips for choosing different models in different cases are given. This research provides references about choosing and designing models, but there are still unsolved challenges in aspects of building standard datasets and public databases. In the future, gRNA design can be optimized by reinforcing the use of Federated Learning (FL) and single-cell CRISPR screening technology.