According to a recent publication in Nature Biotechnology, researchers from New York University, Columbia Engineering, and the New York Genome Center have discovered that artificial intelligence (AI) can accurately predict the on- and off-target activity of CRISPR tools that target RNA instead of DNA. This groundbreaking study combines a deep learning model with CRISPR screens, enabling researchers to control the expression of human genes in various ways. This precise gene control could lead to the development of novel CRISPR-based therapies.

CRISPR technology has garnered significant attention due to its versatility in biomedical applications, ranging from treating genetic diseases like sickle cell anemia to enhancing the characteristics of crops. Traditionally, CRISPR targets DNA using the Cas9 enzyme. However, scientists have recently uncovered an alternative form of CRISPR that targets RNA using the Cas13 enzyme.

RNA-targeting CRISPRs have broad applications, including RNA editing, gene silencing, and high-throughput screening for identifying potential drug candidates. Researchers at New York University and the New York Genome Center have devised a platform for RNA-targeting CRISPR screens using Cas13. This platform aims to enhance the understanding of RNA regulation and uncover the functions of non-coding RNAs. Moreover, since RNA serves as the primary genetic material in viruses like SARS-CoV-2 and influenza, RNA-targeting CRISPRs hold promise for developing new methods to prevent or treat viral infections. Additionally, in human cells, gene expression begins with the creation of RNA from DNA in the genome.

A crucial objective of this study is to maximize the activity of RNA-targeting CRISPRs on the intended RNA target while minimizing off-target activity that could potentially harm the cell. Off-target activity includes mismatches between the guide RNA and the target RNA, as well as insertion and deletion mutations. Previous studies on RNA-targeting CRISPRs primarily focused on on-target activity and mismatches, neglecting the prediction of off-target activity, especially insertion and deletion mutations. However, such mutations account for approximately one in five mutations in human populations, making them significant considerations for CRISPR design.

The researchers conducted a series of pooled RNA-targeting CRISPR screens in human cells as part of the study. They assessed the activity of 200,000 guide RNAs targeting essential genes, evaluating both “perfect match” guide RNAs and those with off-target mismatches, insertions, and deletions.

To facilitate the analysis, the research team collaborated with machine learning expert David Knowles to develop a deep learning model called TIGER (Targeted Inhibition of Gene Expression via guide RNA design). TIGER was trained using the data obtained from the CRISPR screens. By comparing the predictions generated by the deep learning model with laboratory tests conducted in human cells, TIGER proved capable of accurately predicting both on-target and off-target activity. It outperformed previous models designed for on-target guide RNA design using Cas13 and became the first tool capable of predicting off-target activity for RNA-targeting CRISPRs.

The fusion of artificial intelligence and RNA-targeting CRISPR screens enables precise modulation of gene dosage, which determines the level of gene expression. The research team demonstrated that TIGER’s off-target predictions can be employed to achieve partial inhibition of gene expression using mismatched guide RNAs. This breakthrough has potential implications for diseases characterized by an abnormal number of gene copies, such as Down syndrome, certain forms of schizophrenia, Charcot-Marie-Tooth disease (a hereditary nerve disorder), and cancers driven by aberrant gene expression.

The team of researchers believes that their AI-powered predictions, coupled with RNA-targeting CRISPR screens, will help prevent unintended off-target CRISPR activity. This advancement is expected to accelerate the development of a new generation of RNA-targeting therapies.

As larger datasets from CRISPR screens become available, the integration of sophisticated machine learning models offers increasing opportunities for advancement. The collaboration between the research teams involved in this study highlights the power of interdisciplinary cooperation. The deployment of TIGER allows for the prediction of off-target effects and precise modulation of gene dosage, opening up exciting new applications for RNA-targeting CRISPRs in biomedicine, thereby propelling the field forward.

By Impact Lab