Researchers at Stanford Medicine have created a new artificial intelligence tool to make gene-editing experiments easier and faster to design. The technology, called CRISPR-GPT, serves as an AI-powered “copilot” for gene editing, helping researchers — even those with little or no prior experience — generate experimental designs, analyze results, and identify potential flaws.

The tool builds on CRISPR, the groundbreaking gene-editing system widely used to modify genomes and develop therapies for genetic diseases. While powerful, learning to design CRISPR experiments can be challenging and time-intensive, even for experienced scientists. CRISPR-GPT streamlines this process by automating many aspects of experiment planning and refinement.

“Our goal is to accelerate the development of lifesaving therapies,” said Le Cong, PhD, assistant professor of pathology and genetics, who led the project. “We hope CRISPR-GPT will allow us to create new drugs in months rather than years. By speeding up the design process and supporting collaboration between students, trainees, and researchers, this technology could ultimately help save lives.”

The platform also lowers barriers for entry into gene editing. For example, a student in Cong’s lab used CRISPR-GPT to successfully disable several genes in lung cancer cells on the very first try — a task that usually requires extensive trial and error. This ability to flatten CRISPR’s steep learning curve could broaden access to gene-editing tools across biotechnology, agriculture, and medicine.

“Science is often built on trial and error,” Cong noted. “But what if we could make it trial and done?” 

Le Cong, senior author of a study published July 30 in Nature Biomedical Engineering, and lead authors Yuanhao Qu and Kaixuan Huang, have developed CRISPR-GPT, an AI tool that streamlines CRISPR experiment design. Trained on 11 years of data and expert discussions, CRISPR-GPT predicts off-target edits, reduces trial-and-error, and explains its reasoning to users through a chat interface.

In early tests, undergraduate Yilong Zhou used the tool to successfully activate genes in melanoma cells on his first attempt — a rare feat in gene editing. The system offers beginner, expert, and Q&A modes, making it accessible for both novices and advanced researchers while enhancing collaboration and knowledge-sharing.

Built with safeguards to prevent misuse, CRISPR-GPT is designed for responsible application and may serve as a blueprint for AI agents in other areas of genomic research. The team also launched Agent4Genomics, a platform for related AI tools.

Collaborators included Google DeepMind, Princeton, and UC Berkeley, with funding from NIH, the Baxter Foundation, the Weintz Family Foundation, and NSF.

Source: https://med.stanford.edu/news/all-news/2025/09/ai-crispr-gene-therapy