Researchers at Stanford Medicine have created an artificial intelligence (AI) tool called CRISPR-GPT. This tool acts as a helpful guide for gene-editing work. It supports scientists in designing experiments, checking data, and fixing problems. Even those new to the field can use it easily.
The tool is built around CRISPR, a technology that cuts and edits DNA. CRISPR helps develop treatments for diseases caused by faulty genes, like sickle cell anemia, where red blood cells form wrong shapes. But planning CRISPR experiments often takes months of testing ideas, which is hard and slow, even for experts.
CRISPR-GPT changes that by automating much of the work. It uses data from 11 years of scientific talks and papers to suggest smart plans. Users chat with it through a text box, sharing goals and gene details. The tool then outlines steps, explains choices, and warns of past mistakes in similar tests. It also predicts unwanted changes, called off-target edits, where CRISPR might alter the wrong DNA spot.
The researchers hope CRISPR-GPT will cut drug creation time from years to months, saving lives. A student used it to turn off genes in lung cancer cells on the first try, skipping usual trial-and-error. Another student activated genes in skin cancer cells to study why immune treatments fail.
How it works in practice
CRISPR-GPT has three modes. Beginner mode teaches by explaining each step. Expert mode partners equally on tough issues. Q-and-A mode answers quick questions. It feels like chatting with a lab helper, users say. This makes knowledge sharing easy across labs and reduces errors.
The tool checks designs and applies ideas to new diseases. But safety matters. It blocks unethical requests, like editing viruses or embryos, with warnings. The researchers plan to share CRISPR-GPT with agencies for ethical checks.
In the future, CRISPR-GPT could inspire AI for other biology tasks, like growing stem cells or studying heart disease paths. A website called Agent4Genomics offers it and similar tools for free use. This study is published in Nature Biomedical Engineering.