Artificial intelligence designs thousands of new DNA switches to precisely control gene expression

According to a paper published in Nature magazine on the 23rd, teams from the Jackson Laboratory, the Broad Institute of MIT and Harvard, and Yale University used artificial intelligence (AI) technology to design thousands of new DNA switches. These newly designed components can precisely control the expression of genes in different types of cells, providing unprecedented possibilities for human health and medical research.

Graphic representation of how cis-regulatory elements work to turn genes on or off, which could lead to more precise and personalized gene therapies. Image credit: Broad Institute of MIT and Harvard

Although gene editing and other gene therapy methods have enabled scientists to modify genes in living cells in recent years, it is still challenging to perform genetic intervention on only a certain type of cell without interfering with the entire organism. This is mainly because the understanding of the cis-regulatory elements (CREs), the DNA switches that control the turning on and off of genes, is not deep enough. The core of this innovation is that the new method can increase or decrease gene expression in specific cell types without affecting other parts of the body.

The team used a deep learning algorithm to train a model based on hundreds of thousands of DNA sequences in the human genome. Through this model, they can measure the activity of CRE in three types of cells (blood, liver and brain) in a laboratory environment. The AI ​​model can predict the activity of any sequence, thus revealing new patterns in DNA and how the syntax of CRE sequences affects the amount of RNA produced.

Based on the above findings, the team built a platform called CODA (Computational Optimization of DNA Activity). Through an iterative process combining experimental data and computational modeling, the platform continuously improved its ability to predict the biological effects of CREs and successfully designed CREs that have never appeared in nature.

After testing, the newly designed synthetic CREs showed better cell type specificity than naturally occurring CREs. They not only contain sequences that promote gene expression in target cell types, but also contain elements that inhibit gene expression in non-target cell types. Currently, the team has verified the effectiveness of several synthetic CRE sequences in zebrafish and mice.

This breakthrough means that more precise and personalized gene therapies may be possible in the future, opening up new avenues for preventing and treating diseases.

In organisms, not all genes work in the right cells and at the right time. The function of CRE is to ensure that the right genes are activated at the right time and in the right place. For example, it prevents skin cells from using genes needed by the brain, or prevents the activation of genes that are only needed in early development stages in adults. This time, the team not only synthesized new CREs, but also demonstrated how to use these CREs to selectively activate genes in brain, liver or blood cells, while ensuring that these genes remain silent in other types of cells. This is of doubly important significance in biomedicine and technology.