Scientists from leading institutions worldwide have developed a method to precisely control gene expression in various cell types, marking a groundbreaking step in the treatment of genetic diseases. This innovation, driven by artificial intelligence, addresses a long-standing challenge in genetics.
All living organisms possess the same genetic code, or DNA, but the activation of specific genes varies according to the cell's function. Cis-regulatory elements (CREs), often referred to as 'DNA buttons,' play a crucial role in ensuring that the right genes are activated at the right time.
Despite advances in manipulating genes within living cells, selectively turning genes on and off in specific cell types has proven difficult. A significant barrier has been the unpredictability of how different CREs behave. Although the human genome contains thousands of distinct CREs, scientists have struggled to interpret their 'language' effectively.
To tackle this issue, researchers from Jackson Laboratory, the Massachusetts Institute of Technology (MIT), the Broad Institute of Harvard University, and Yale University turned to artificial intelligence. Their study, published in the prestigious journal Nature on October 23, 2024, introduced a deep learning model capable of predicting CRE activity.
By measuring the activity of CREs in blood, liver, and brain cells, the AI model was trained on hundreds of thousands of DNA sequences from the human genome. Following this training, the AI demonstrated the ability to accurately predict CRE activities across an almost infinite array of combinations.
Moreover, the research team designed thousands of new synthetic CREs that can control gene expression in selected cell types. Testing these synthetic CREs in live animals, they successfully activated a protein exclusively in the livers of developing zebrafish, without affecting other cells or tissues.
This advancement opens up new avenues for developing treatments for various genetic disorders by enabling precise control over gene expression in targeted areas. Ryan Tewhey, a co-author of the study from Jackson Laboratory, stated, 'This technology paves the way for writing new regulatory elements with predefined functions.'