Revolutionary AI-Designed DNA Controls Genes in Healthy Mammalian Cells for First Time

Researchers at the Centre for Genomic Regulation (CRG) have developed an AI tool capable of designing synthetic DNA molecules that control gene expression in healthy mammalian cells, marking a first reported instance in the journal Cell. The tool generates custom regulatory sequences not found in nature, enabling precise control over gene activity—such as activating specific genes in stem cells destined to become red blood cells but not platelets. As proof-of-concept, the AI-designed enhancers successfully activated fluorescent proteins in mouse blood cells without disrupting other genetic functions. This breakthrough could lead to ultra-selective therapies with fewer side effects by engineering DNA sequences tailored to specific cellular needs. The study involved synthesizing over 64,000 synthetic enhancers and analyzing their interactions across seven stages of blood-cell development, revealing mechanisms like negative synergy where transcription factors can repress genes when combined. Funded by ERC and Spanish National Agency grants, the research underscores the potential for AI-driven solutions in gene therapy and precision medicine.

AI-designed DNA controls genes in healthy mammalian cells for first time

The study published in Cell marks a significant advancement by successfully using generative AI to design synthetic DNA molecules that control gene expression in healthy mammalian cells for the first time. Researchers at CRG developed an AI tool capable of generating novel DNA regulatory sequences tailored to specific criteria, such as activating genes in particular stem cell types while leaving others unaffected.

The AI model predicts optimal DNA combinations, which are synthesized and introduced into cells via viral vectors. In a proof-of-concept experiment, these enhancers were tested on mouse blood cells, successfully activating fluorescent proteins without disrupting other genetic activities. This precise control opens promising avenues for gene therapy, enabling targeted modulation of gene activity in specific cell types or tissues.

The research highlights the potential for AI-designed enhancers to address diseases caused by faulty cell-type-specific gene expression. By constructing a large library of synthetic enhancers and analyzing their effects across different stages of blood-cell development, the team uncovered mechanisms such as negative synergy, where two factors that individually activate genes can suppress them when combined.

Using healthy cells rather than cancer lines provided more representative insights into human biology, achieved through extensive experiments over five years. The data from these studies trained the AI model to predict new enhancer designs capable of achieving specific on-off patterns, even if they don’t naturally occur.

This method has potential applications in gene therapy for diseases involving overactive or underactive genes, such as cancer or genetic disorders. The use of AI accelerates design processes and allows for complex sequences beyond human capability. However, considerations remain regarding scalability, integration into diverse cell types, and ensuring specificity to avoid unintended genomic disruptions.

Overall, this research represents a promising advancement in gene therapy, offering precise targeting of specific genes with potential for safer and more effective treatments.