AI designs whole genomes but synthetic life remains out of reach
Mar 6th 2026
The Evo2 AI model can read and write DNA and has produced genome‑scale designs, but researchers warn that synthesis, functional testing and biological complexity still block AI from creating new living cells.
- Evo2 is a DNA language model trained on trillions of genetic letters that can generate full genome sequences.
- Researchers used Evo2 to design genomes inspired by Mycoplasma genitalium, human mitochondria and a yeast chromosome.
- Computational analysis found about 70 percent of genes in the M. genitalium-inspired designs looked realistic.
- Earlier work in 2008 chemically synthesized a M. genitalium genome and later rebooted genomes inside cells to create what was called synthetic life.
- In 2025 Evo models generated phage genomes that produced 16 functional viruses out of 285 designs when tested in E. coli.
- Scientists say major hurdles remain, including the need to chemically synthesize sequences and test them at scale inside living cells.
- Even one missing or poorly modeled essential gene or incorrect gene ordering can make a designed genome nonfunctional.
- Researchers call the progress transformative for genome engineering but stress that AI-designed living microbes have not yet been created