AI can write genomes — how long until it creates synthetic life?

2026-03-04 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Mathematics & Computational Sciences, Research Methodology & Innovation · Depth: Intermediate, quick

Summary

The Evo2 genomic language model, an AI designed to read, interpret, and generate DNA, RNA, and protein sequences, marks a significant advancement in synthetic biology by designing whole genome sequences, including one inspired by *Mycoplasma genitalium*. Trained on trillions of DNA letters, Evo2 previously generated functional phage viruses, with 16 out of 285 designs successfully killing bacteria, as detailed in a *Nature* paper published on March 4. However, experts like Nico Claassens highlight major challenges, such as the need for scalable synthesis and testing, and the difficulty in designing genomes that direct *all* essential functions for even simple life forms, noting that "You cannot design life 70%". Despite these hurdles, researchers like Patrick Yizhi Cai consider these AI models the "ChatGPT moment" for synthetic genomics, making the creation of novel biological entities a tangible goal. The ultimate aim is to create AI-generated microbial life, moving beyond mere genome editing to designing entirely new functional genomes.

Key takeaway

The Evo2 DNA language model now generates entire genome sequences, including bacterial and mitochondrial designs, marking a "ChatGPT moment" for synthetic genomics. While previous Evo models produced functional phage viruses (16/285 designs), current bacterial genome designs like the *M. genitalium*-inspired sequence only show ~70% predicted realism, highlighting challenges in ensuring full functionality. This advance offers AI/ML professionals a powerful tool for novel biological system design, despite significant hurdles in scaling synthesis and achieving fully functional synthetic life.

Topics

• Genomic Language Models
• Synthetic Genomics
• Evo2 AI Model
• Synthetic Biology
• Genome Design

Best for: AI Researcher, AI Scientist, Research Scientist

Related on AIssential

• Researchers try to cut the genetic code from 20 to 19 amino acids — Researchers from Columbia and Harvard successfully engineered *E.
• [R] Genomic Large Language Models — Genomic LLMs can identify functional similarities in DNA beyond sequence alignment, indicating learned regulatory patterns.
• Leap in DNA Synthesis Slashes Time to Build New Genetic Sequences — Sidewinder enables rapid, high-fidelity DNA synthesis, resolving a key bottleneck for generative AI in biology.
• Diffusion model generating regulatory DNAs — DNA-Diffusion is a diffusion model generating functional regulatory DNA sequences conditional on cell type.
• RNA design across eras: from covariance models to modern generative AI — RNA's functional versatility relies on 3D folding, with generative AI building on foundational probabilistic models for structure prediction.
• 😺 🎙️ Watch: It takes 2 years to build a $1B startup now (and 2 weeks to replace 8 years of R&D) — AI is profoundly reshaping venture capital and startup dynamics, with half of all funding now directed to AI-native companies, accelerating the path to a \$1B valuation to just 2-3 years and boosting solo founder success.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.