AI just designed NEW viruses

AI is beginning to move beyond predicting what biological molecules might look like.

In a new study published in Science, researche used AI Models Evo 1 and Evo 2 which are trained on huge amounts of DNA sequence data and learn patterns that have emerged through evolution. The researchers wanted to find out whether such models could use those patterns to create a complete, functional viral genome.

The team generated thousands of potential phage genomes using AI. They then chemically synthesized nearly 300 of the designs and tested them in the laboratory. Sixteen of the AI designed genomes produced viable phages.

The successful designs contained new combinations of mutations, genes and regulatory elements. Some also differed in genome length from their natural counterparts. One design included a DNA packaging protein related to a much more distant phage.

The researchers also found that some of the designed phages could infect their intended bacterial host effectively. 

So, what does “AI-designed viruses” actually mean here?

It does not mean that AI has suddenly created a virus capable of infecting people. The study involved bacteriophages that target bacteria, and the work was conducted to explore how AI could help design useful biological systems.

The bigger significance is that AI may be making genome engineering more flexible. Instead of changing one gene at a time, researchers may eventually be able to design larger portions of a genome according to specific goals.

The study therefore represents a milestone in synthetic biology: AI is beginning to demonstrate that it can generate functional genomes that are different from those found in nature. The researchers suggest that this approach could eventually support applications such as developing phage-based therapies against bacteria that have become resistant to existing treatments.

Rather than being a story about AI “creating dangerous viruses,” this is better understood as a demonstration of a new scientific capability: using AI to explore biological designs that evolution has not necessarily produced naturally. Refer following link for more details: https://www.science.org/doi/10.1126/science.aec2657