How Are AI Models Designing New Viruses?

The intersection of artificial intelligence and biology has been a game-changer, especially when it comes to designing new proteins - and now, a new frontier has emerged, focusing on large genome models that can actually design new viruses. These models are trained on DNA sequences, and what's really cool is that they can output new sequences that encode functional proteins in bacteria. Take the researchers at Stanford University, for instance - they've used these models to design new viruses that infect bacteria, including one that's genetically distant from an existing bacteria-killing virus.

The tech behind this is based on large genome models, which are kinda like large language models, but applied to DNA - so, they're trained on a vast amount of genome sequences and can recognize the biological context of the sequence they're outputting. By prompting these models with a sequence of part of a gene cluster, they can output DNA sequences that encode proteins with related functions, which has the potential to create new proteins that haven't been identified before. It's a pretty powerful tool, and the possibilities are endless.

But, as with any new technology, there are limitations - particularly when it comes to understanding the functions of the outputted genes. I mean, while the models can output sequences that contain genes and regulatory DNA, the lack of knowledge about the functions of these genes limits the ability to test them. To mitigate this risk, researchers have been careful about what they feed into the models, avoiding sequences from viruses that target complex cells and focusing instead on viruses that attack bacteria.

The potential risks and benefits of this technology are still being explored, and it's a bit of a double-edged sword. On one hand, the ability to design new viruses could lead to significant advancements in fields like biotechnology and medicine - but on the other hand, there's a risk that this technology could be used to create viruses that target vertebrates, including humans. So, researchers are calling for caution and careful consideration of the potential consequences of this technology.

Source: Ars Technica