Biological research

Gigantic AI system promises new era of medicine

The artificial intelligence (AI) model called Evo 2 was trained with almost nine trillion pieces of genetic information from more than 128,000 species.(Bild: Deemerwha studio)

A research group says it has created the largest AI system for biological research to date using a huge amount of genetic information to help develop medical treatments.

The artificial intelligence (AI) model, called Evo 2, was trained with almost nine trillion pieces of genetic information from more than 128,000 species, the team, funded by the Arc Institute research organization, announced on Wednesday (local time).

2000 Nvidia processors do the math
The model was created using 2000 H100 processors from the US semiconductor giant Nvidia. The extensive genetic data is stored in an Amazon Web Services (AWS) cloud. They are freely accessible to scientists all over the world via the Nvidia research platform BioNeMo.

With the help of Evo 2, knowledge of "complex" diseases in humans should be improved quickly, explained Arc Institute Director Silvana Konermann. Specifically, the aim is to identify "which variant of a gene is associated with a particular disease". This knowledge could then be used to develop new molecules that specifically target the cause of the disease.

First results for cancer research
Researchers from the Californian elite universities Stanford and Berkeley as well as the University of California in San Francisco (UCSF) are involved in the project. According to the Arc Institute, the AI model has already delivered results for cancer research. For example, Evo 2 has been used to determine with 90 percent certainty which mutation of the BRCA1 gene, which is associated with breast cancer, is benign and which is malignant.

According to its developers, the model could also be used to modify the genetic material of crops so that they are better adapted to climate change and contain more nutrients. According to Nvidia, Evo 2 could also help with the development of enzymes that break down plastic.