They present a model capable of understanding and generating biological information.

2024/11/18 Etxebeste Aduriz, Egoitz - Elhuyar Zientzia Iturria: Elhuyar aldizkaria

• Genetika
• Adimen artifiziala

They've presented a machine-learning model that can accurately decode and design DNA, RNA and protein sequences. They call it EVO. With the analysis of millions of microbial genomes, EVO has developed the ability to understand the genetic code. Thus, for example, it is able to predict whether mutations will have a biological effect or design new sequences. The work has been published in the journal Science.

The EVO was designed to generate full-genome DNA sequences, and was trained with microorganism genome data for 2.7 million euros. And according to the developers, it has a great ability to interpret and generate biological information with great precision. It accurately predicts the influence of mutations on bacterial proteins and RNA, as well as on the modeling of genetic regulation. Likewise, EVO understands the complex co-evolution between encoding and non-coding sequences. And it's able to generate sequences of over 1 megabase.