EvolutionaryScale, a pioneering AI-biotech startup, has been established by former Meta researchers. The company is set to revolutionize the field with its innovative tool, which empowers scientists to craft entirely novel molecules. This cutting-edge technology is poised to expedite the development of superior medicines, biofuels, and materials by emulating the evolutionary processes of billions of years.
Researchers increasingly use AI to develop new molecules for various applications, from improving protein structures to designing gene editors. EvolutionaryScale's tool, powered by a language model similar to those behind AI chatbots, helps create new proteins by predicting patterns in extensive biological data.
Proteins’ cell functions are primarily determined by their structures, which have evolved over billions of years. EvolutionaryScale's generative AI model, ESM3, was trained on data from 2.78 billion natural proteins, including their amino acid sequences, 3D structures, and functions. The model uses this data to design new proteins with specific functions.
In one test, the AI model created a new green fluorescent protein (GFP) with a sequence that was only 58% identical to existing proteins, simulating over 500 million years of evolution. "We think ESM3 makes a meaningful step toward tools that can allow you to design biology from first principles — to engineer biology," said Alexander Rives, EvolutionaryScale’s co-founder and chief scientist.
EvolutionaryScale is partnering with Amazon Web Services and Nvidia to make the AI model available to select customers. The model is also accessible for non-commercial use via an API, and the code and weights will be released for a smaller open version of ESM3.
As AI continues to make strides in the biotech sector, the issue of responsibility looms large. Alexander Rives, co-founder and chief scientist of EvolutionaryScale, along with a consortium of over 160 researchers, has issued a call for the establishment of ethical principles to guide the responsible development of AI for protein design. This proactive stance underscores EvolutionaryScale's commitment to anticipating and mitigating potential risks.
Ultimately, these AI tools must prove their value to the pharmaceutical industry, which has historically been skeptical of computational methods. Rives believes new predictive AI tools can significantly impact the industry by enabling more rational engineering approaches.
Comments