A new startup, EvolutionaryScale, has made a significant entrance into the biotech industry with a remarkable $142 million seed funding round. This funding aims to support the development of advanced AI models that can generate novel proteins for scientific research. The startup announced its ambitious plans and introduced ESM3, a pioneering AI model designed to revolutionize protein design for applications in drug discovery and materials science.
Massive Funding and Prominent Backers
EvolutionaryScale’s seed round was led by former GitHub CEO Nat Friedman, Daniel Gross, and Lux Capital, with participation from Amazon and NVentures, Nvidia’s corporate venture arm. The substantial investment underscores the potential seen in EvolutionaryScale’s innovative approach to biotechnology.
Unveiling ESM3: A Frontier Model in Biology
ESM3, the AI model introduced by EvolutionaryScale, represents a significant advancement in protein design. Co-founder and chief scientist Alexander Rives described ESM3 as a tool that brings AI into the realm of biological engineering, akin to how AI is used in engineering structures, machines, and microchips.
“ESM3 takes a step toward a future of biology where AI is a tool to engineer from first principles,” Rives stated. This model has the capability to generate proteins that can be utilized in various scientific fields, potentially leading to new classes of drugs, tools, and therapeutics.
The journey of EvolutionaryScale’s founders began at Meta’s AI research lab, FAIR, in 2019. Rives, along with Tom Secru and Sal Candido, initially developed generative AI models to explore proteins. After their team at Meta was disbanded, they decided to continue their groundbreaking work independently, leading to the creation of EvolutionaryScale.
The Science Behind Protein Design
Protein design is a complex process that involves creating a structure capable of performing specific tasks within the body or a product, followed by identifying a protein sequence likely to fold into the desired structure. Proteins must fold correctly into three-dimensional shapes to function effectively. ESM3, trained on a dataset of 2.78 billion proteins, can “reason over” the sequence, structure, and function of proteins, enabling the generation of new proteins similar to Google DeepMind’s AlphaFold.
EvolutionaryScale has already demonstrated the potential of ESM3 by generating a new variant of green fluorescent protein (GFP), known for its role in the luminescence of jellyfish and coral. A preprint paper detailing this work is available on the company’s website, showcasing the practical applications of their AI model.
EvolutionaryScale plans to make ESM3 available for non-commercial use through its cloud Forge developer platform. Additionally, a smaller version of the model will be released for offline use. The company aims to generate revenue through partnerships, usage fees, and revenue sharing, potentially collaborating with pharmaceutical companies and researchers to integrate ESM3 into their workflows.
To further extend its reach, EvolutionaryScale will bring ESM3 to select AWS customers via the SageMaker AI dev platform, Bedrock AI platform, and HealthOmics service. The model will also be available to Nvidia customers through the NIM microservices, supported by an Nvidia enterprise software license. Both AWS and Nvidia customers will have the ability to fine-tune ESM3 using their own data, enhancing the model’s utility for specific applications.
Despite the promising technology and substantial funding, EvolutionaryScale faces significant challenges. The company acknowledges that it may take up to a decade for generative AI models to assist in designing effective therapies. Additionally, EvolutionaryScale must contend with established competitors such as DeepMind’s spin-off Isomorphic Labs, Insitro, Recursion, and Inceptive, all of which have made strides in the biotech industry.
A Vision for the Future
EvolutionaryScale’s long-term vision involves scaling its model training beyond proteins to create a general-purpose AI model for various biotech applications. The company believes that the rapid advancements in AI, driven by larger models, expansive datasets, and increasing computational power, will also drive breakthroughs in biology.
“The incredible pace of new AI advances is being driven by increasingly large models, increasingly large data sets, and increasing computational power,” an EvolutionaryScale spokesperson said. “The same holds true in biology. As language models scale, they develop an understanding of the underlying principles of biology and discover biological structure and functions”.
EvolutionaryScale’s ambitious plans, supported by deep-pocketed investors, aim to push the boundaries of what is possible in protein design and biotechnology. While the road ahead may be challenging, the potential for groundbreaking discoveries and new therapeutics offers a compelling vision for the future of AI in biology.
