In a significant development for biotechnology, EvolutionaryScale, an advanced AI research lab, launched with ESM3, an AI model capable of generating novel proteins. This innovative model has already achieved a remarkable feat: the creation of a new Green Fluorescent Protein (GFP), a process that would take 500 million years of natural evolution to occur.

The ESM3 model represents a major leap in generative AI, enabling scientists to interactively prompt the creation of proteins. This advancement promises to impact fields such as drug discovery, materials science, and carbon capture. The team behind ESM3, pioneers in the application of AI to biology, previously developed the first transformer language model for proteins, ESM1. Their models have facilitated significant research progress, including advances in protein folding that have illuminated the structures of hundreds of millions of metagenomic proteins. These tools are now used globally by scientists to model and understand protein structures.

Trained with 1 trillion teraflops on a dataset of 2.78 billion proteins, ESM3 utilizes more computational power than any other known biological model. It is the first generative model for biology that simultaneously considers the sequence, structure, and function of proteins, allowing scientists to both understand and create new proteins, thus making biology programmable.

“ESM3 takes a step toward a future of biology where AI is a tool to engineer from first principles, the way we engineer structures, machines, and microchips, and write computer programs,” said EvolutionaryScale co-founder and chief scientist, Alexander Rives. “We’ve been working on this for a long time, and we’re excited to share it with the scientific community and see what they do with it.”

The potential applications of ESM3 are vast, ranging from developing new cancer treatments to creating proteins that capture carbon. By reasoning over possible sequences and structures, ESM3 bypassed 500 million years of evolution to create a new fluorescent protein. GFP, known for its role in jellyfish and coral fluorescence, has become a crucial tool in molecular biology for visualizing cellular processes. The creation of a new GFP variant by ESM3 underscores the model’s impressive capabilities in advancing biological research.

EvolutionaryScale is opening an API for closed beta today, with code and weights available for a smaller, non-commercial version of ESM3. Collaborating with Amazon Web Services (AWS) and NVIDIA, EvolutionaryScale aims to accelerate applications from drug discovery to synthetic biology using AI.

By partnering with AWS, EvolutionaryScale ensures that the full ESM3 model family is accessible to hundreds of thousands of researchers worldwide and nine of the top ten global pharmaceutical companies. This collaboration allows researchers to securely fine-tune the ESM3 models using their proprietary data, at scale. Additionally, ESM3 will be optimized for training and inference performance through NVIDIA BioNeMo NIMs, supported by the NVIDIA AI Enterprise software license.

EvolutionaryScale also announced a seed funding round of more than $142 million, led by Nat Friedman and Daniel Gross, along with Lux Capital, Amazon, NVentures (NVIDIA’s venture capital arm), and angel investors. This funding will be used to further expand the capabilities of its innovative models.

‍