top of page

ESM3 - Simulating 500 Million Years of Evolution to Create New Proteins

Simulating 500 Million Years of Evolution to Create New Proteins

In a groundbreaking leap for artificial intelligence and biotechnology, the newly unveiled ESM3 model has simulated an astounding 500 million years of evolution to create new proteins. This language model, designed specifically for biological engineering, promises to revolutionize our approach to understanding and programming life itself.

At its core, ESM3 is a frontier language model akin to ChatGPT and Gemini, but with a singular focus: to read and write the code of life. By simulating evolutionary processes, ESM3 can design proteins with unprecedented precision and efficiency. Proteins, the molecular engines that power life, have traditionally been understood and manipulated through painstaking experiments. ESM3, however, bypasses this trial-and-error method by using advanced AI to predict and create protein sequences, structures, and functions.

One of ESM3's most impressive feats is the creation of a new green fluorescent protein (GFP). GFPs are vital in biotechnology because they can easily be visualized, making them ideal for tracking and studying biological processes. ESM3 generated a GFP with a sequence only 58% similar to the closest known variant, a task that would naturally take hundreds of millions of years of evolution. This breakthrough exemplifies ESM3's potential to revolutionize biological research and applications.

The ability to program life opens up vast possibilities. From developing new medicines and creating enzymes that can break down plastics to engineering life forms that capture carbon, the potential applications are endless. ESM3's ability to reason over the fundamental biological properties of proteins—sequence, structure, and function—marks a significant advancement in our quest to understand and manipulate the building blocks of life.

In a commendable move towards open science, the team behind ESM3 is releasing the model's weights and code to the public. This transparency aims to accelerate progress, reduce risks, and maximize positive impacts globally. Researchers and developers can access the ESM3 model on GitHub under a non-commercial license, potentially spurring a wave of innovation in biotechnology.

The advent of ESM3 represents a confluence of AI and biology that could redefine our understanding of life. By making biology programmable, we stand on the brink of monumental advancements in medicine, environmental science, and beyond. With open science leading the charge, the future of programmable biology is bright, promising a world where we can engineer life to address some of humanity's most pressing challenges.


bottom of page