New Delhi:
A New York-based AI research company focused on biotechnology has generated a new protein using a Protein Language Model (ProLLMs) that works on a similar transformer architecture to that of ChatGPT.
EvolutionaryScale on June 25 unveiled a first-of-its-kind AI that has generated a new protein molecule that glows – and mimics the bioluminescence of a jellyfish molecule called green fluorescent protein. The new protein sequence is significantly different (less than 60 percent similarity) to the natural protein: a difference that the company may call “more than 500 million years of (natural) evolution.”
The company used its groundbreaking AI language model called EvolutionaryScale Model-3 (ESM3) to achieve this feat, raising $142 million in a seed funding round, including investments from industry giants like Nvidia and Amazon.
We’re excited to partner with AWS and NVIDIA to push the boundaries of AI for the life sciences.
— Evolutionary Scale (@EvoscaleAI) June 25, 2024
The ESM3 differs from ChatGPT because it is trained on parameters (internal variables) of three fundamental biological properties of proteins: sequence, structure and functions. The model was trained on 98 billion parameters, making it the largest biological AI model to date.
EvolutionaryScale calls this a “model trained throughout evolution.” The training set included 2.78 billion natural proteins ranging from “the Amazon rainforest to the depths of the ocean, extreme environments such as hydrothermal vents and microbes in a handful of soil.”
ESM3 allows users to generate proteins, using clues with partial information (sequence, structure and function keywords) and iterate the model to make predictions until the entire sequence is completed. The model is mainly intended for scientists and gives them unprecedented control over the process of generating proteins.
We’ve been training ESM3 and we’re excited to introduce EvolutionaryScale.
ESM3 is a generative language model for programming biology. In experiments, we found that ESM3 can simulate 500 million years of evolution to generate new fluorescent proteins.
Read more: https://t.co/iAC3lkj0iVpic.twitter.com/AhWtC4vxlF
— Alex Rives (@alexrives) June 25, 2024
EvolutionaryScale states that their goal is to make biology programmable. “ESM3 takes a step toward the future where AI is a tool to engineer biology from first principles, in the same way we design structures, machines and microchips and write computer programs,” according to the company’s website.
The application of this technology can lead to breakthroughs in several areas, such as drug discovery and development, biomedical research and sustainability. An example of this is already being demonstrated by EvolutionaryScale by demonstrating a protein prototype capable of breaking down plastic waste.
The possibilities are endless, with every cell in every organism containing ribosomes (protein complexes responsible for protein synthesis). However, there are also concerns that AI could be misused to create biological weapons.
Scientists have taken a proactive approach and established “Common Values, Guiding Principles and Commitments for the Responsible Development of AI for Protein Design” in March, in an effort to guide developments in this field for the betterment of humanity.
We are advancing a new global agreement, signed by more than a hundred leading scientists, to ensure that AI technologies for protein design are developed responsibly. This field can produce medicines, vaccines and other innovations that benefit everyone. https://t.co/pTlFtBWHNhhttps://t.co/XYn986dEAt
— Institute for Protein Design (@UWproteindesign) March 8, 2024
EvolutionaryScale has also been praised by experts for releasing a smaller open source version that others can use freely. The large-scale complete model has not been released, although its training process has been made public in an effort to remain transparent and share the technology freely.