ESM3 is a groundbreaking multimodal generative language model designed to reason over protein sequence, structure, and function simultaneously1. Developed by researchers at EvolutionaryScale and the Arc Institute, this AI model represents a significant advancement in computational biology:

• Trained on an unprecedented 771 billion unique tokens derived from 3.15 billion protein sequences, 236 million protein structures, and 539 million proteins with function annotations2

• Utilizes up to 98 billion parameters, making it one of the most computationally intensive biological models to date3

• Employs a "chain of thought" process for protein design, iteratively adjusting sequences and structures to optimize functionality4

• Available in public beta via an API, allowing scientists to engineer proteins programmatically or through interactive browser-based applications2

ESM3's ability to generate functional proteins far removed from known sequences demonstrates its potential to revolutionize protein engineering and expand our understanding of evolutionary processes5.

esmGFP, the novel protein created by the ESM3 AI model, represents a significant leap in protein engineering. This green fluorescent protein shares only 58% sequence similarity with its closest known natural counterpart, a modified version of a protein found in bubble-tip sea anemones12. The creation of esmGFP required simulating 96 different genetic mutations, a process that would have taken over 500 million years to occur naturally1.

Key features of esmGFP include:

• Its unique genetic sequence, which exists only as computer code but contains the blueprint for a previously unknown type of green fluorescent protein1

• A maturation process that takes about a week, compared to less than a day for natural GFPs3

• Initial brightness levels 50 times lower than natural GFPs, though subsequent iterations achieved comparable brightness3

• Potential applications in medicine, environmental research, and various other scientific fields4

The development of esmGFP demonstrates the power of AI in exploring vast protein sequence spaces and uncovering functional proteins that nature may never have produced56.

The creation of esmGFP marks a significant milestone in the field of protein engineering, demonstrating the potential of AI to accelerate evolutionary processes that would take eons in nature. This breakthrough has far-reaching implications for biological research and biotechnology:

• It showcases the ability of language models to generate functional proteins that are vastly different from known proteins, opening up new avenues for exploring protein space12

• The success of esmGFP validates the approach of using AI to simulate long-term evolutionary processes, potentially revolutionizing how scientists approach protein design and engineering34

• By creating a protein that nature may never have produced on its own, this research challenges our understanding of the limits of natural protein evolution and expands the possibilities for synthetic biology56

The development of AI-driven proteins like esmGFP opens up a wide range of potential applications across various scientific and industrial fields:

• Medical research: AI-generated proteins could lead to new fluorescent markers for tracking disease progression or drug efficacy in living tissues1. These novel proteins may also serve as the basis for developing targeted therapies or diagnostic tools2.

• Environmental monitoring: Engineered fluorescent proteins could be used to detect pollutants or track environmental changes in ecosystems3. Their unique properties might allow for more sensitive and specific detection methods compared to existing technologies.

• Biotechnology: The ability to rapidly design and synthesize new proteins could revolutionize industrial processes, potentially leading to more efficient biofuels, enzymes for waste degradation, or bio-based materials4. This could contribute to more sustainable manufacturing practices and reduce reliance on petrochemicals.

• Fundamental research: AI-driven protein engineering provides a powerful tool for exploring protein structure-function relationships, potentially uncovering new insights into evolutionary biology and the origins of life5. This could help scientists better understand how proteins evolve and function, leading to breakthroughs in fields like structural biology and biochemistry.