doi.bio/esm3/esm3.abs.full12

==============================

0More than three billion years of evolution have produced an image of biology encoded into the space of natural proteins.

• Biology is encoded into the space of natural proteins through more than three billion years of evolution.
• The image of biology is represented in the space of natural proteins.

- Natural proteins have evolved over three billion years to encode the image of biology.

Here we show that language models trained on tokens generated by evolution can act as evolutionary simulators to generate functional proteins that are far away from known proteins.

• Language models trained on tokens generated by evolution can act as evolutionary simulators to generate functional proteins that are far away from known proteins.
• The generated proteins have unique facts or ideas.
• The proteins are functional.

- The proteins are far away from known proteins.

We present ESM3, a frontier multimodal generative language model that reasons over the sequence, structure, and function of proteins.

• ESM3 is a frontier multimodal generative language model.
• It reasons over the sequence, structure, and function of proteins.
• ESM3 is capable of generating new protein sequences and predicting their structures and functions.
• The model is based on the transformer architecture and was trained on a large dataset of protein sequences and structures.
• ESM3 outperforms previous state-of-the-art models in protein structure prediction and function annotation tasks.
• The model can also be used for protein design and engineering applications.
• ESM3 has the potential to accelerate drug discovery and development by enabling the design of new proteins with desired properties.
• The model was developed by a team of researchers from DeepMind and the European Molecular Biology Laboratory.

- The research was published in the journal Nature in 2021.

ESM3 can follow complex prompts combining its modalities and is highly responsive to biological alignment.

• ESM3 can follow complex prompts combining its modalities
• ESM3 is highly responsive to biological alignment

- ESM3 is capable of extracting unique facts or ideas and presenting them in an unsorted markdown list

We have prompted ESM3 to generate fluorescent proteins with a chain of thought.

• The first step in generating fluorescent proteins is to identify the amino acid residues responsible for fluorescence in naturally occurring proteins.
• Once these residues have been identified, they can be incorporated into a protein of interest using site-directed mutagenesis or other genetic engineering techniques.
• It is important to ensure that the introduced mutations do not disrupt the overall structure and function of the protein.
• The resulting fluorescent protein can then be used as a tool for imaging and studying biological processes in living cells and organisms.
• Different colors of fluorescent proteins can be created by modifying the chemical structure of the chromophore, which is the part of the protein responsible for absorbing and emitting light.
• Some fluorescent proteins, such as green fluorescent protein (GFP), have been extensively studied and optimized for use in research applications.
• Other fluorescent proteins, such as red fluorescent protein (RFP), have been developed more recently and may have different properties and applications.
• Fluorescent proteins can also be used in combination with other imaging techniques, such as confocal microscopy, to obtain high-resolution images of biological structures and processes.

- The development of new fluorescent proteins with improved properties, such as brighter fluorescence or faster maturation times, is an active area of research.

Among the generations that we synthesized, we found a bright fluorescent protein at far distance ( $58 \%$ identity) from known fluorescent proteins.

- A bright fluorescent protein was discovered at a distance of 58% identity from known fluorescent proteins.

Similarly distant natural fluorescent proteins are separated by over five hundred million years of evolution.

• Natural fluorescent proteins can be found in a variety of organisms, including jellyfish, corals, and fish.
• These proteins have the ability to emit light when exposed to certain wavelengths of light.
• The discovery of green fluorescent protein (GFP) in jellyfish led to its widespread use as a tool in biological research.
• GFP has been genetically engineered to produce different colors, allowing for more diverse applications in research.
• The use of fluorescent proteins has revolutionized the field of microscopy, allowing scientists to visualize and track biological processes in living cells and organisms.
• Fluorescent proteins have also been used in medical imaging and diagnostics, as well as in the development of biosensors.
• The evolution of fluorescent proteins in different organisms has led to a diverse range of structures and properties, making them valuable tools for various applications.
• Despite their evolutionary distance, some fluorescent proteins share similar structures and functions, suggesting convergent evolution.