BTN News: Did you know that artificial intelligence (AI) can now make new proteins that never existed before? This amazing step in biotechnology was announced by EvolutionaryScale, a company started by former Meta researchers. They made an AI model that created a fluorescent protein with a sequence only 58% similar to natural ones.
Fluorescent Proteins That Glow in Unique Colors and Aid Biological Research
Fluorescent proteins take in light and give it out at a longer wavelength, making them glow in different colors. These proteins are very important in biological research because they can be added to specific molecules to track and study them. According to Live Science, the new protein made by EvolutionaryScale glows in a special shade of green, making it very useful for scientists.
AI Model and the Method Used to Create This New Protein
The team at EvolutionaryScale used an AI language model called ESM3 to make the new protein. They trained ESM3 with data from 2.78 billion proteins, giving information about their sequences, structures, and functions. They hid some of this information and asked ESM3 to guess the missing parts.
This method led the AI to create 96 possible proteins that could glow. The researchers chose one that was least like natural fluorescent proteins. ESM3 then made new versions of this protein, making it brighter, and this resulted in a new green fluorescent protein called “esmGPF”.
Potential Benefits for Medicine and Chemical Development
The study published on bioRxiv suggests this new protein could help make new drugs and chemicals to break down plastics. EvolutionaryScale points out that the AI made these new proteins in seconds, which would take nature 500 million years to evolve.
The ESM3 model does more than guess existing proteins. With data from 771 billion unique sequences, structures, and functions, ESM3 can create completely new proteins with specific jobs. Lux Capital, a sponsor of the study, called this a “ChatGPT moment for synthetic biology”.
Challenges in AI-Driven Protein Creation and Future Possibilities
Though AI in protein design has huge potential, there are challenges. The accuracy of AI predictions for new proteins needs to be checked. Luckily, AI also speeds up finding protein structures, avoiding old methods that use X-rays.
In conclusion, EvolutionaryScale’s work shows how AI can change biotechnology. As AI gets better, it will open new doors in science and medicine, giving us tools and solutions we never thought possible.
