Profluent, a biotech company based in California, has utilized an AI model similar to ChatGPT to design novel antimicrobial proteins that have demonstrated the ability to kill bacteria in laboratory experiments. These proteins, published in Nature Biotechnology, were the first set of designs produced by Profluent’s AI platform, ProGen. ProGen is a large language model that uses deep learning AI and a vast amount of text data to analyze and generate language, specifically in the context of proteins. By harnessing the power of AI and language models, Profluent aims to develop new proteins that can potentially cure diseases. The researchers at Profluent created an AI model based on the language of proteins, which enabled them to generate new protein sequences with desired shapes and features. ProGen was trained on 280 million protein sequences and was guided to focus on specific protein properties. The team selected sequences related to a group of antimicrobial proteins called lysozymes and synthesized 100 artificial sequences, of which 66 showed chemical reactions similar to the positive control. Two of the novel antimicrobial proteins successfully killed E. coli bacteria. X-ray imaging revealed that these proteins, despite having significantly different amino acid sequences from known natural proteins, folded into shapes similar to their natural counterparts. Although the antimicrobial proteins are not ready for clinical use, this study highlights the potential of using language AI models to precisely design proteins for various biological, medical, and environmental applications.
| Signal | Change | 10y horizon | Driving force |
|---|---|---|---|
| ChatGPT-like AI creates new bacteria-killing proteins | Creation of novel antimicrobial proteins | More effective and targeted antibiotics | AI and language models for protein design |
| AI model similar to ChatGPT designed new proteins | AI-driven protein design | Precise de novo protein design | Advancements in AI and protein engineering |
| ProGen trained on protein language | Language-based protein design | Improved understanding of protein structure | Enhancing protein engineering capabilities |
| Novel antimicrobial proteins generated | Discovery of new antimicrobial proteins | Potential for new antibiotics | Addressing antibiotic resistance |
| Artificial sequences synthesized and tested | Identification of effective antimicrobial proteins | Development of new treatments | Advancements in laboratory synthesis and testing |
| Proteins folded into shapes similar to natural counterparts | Retention of natural protein folding | Enhanced functionality and stability | Conserving protein structure and function |
| Language AI models used for precise protein design | Application of AI in biology and medicine | AI-driven advancements in biotechnology | Solving problems in healthcare and the environment |