Profluent, an AI-powered protein design company based in California, has unveiled a groundbreaking AI model that can generate CRISPR-like proteins not found in nature. CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) is a complex containing proteins that enable scientists to perform precise gene editing in organisms. Profluent’s AI model leverages an in-house large language model (LLM) trained on extensive sequence and biological context data. This allows the model to generate millions of diverse CRISPR-like proteins that do not occur naturally. These synthetic gene editors have the potential to revolutionize medical research, opening up new avenues for treating previously incurable diseases.
Ali Madani, founder and CEO of Profluent, announced the AI model in a series of posts on X (formerly known as Twitter). The company has also published a blog post detailing the initiative and released a pre-print version of its research paper on bioRxiv. Profluent’s AI model represents a significant advancement in the field of gene editing, where scientists have traditionally relied on naturally occurring proteins like Cas9. The ability to generate synthetic gene editors through AI greatly expands the toolkit available to researchers and accelerates the pace of discovery.
In addition to announcing the AI model, Profluent has also launched OpenCRISPR-1, one of the AI-created gene editors. OpenCRISPR-1 is an open-source release licensed for ethical research and commercial uses. The company’s commitment to open science will foster collaboration and innovation within the scientific community. CRISPR technology has a wide range of applications, including the development of disease-resistant crops, the alteration of mosquito DNA to prevent disease transmission, and potential therapies for genetic diseases like sickle-cell anemia. The ability to generate CRISPR-like proteins through AI holds immense promise for pushing the boundaries of gene editing and unlocking the potential to cure previously untreatable diseases.