Salesforce project builds proteins with generative AI

Why is it important: Generative AI technology may be more than just a threat to the livelihoods of countless artists and writers. According to Salesforce, a perfectly trained machine learning algorithm could help create artificial proteins that are beneficial to health care or the environment.

Salesforce Research is working on an innovative bio-application for artificial intelligence algorithms. ProGen is an artificial intelligence model designed to create synthetic proteins. It was trained on hundreds of millions of protein sequences in text form, resulting in artificial proteins that were just as effective as natural ones at removing waste.

ProGen was introduced in this month’s issue of Nature Biotech, where the researchers also described how the first known 3D structure of an artificial protein was designed entirely by an AI system. ProGen is a language model that can “generate protein sequences with predictable function across large protein families,” just as ChatGPT can combine different pieces of text to produce “grammatically and semantically correct natural language sentences on a variety of topics.”

According to the researchers, the new AI model was trained on 280 million protein sequences from more than 19,000 families and “padded” with check marks containing protein properties. By having enough homologous samples from different protein families, ProGen can be further tuned to “improve controlled protein generation performance.”

In other words, ProGen’s generative AI gives researchers ability to design highly adapted proteins “with desired properties” with a controlled tool. ProGen “learned” the rules of protein synthesis by looking up a database of protein sequences and was able to generate a set of proteins that the researchers later tested for their actual antibacterial properties in the lab.

According to Salesforce, test results showed that 73 percent of the artificial proteins created by ProGen were “functional” compared to 59 percent of natural proteins.

ProGen shows how scientists can develop a “preemptive approach” to protein design, according to Salesforce. The company hopes that in the future this new approach will help accelerate the development of “drugs for diseases” and enzymes for industrial or environmental applications. Plastic-eating proteins are another potential game-changing application.

Salesforce said it is already using the ProGen generative model to identify “potential treatments” for neurological and autoimmune diseases such as rheumatoid arthritis and multiple sclerosis.