Chatbots like ChatGPT have made it easy for users to access complex information through simple questions. The same principle is now being applied to one of biology’s most complex challenges: protein design. Traditionally, creating custom proteins required deep technical expertise and reliance on naturally evolved templates. But scientists are now building AI models that can generate novel proteins from plain English prompts, much like asking ChatGPT for a summary or essay.

Enter Pinal, a new AI designed to act as a conversational protein engineer. Developed by an international team of researchers, Pinal allows scientists to describe the desired type, function, or structure of a protein in natural language. In response, the AI generates candidate proteins that can be tested in living cells. In one demonstration, Pinal successfully designed enzymes that broke down alcohol, some even functioning at high temperatures.

Unlike older protein design tools, which require structural inputs or advanced coding skills, Pinal simplifies the process through a two-step model. It first converts the prompt into structural elements (like loops or folds), then uses a protein-language model called SaProt to generate full amino acid sequences. Compared to other models like ESM3, Pinal showed greater accuracy and novelty, meaning it was better at producing useful, never-before-seen proteins.

Pinal was trained using a massive dataset of 1.7 billion protein-text pairs, enabling it to interpret prompts with biological context. It operates with 16 billion parameters, drawing parallels to large language models in terms of scale and learning architecture.

This new class of AI systems aims to democratize synthetic biology, making protein design as accessible as using a search engine or writing assistant. While the technology still faces challenges—such as generating non-functional or redundant sequences—it marks a major step toward designing proteins untethered from evolution, guided instead by human intent and natural language.

Other startups, like 310 AI with its MP4 model, are also entering the space, showing that text-to-protein design is a fast-growing frontier with implications for medicine, sustainability, and beyond.

By Impact Lab