Protein design tool accepts natural language requests

Researchers have created a tool that lets people design proteins by describing what they want in natural language rather than specifying technical details about amino acid positions and structures. Traditionally, protein design required sophisticated knowledge of protein chemistry and enzyme mechanisms, but this new approach uses databases of millions of protein sequences paired with function descriptions to translate plain language requests into actual protein designs. The tool works in two stages: first converting text descriptions into a structural blueprint showing spatial relationships between amino acids, then constructing the actual amino acid sequence. While the success rate appears low when tested, this reflects the filtering process that removes designs resembling known proteins, and a similar natural language approach has also been developed for designing DNA sequences.