Building Natural Language Interface for Human Protein Atlas Data in 18 Months (axonagentic.ai)

Jonathan Agoot has announced the completion of an innovative AI-powered search engine designed to help scientists find commercial research use-only (RUO) antibodies by employing natural language queries. The project, which began in April 2024, leveraged the Human Protein Atlas (HPA) data, a respected resource for antibody validation. As researchers typically struggle with cumbersome manufacturer databases, Agoot's system aims to streamline the antibody discovery process, transitioning from complicated interfaces to a more intuitive, conversational style of interaction. The significance of this development lies in its potential to improve accessibility and accuracy in scientific research. By implementing a multi-agent architecture that includes specialized AI roles, Agoot's system enhances the precision of search results using advanced models like OpenAI's gpt-5. These models, with their large context windows and innovative Model Context Protocol (MCP), enable better integration with external data sources. The final prototype also features a verification-first approach, increasing the reliability of results and demonstrating how an individual without formal scientific training can contribute effectively to high-stakes scientific domains.