Large language model-integrated automatic designer for species-specific primers in food authentication

Species-specific PCR is widely used for food authentication and allergen detection, but primer design remains laborious and expertise-dependent. Existing design tools require users to select the target gene, obtain its sequence, and identify species-specific regions through separate alignment. Here, we developed a large language model-integrated web platform that automates the entire design process. Using only target and non-target species names, the platform retrieves marker gene sequences from the National Center for Biotechnology Information, aligns them, identifies species-specific regions, and designs and screens candidate primer pairs. Subsequently, an integrated large language model uses these computed results to generate a natural-language report of the candidates’ properties and recommended assay conditions. The platform was applied to three oilseeds: sesame ( Sesamum indicum ), perilla ( Perilla frutescens ), and sunflower ( Helianthus annuus ). Real-time PCR confirmed four species-specific primer pairs per species, with amplification efficiencies of 90.6–104.4% and limits of detection of 0.00001 ng (sesame and sunflower) and 0.0001 ng (perilla). In 17 commercial processed foods, the assays identified all labeled species and detected undeclared sesame and perilla, likely from cross-contact during production. The platform thus enables users without bioinformatics expertise to rapidly develop species-specific assays for food authentication.