READRetro Web: A User-Friendly Platform for Predicting Plant Natural Product Biosynthesis.

Abstract

Natural products (NPs), a fundamental class of bioactive molecules with broad applicability, are valuable sources in pharmaceutical research and drug discovery. Despite their significance, the large-scale production of NPs is often limited by its availability and scalability, requiring alternative approaches such as metabolic engineering or biosynthesis. To identify ideal pathways for the mass production of NPs via, deep learning-based retrosynthesis models have been recently developed. Such models accelerate NP discovery, however, these tools are often not easy-to-use for researchers with limited computational background, because they require complex environment configurations, command-line interfaces, and substantial computational resources. Here we introduce READRetro Web, a user-friendly web platform that integrates the READRetro ML model for retrosynthesis prediction. Based on modern web technologies our web platform provides a fast and responsive user experience. READRetro Web bridges the gap between advanced ML-driven retrosynthesis and practical research workflows, making retrosynthesis prediction accessible to a broader range of researchers. Our platform demonstrates high predictive accuracy and computational efficiency, offering well-organized results to facilitate NP retrosynthetic pathway design. READRetro Web is freely accessible via: https://readretro.net.