Samsung Advanced Institute of Technology

In synthesis planning, identifying and optimizing chemical reactions is important for the successful design of synthetic pathways to target substances. Chemical reaction databases assist chemists in gaining insights into this process. Traditionally, searching for relevant records from a reaction database has relied on the manual formulation of queries by chemists based on their preferences, which is however challenging without explicit knowledge of what they are searching for. In this study, we propose an intelligent chemical reaction search system that automatically customizes the search results based on user feedback. Contrastive representation learning is used to train a representation model that embeds each record in the reaction database into a numeric vector. When a query is submitted, the search algorithm retrieves the nearest records based on their distance in the vector representation. Users are allowed to provide binary ratings to individual retrieved results, inspired by how recommender systems work. Human-in-the-loop learning is integrated to reflect user feedback in the search results. We demonstrate through experimental investigations that the proposed system effectively improves chemical reaction searches toward better aligning with user preferences.