Assay2Mol is a large-scale language model-based workflow that aims to accelerate early-stage drug discovery by leveraging the vast existing data set of biochemical screening assays. It searches for assays that engage targets similar to existing targets and uses the retrieved assay screening data to generate candidate molecules through contextual learning. By leveraging information such as biological mechanisms and experimental screening protocols in unstructured text, Assay2Mol outperforms existing machine learning approaches that simply use the target protein structure, accelerating the generation of highly synthetically viable molecules.
