Prediction of Molecular Drug Properties

In drug chemistry research, elucidating the relationship between molecular structure and activity has always been of great significance. Early studies generally analyzed this issue by constructing quantitative structure–activity relationship models. In recent years, new methods based on artificial intelligence, especially deep learning models, have achieved remarkable results in many fields and have attracted increasing attention in the fields of chemistry and pharmacy (Klebe and Abraham in J Med Chem 36:70–80, 1993). Molecular property prediction is one of the key tasks in the computer-aided drug discovery process and plays a crucial role in many downstream applications, such as drug screening and drug design. Its main purpose is to predict the physical and chemical properties of molecules based on internal information such as atomic coordinates and atomic numbers, enabling the identification of compounds with desired properties from a large number of candidates and accelerating the speed of drug screening and design (Shen and Nicolaou in Drug Discov Today Technol 32:29–36, 2019).