AlphaFold for a medicinal chemist: tool or toy?

Abstract

The development of novel small drug molecules is a complex and important cross-disciplinary task. In the early stages of development, chemoinformatics and bioinformatics methods are routinely used to reduce the cost of finding a lead compound. Among the tools of medicinal chemistry, docking and molecular dynamics occupy a special place. These methods are used to predict the possible mechanism of binding of a potential ligand to a protein target. However, in order to perform a docking study, it is necessary to know the spatial structure of the protein under investigation. Although databases of crystallographic structures are available, the three-dimensional representations of many protein molecules have not been reported. There is therefore a need to model such three-dimensional conformations. Several computer algorithms have been published to solve this problem. AlphaFold is considered by the scientific community to be the most effective approach to predicting the three-dimensional structure of proteins. However, the scope of its application in medicinal chemistry, especially for virtual screening, remains unclear. This review describes methods for predicting the three-dimensional structure of a protein and provides representative examples of the use of AlphaFold for the design and rational selection of potential ligands. Special attention is given to publications presenting the results of experimental validation of the approach. On the basis of performed analysis, the main problems in the field and possible ways to solve them are formulated.<br> The bibliography includes 154 references.