Virtual Screening Algorithms in Drug Discovery: A Review Focused on Machine and Deep Learning Methods

Drug discovery and repositioning are important processes for the pharmaceutical industry. These processes demand a high investment in resources and are time-consuming. Several strategies have been used to address this problem, including computer-aided drug design (CADD). Among CADD approaches, it is essential to highlight virtual screening (VS), an in silico approach based on computer simulation that can select organic molecules toward the therapeutic targets of interest. The techniques applied by VS are based on the structure of ligands (LBVS), receptors (SBVS), or fragments (FBVS). Regardless of the type of VS to be applied, they can be divided into categories depending on the used algorithms: similarity-based, quantitative, machine learning, meta-heuristics, and other algorithms. Each category has its objectives, advantages, and disadvantages. This review presents an overview of the algorithms used in VS, describing them and showing their use in drug design and their contribution to the drug development process.