Abstract
De novo drug design is a computational technique to develop novel chemical
compounds from scratch without prior knowledge. Traditionally, structural
bioinformatics approaches used either structure-based or ligand-based design; the
former uses the active site information of the protein, and the latter uses known active
binders. Modern methods based on artificial intelligence help design de novo drugs in
less time by using pre-trained models. One of the major bottlenecks of the de novo drug
design is the synthetic feasibility of the active compounds, which is addressed using
AI-based methods that help reduce the time and cost of analysis of those compounds.
Recent success stories from several companies show the strength of the AI-based de
novo drug design programs, and many advances can be expected shortly.