Chapter 4 - The impact of artificial intelligence methods on drug design

Abstract

Artificial intelligence (AI) methods, and in particular machine learning (ML), deep learning (DL), Bayesian nets (BNs), and probabilistic reasoning are offering new tools for computer-assisted drug design (CADD). AI methods are accompanying, extending, or even changing existent practices in generating and searching chemical libraries in de novo drug design, in checking for properties, in optimizing drug candidates. After considering in short the computer practices in the drug development process, open challenges are individuated. The AI methods considered include the learning methods loosely inspired by the brain structure, such as the neural networks (NN) and the derived deep learning, the applications of mathematical logic to express hierarchical knowledge, and the integration of symbols and probability to reason on data. How those AI methods are working in various CADD and clinical studies are finally presented through the analysis of some recent literature. Warnings about the limitations of the AI algorithms are also reported.