Phenotypic Drug Discovery
Chapter 6: Artificial Intelligence as an Enabler for Phenotypic Drug Discovery
Published:09 Dec 2020
Special Collection: 2020 ebook collectionSeries: Drug Discovery
N. Renaud and Y. Wang, in Phenotypic Drug Discovery, ed. B. Isherwood and A. Augustin, The Royal Society of Chemistry, 2020, ch. 6, pp. 104-117.
Download citation file:
Phenotypic drug discovery focuses on screening for an effect in cells (the phenotype) while being agnostic to the mechanism and target of the compound. Here we describe the use of computational methods, notably artificial intelligence, to accelerate each step of phenotypic drug discovery. Starting with assay development, machine learning can be used to prioritize good chemical probes to help the project team design and validate a robust screen. During hit discovery and triage, an iterative screening and machine-learning strategy can enable screening in complex cell models that would otherwise not be amenable. During target deconvolution and lead optimization computational models capitalize on decades of science to prioritize target hypotheses using data from orthogonal platforms both within our organization and outside. Finally, we end with an outlook and overview of emerging methods. Critical to these steps are not only the computational methods but also well-organized, curated data that capture the history of experiments. The acceleration is enabled by first making our data machine-learnable in order to apply machine learning.