Phenotypic profiling of cells in drug discovery

Abstract

In the context of the current crisis in the pharmaceutical industry, when the profit from new drugs no longer justifies the huge investments that are made in research and development, finding reliable disease models is one of the main tasks. The review considers phenotypic drug development and an important element of development - a pathological phenotype (model) suitable for high-throughput screening. In particular, one of the ways to solve the problem of physiologically relevant cell phenotype is phenotypic profiling aimed at recording a wide range (up to one and a half thousand) of phenotypic characteristics of a model cell, some or even all of which may not have previously had any confirmed significance for a particular disease or potential treatment, and identifying - using AI - hidden, non-obvious, practically useful profiles/“fingerprints” of cells associated with diseases and various impacts, and also a special case of phenotypic profiling - morphological profiling based on the analysis of images of cells stained with fluorescent dyes, using machine learning methods (AI, deep learning of neural networks, reinforcement learning), similar to face recognition and personal identification methods. Morphological profiling data are widely used both for drug development and for fundamental research.