Image analysis methods in high-content screening for phenotypic drug discovery

摘要

High content screening (HCS) of microscopic images is a very active field in computational cell biology and a powerful technique to reveal how chemical, genetic, and environmental perturbations affect cellular state. HCS has been effectively used to study organelle morphology, drug discovery, signaling pathways, sub-cellular protein localization, and functional genomics. Hence, high content screening is a type of phenotypic screen conducted in cells. The increased throughput characteristic of HCS experiments is due to automation of sample handling and microscopy; the development of robotic controlled stage positioning, fluorescence filters, camera acquisition and auto-focusing. High-content screening microscopy experiments generally require some steps: sample preparation, image acquisition, image analysis, image data management and image analysis. Each processing stage poses a number of computational challenges. The success of any high-content screening-imaging experiment relies on thoughtful assay design and appropriate image analysis approaches. We have created a number of methods for evaluating cell toxicity to determine nuclear area and cell number, and plasma membrane permeability. Methods developed for the characterization of bone marrow fractions obtained by density gradient centrifugation, comprising determining cell size, toxicity, assessment of the Hoechst dynamics absorption by living bone marrow cells. To investigate the influence of various factors on cell migration and cell-cell interaction we developed high-content analysis wound healing assay leading to increased assay precision and accuracy. Methods for the intracellular reactive oxygen species dynamic have evaluated for drug-screening procedure for photosensitizing agents used in photodynamic therapy. HCS experiments can contain tens of thousands of images including millions of cells and researchers must utilize machine-learning algorithms to translate morphological features into meaningful biological information. Machine learning is widely used in image-based screening to classify cell morphologies. The principal objective of the screening is to determine whether an experimental perturbation leads to a cellular phenotype. The most commonly used machine-learning method, classification, is based on the definition of phenotypes by representative examples. Thus, before a screen can be conducted, for negative controls as well as for expected classes of phenotypes. If representative examples for phenotypes cannot be obtained, supervised machine learning is not applicable and unsupervised methods need to be used instead. Also among these analyses are machine-learning methods that encompass data-driven models for deep learning. By further improving the usability of software interfaces, machine learning could eventually facilitate assay development and increase processing throughput, accuracy and objectivity.