Main

In this era of high-throughput methods, quantification of cellular changes caused by external perturbations is generally limited to population-level transcriptional and proteomic profiling techniques. However, cells are much more complex than such methods imply. In a recent issue of Science, Steven Altschuler and colleagues at Harvard describe a method of high-throughput cell profiling that enables a far wider range of phenotypic responses in cells to be quantified and analyzed. Rather than grinding up cells to get population averages, this multidimensional approach allows the investigator to measure changes in individual cells.

This breakthrough was the result of a broad collaboration that combined immunocytochemistry, microscopy, image processing and custom data analysis techniques to create an automated microscopy platform that can quantify changes in tens of millions of individual cells in one study. Cells are labeled with fluorescent antibodies to specific proteins and image processing identifies the nucleus, cytoplasm and other regions. An investigator then defines a set of descriptors to track changes in protein expression and cell morphology. Through the use of some clever data processing, graphical representations of the response of each descriptor to external perturbation are constructed. These color-coded profiles, capable of summarizing almost a billion separate data points, describe the cellular response to any perturbation.

In this initial report the authors examined the individual effects of 13 dilutions of 100 drugs on 93 descriptors in one cell type. It turns out that it is possible to group drugs with different structures but common targets by simply examining the color-coded profiles generated. According to coauthor Wu, “It was surprising to many people that we could identify the broad mechanisms of drug action with one cell type.”

Despite the potential of cell profiling for drug testing and discovery, this was not the impetus of the work. Altschuler says, “We are actually interested in how cells organize and this was a very pretty detour on the way to building network models that incorporate space and time.” They hope to next pick a single important pathway and use descriptors that will extensively mark it, allowing them to deeply probe the network and show how it is organized on a cellular level.