Imaging Platform

Conventional flow cytometry is a powerful technique for measuring cell phenotype and function, but it relies on fluorescent stains, or labels, to identify particular cell subpopulations. At times, those labels can be incompatible with live cells, or unavailable to researchers. Now, researchers from the Broad Institute’s Imaging Platform, Swansea University’s College of Engineering, and fellow international collaborators have found a new way to detect these cellular subpopulations, by applying machine learning to the hidden information in images of unlabeled cells generated from image flow cytometry. Their method is described in Nature Communications and a Swansea University press release, and their open-source workflow is available online.

Images of cells, when suitably stained, contain a vast amount of biological information. These images are now generated by the hundreds of thousands in experiments that seek to determine the functions of genes and to identify useful chemicals for research and for potential therapeutics.