Scoring diverse cellular morphologies in image-based screens with iterative feedback and machine learning

Note: This is a review of the published article listed below. All information, quotes, figures, methods, and findings mentioned in this review are from that article, and are the property of its authors and/or the publication in which the article originally appeared.

Original Research Paper

Scoring diverse cellular morphologies in image-based screens with iterative feedback and machine learning

Thouis R. Jones, Anne E. Carpenter, Michael R. Lamprecht, Jason Moffat, Serena J. Silver, Jennifer K. Grenier, Adam B. Castoreno, Ulrike S. Eggert, David E. Root, Polina Golland and David M. Sabatini

PNAS February 10, 2009 vol. 106 no. 6 1826-1831
http://www.pnas.org/content/106/6/1826.full

Review

Many cell biology pathways have been identified using mutants that have visible phenotypes. These phenotypes have traditionally been measured by subjective, visual and manual scoring methods that are laborious and subject to user errors. This study, from a group at the Broad Institute in Cambridge MA, describes a new method that lets you score cells with diverse morphologies using novel software approaches in cellular imaging technology. The group to identified multiple and complex morphological changes in high-throughput image-based screening assays. The approach took advantage of being able to undertake large-scale experiments in cells transfected with RNAi libraries, and to classify cells accurately based on subtle morphological changes. The group describes how this approach can be used for large-scale chemical or genomic cellular image-based screening projects, even when the cell phenotypes to search for are rare or when there are no positive controls available for comparison.

• View Larger

  Fig. 1. Scoring cell morphologies via cytological profiling, iterative feedback and machine learning. (A) Images of cell populations for each treatment condition
  (RNAi or chemical) are processed with cell-image analysis software like CellProfiler. This processs identifies and measures individual cells, in order to generate a cytological profile, containing a collection of measurements of features of each cell, represented schematically here as a bar code. (B) The software presents researchers with individual cells for classification, sampled randomly from the screen-wide population. After a few dozen cells are classified, the researcher can begin the iterative machine learning phase. This is whenthe computer generates a tentative rule based on the classified cells, and presents the researcher with cells classified according to that rule. In general, larger training sets produce more accurate rules, and using too small a training set can result in the computer training to an overly narrow definition of the phenotype (Fig. S10). Generating a large training set without iterative feedback can be difficult when the phenotype is rare or no positive control samples are available. These are the cases where the iterative nature of our approach is most useful. The optimal initial training set size depends on the complexity of the phenotype and the scarcity of positive cells in the experiment. After the researcher corrects errors and retrains for several rounds, the rule becomes more accurate. (C) When the rule is sufficiently accurate, it is used to classify all cells in the experiment, and calculate the number of positive and negative cells in each sample.