Our lab has developed CellClassifier to enable cell biologists to apply supervised machine learning in their experiments. The main features are:
- An intuitive and easy-to-use graphical user interface developed in MatLab to upload and visualize microscopy images. CellClassifier displays each original image interactively, so that each cell is recognized and made 'clickable'. Each cell can therefore be easily annotated, rare phenotypes within a population can be found, and it can be observed if single-cell phenotypes depend on the population context of that cell (Snijder B, Sacher R, Rämö P, Damm EM, Liberali P, Pelkmans L. "Population context determines cell-to-cell variability in endocytosis and virus infection." Nature. 2009 Sep 24;461(7263):520-3. pdf
- Three methods of supervised machine learning, namely Support Vector Machines (SVM), including multiclass SVMs, multilinear perceptrons and k-nearest neighbors. Developers can easily add their own methods.
- Possibility to display fractions of correctly and incorrectly classified cells during the course of training in order to evaluate the classification performance and its improvement through training iterations.
- Export of classification results at the single-cell level to MatLab-readable files and at the population level to standard spreadsheet programs.
Original publication: Rämö P, Sacher R, Snijder B, Begemann B, Pelkmans L. "CellClassifier: supervised learning of cellular phenotypes." Bioinformatics. 2009 Nov 15;25(22):3028-30. pdf
Screenshot of CellClassifier's user interface