Fig. 2

Analysis of simulated data. a Analysis with previously published mRNA detection approach and localization features. In the t-SNE plot, each dot is a cell colored with its localization pattern. b Confusion matrix for k-means clustering with 8 classes for data in a. Rows show identified classes, columns the known localization pattern. Numbers in each square indicate percentage of cells for a localization pattern (column) that was assigned to a given class (row). Off-diagonal elements are mis-classifications. c, d t-SNE projection and confusion matrix as in a and b but determined with our 3D spot detection with GMM and the new localization features. e Analysis of pooled simulated data (different pattern strength and expression levels). Analysis and t-SNE projection as in c. f Confusion matrix for data in e with k-means clustering performed on t-SNE projected data in six dimensions