Fig. 3From: Aristotle: stratified causal discovery for omics dataOverview of Aristotle. The data are shown by rectangular blocks and the methods are shown by ovals. P, R, G, and Z are defined in “Problem definition” section \(G_1\) to \(G_D\) are the subsets of G corresponding to feature groups produced in the Feature Grouping phase (see “Group Features Based on Background Knowledge” section). \(W_1\) to \(W_D\) are the feature weights produced by biclustering for each feature group. \(C_1\) to \(C_D\) are the sets of the top weighted features in each of the feature groups that the feature selection method (indicated by FS in the diagram) chooses from groups \(G_1\) to \(G_D\) given the weights \(W_1\) to \(W_D\), respectively.\(P_1\) to \(P_K\) are the sample strata produced by biclustering. Quasi-experimental design (indicated by QED in the diagram) evaluates the candidate features in the set \(C=\cup _{d=1}^{D} C_d\) for causality with respect to their corresponding sample strata and confounders Z. The outputs of QED are tuples that indicate the causal pairs of feature and stratumBack to article page