Fig. 1
From: CoINcIDE: A framework for discovery of patient subtypes across multiple datasets
CoINcIDE steps. The four main steps to CoINcIDE, a method to discover replicable patient subtypes by finding consensus across dataset-specific clusterings from multiple datasets. The first step of CoINcIDE is to derive these dataset-specific clusterings. The second step is to compute cluster-cluster similarities between each cluster-cluster pair, resulting in an adjacency matrix (clusters within the same dataset are not compared). The third step is to assign an edge between cluster pairs whose similarity passes set magnitude and significance thresholds, with the weight of the edge equaling the similarity magnitude. This creates a network. The fourth step is to then discover meta-clusters via network community detection methods. The resulting meta-clusters are the final subtypes