Fig. 3: Applying the VoPo framework across clustering algorithms.
From: VoPo leverages cellular heterogeneity for predictive modeling of single-cell data
Implementing repeated metaclustering within the VoPo framework improved accuracy and increased robustness across datasets and clustering algorithms. Regardless of whether PhenoGraph, FlowSOM, or k-means was used in the clustering and metaclustering components of VoPo, repeated metaclustering with VoPo (pink boxplots) lead to higher average classification accuracy (AUC) with less variability over the baseline case where VoPo was not applied (black boxplots). The boxplots show median values, interquartile range, whiskers of 1.5 times interquartile range, and all individual points.
