Figure 3

The 200 most variable CpG sites from 46 samples were clustered using Recursively Partitioned Mixture Model (RPMM) for Beta and Gaussian Mixtures from R package, a model-based clustering algorithm that returns a hierarchy of classes, similar to hierarchical clustering and finite mixture models. (A) RPMM clustering divided 46 samples into 3 clusters (1, 2, and 3), containing 12, 15, and 19 samples, respectively. (B) More than half of the BPA-high samples were clustered within cluster 1, while BPA-low and –intermediate samples were more evenly distributed across all 3 clusters.