MOESM10 of “Gap hunting” to characterize clustered probe signals in Illumina methylation array data

Additional file 10: Table S4. Alternatives to gap hunting do not correctly identify polymorphism-affected clusters. For the probes shown in Fig. 7 and the gap signal in Fig. 1, we explored other ways of identifying clusters. Specifically we examined a Gaussian mixture model clustering algorithm that selects an optimal number of clusters based on the Bayesian information criterion, and the dip test for unimodality (alternative hypothesis is that distribution is multi-modal). We recorded the number of clusters selected by the mixture model algorithm and the dip test p value.