Skip to main content
Fig. 4 | BMC Bioinformatics

Fig. 4

From: A comparison of reference-based algorithms for correcting cell-type heterogeneity in Epigenome-Wide Association Studies

Fig. 4

Comparison of EpiDISH to other reference-based algorithms. a Barplots of RMSE and R2 values for the estimated cellular proportions for the four different algorithms as applied to the reconstructed blood mixtures of Koestler et al. All estimates were obtained over the 12 reconstructed samples. Right panels show the average values over all blood cell subtypes. (b) As (a), but now for the in-silico generated mixtures of 5 blood cell subtypes from Zilbauer et al. All RMSE and R2 estimates were obtained over 100 randomly generated in-silico mixtures, averaged over the 5 cell-types, and this experiment was repeated a total of 25 times (“Runs”). P-values comparing EpiDISH to each of the other 3 methods are from a paired one-tailed Wilcoxon rank sum test. (c) As (b), but now for in-silico generated mixtures of 3 epithelial/non-epithelial cell types: fetal lung fibroblast, liver and B-cells. All RMSE and R2 estimates were obtained over 100 randomly generated in-silico mixtures, averaged over the 3 cell-types, and this experiment was repeated a total of 25 times (“Runs”). P-values comparing EpiDISH to each of the other 3 methods are from a paired one-tailed Wilcoxon rank sum test

Back to article page