Fig. 5.

The proposed method could improve the deconvolution performance of BayesCCE in simulation studies. BayesCCE is the BayesCCE results without applying TOAST. BayesCCE + T is the BayesCCE algorithm with the proposed method TOAST. Results are based on 100 DNA methylation simulation datasets with four cell types. Panels from left to right demonstrate correlations of estimated versus true proportions (Corr with true prop), root mean squared error, and root means squared bias. 200 samples are simulated in each dataset. The p values are obtained from paired t test. Red fonts indicate significant test results. Boxplots are summarized from 100 Monte Carlo experiments