Figure 3

Investigating batch effects. (A) A plot of the hierarchical clustering of 4 different tissues (CD19, Whole Blood, Brain and Lung) from 8 different studies. A modified Euclidean distance was used so that only those probes with a 0-0.1 beta difference contributed (left panel), 0.1-0.2 beta difference (middle panel) or a 0.2-0.3 beta difference (right panel) contributed. (B) Breast cancer DMPs were called using an F-Test with corrected p-value < 0.01 for tumour samples against normal for a single dataset contained in Marmal-aid (GSE29290). The beta value difference for these DMPs was then plotted for another dataset of breast tumors against normal taken from TCGA. The red lines represent a threshold of an absolute beta value difference >0.1 and >0.2.