Shen, 2007 attempt to reproduce results with 27k platform

The data was normalized two ways: batch removed and batch/age/gender removed. Number of patients: 166

No CpGs that were within the MINT marker regions were found.

Found 70 CpGs corresponding to the genes in Shen, 2007, selected the most variable CpGs per gene. Heatmap for both datasets (distance: euclidean, clustering: average; note: data weren't scaled)

Used bootstrapping to verify cluster stability (10000 bootstraps) (left - only batch removed; right - batch/age/gender removed):

Conclusions: these markers don't separate the data