Try clustering

[wbazant]

• within a dataset: could use SOMs to check whether the replicates are any good

• globally: clust would probably do something interesting for all datasets together:
  http://clust.baselabujamous.com/Home.aspx