And they should also get some known-innocent control samples. (Of course, properly blinded.)