This is very interesting and almost too good to be true, though it is what you would expect if researchers and patients report observations of the same clinical reality, i.e. gene expression measurements corroborating patient reported symptoms. However the big question is, can it be replicated? This was a 10 patient study and it looks like the patients were very carefully selected on a clinical basis, which is the kind of method required if we are to get past the deresolving effect of mixing different subtypes in the same studies. What concerns me is that replication attempts might fail to select their patients according to the same criteria.