When I feel my most cynical perhaps, but no I don't think that is the bottom line of the paper. "For example, disproportionate megalencephaly at age 3 may provide early clues to parents about children who may require higher levels of support, or individuals with autism distinct forms of co-occurring anxiety may benefit from autism-specific anxiety interventions." The naive hope we all had was that we'd collect a bunch of data and identify all the important clusters unsupervised. Reality is that identification meaningful subtypes will involve examining one or several at a time (per grant), even if that cluster might only represent <1-5% of all autistics.