Polygenic Scores are the New Black?

When I was criticizing genetic linkage studies in the early 2000's,  genome-wide association studies (GWAS) were all the rage.  These appeared to FINALLY be falling out of favor, due to their propensity to generate false positives.  Good riddance...  Unfortunately, like a hydra, They return with a vengeance, and with a second head:  Genome-wide Polygenic Scores.  The idea here is that one can take  thousands of genetic markers from a GWAS and find common matches for individuals sharing a particular trait.  I can see where this idea is going to appeal to a younger set of scientist brought up in the age of artificial intelligence, but I believe that we are again looking at a clever way to suggest that traits have genetic linkages, when in fact this type of analysis, at least on my first glance, appears to do nothing of the kind.

Again, I am just getting back into putting out critiques of genetic linkage studies of mental illness, personality traits and intelligence, so I might have more to say about this as I learn more, but I have some immediate concerns about these studies.  The first is that none of these genetic markers are shown to have any direct effect on a trait.  The second is that there is really no evidence for the idea that there are multiple polymorphisms that work to contribute to a trait.  It is an assumption based on our inability to find one, or a small number, of genes for really any trait related to mental illness, intelligence or personality.  Not one.
So let me make my point through a hypothetical example:
Suppose you have partially heritable disorder I'll call Disorder X.  You do a study finding a high Genome-wide polygenic score for a particular group of genetic polymorphisms.  Voila! Disorder X has a genetic component and now we have specific genes that contribute to Disease X.  We have no idea what mechanism these individual polymorphisms use in creating Disorder X (or not creating Disorder X in "normals").  We have no evidence or understanding of how this large group of polymorphisms might interact with other genes and each other but, the logic goes, now we can try to study these polymorphisms and learn how they contribute to Disorder X...  No, not really.  These genes could be completely unrelated to the disorder.
Let's suppose Disorder X is more common in Europeans.  We can't rule out the possibility that all the genetic polymorphisms we have identified are simply more common in Europeans and therefore flag people of European descent (much as Ancestry.com and the like, do).  All we have really done is find an expensive genetic test to determine that a person is of European descent and thus, more likely to have Disorder X, for still unknown reasons.  So, in a follow-up study, we only include Europeans.  And Voila!, we find a more specific subset of genetic polymorphisms giving us an even higher Polygenic Score.  Are these polymorphisms definitively related to Disorder X?  Again, no.  Let's say that Disorder X is more common in Scandinavians than the rest of Europeans.  What these polymorphisms might really be doing is indirectly identifying Scandinavian heritage, and so on...
Suppose also, there is another subset of, say, North African descent, for whom this disorder is also present, but they have a completely different subset of polymorphisms related to their Polygenic Score.  Researchers might conjecture that they suffer from Disorder X due to an entirely different  genetic cascade mechanism.  That's possible, I suppose, but more likely we are simply identifying genetic polymorphisms for North Africans who also have a higher incident of Disorder X.
Thus, we potentially have not found a single gene related to the actual disorder.  That doesn't mean that this type of analysis might not have some utility in determining mathematical risk for certain disorders, probably similar to the way Amazon.com figures out you might like an ice cream maker.  However, it does not really tell us whether any of these genes are related to the disorder, and certainly gives us no understanding of what the mechanism would be. 
Now, it doesn't really have to be some obvious delineation like culture background.  We can also have a relatively random group in which there is an aggregation of genetic polymorphisms related to a disorder or trait.  This might seem more promising.  Some might assume that if such a study is replicated once and even a small association is found, then we have something more than just random data.  This was common with genome wide association studies as well.  Every once in awhile, one of the genetic loci was noted to be associated with a disorder or trait in a follow-up study.  Since none of these "replications" have held up, we know that they were false positives.  What likely has happened is that, with the large number of studies of this nature being done in the early 2000's (believe me, there were a lot!), the law of averages makes it likely you get a false positive "replication."  So when I see someone claim that even a small association that is replicated is proof of a genetic link for a disorder or trait, I would suggest they consider the possibility that they happened to be the one who drew the Ace of Diamonds...