Agreed. To put it simply, if you model a complex system with another complex system, of which you have a similar level of understanding, then even if the model is very faithful, you gain very little. Unless the model is transferable, so you can predict, which AFAIK is not the case here.

I think the point is not that GWAS can do everything, but that many people 5-10 years ago thought that they had a fundamental flaw - where was all the heritability? It has turned out so far that the heritability is there and discoverable with big enough sample sizes. But of course, you are right that GWAS don't tell us much about the very large space between strands of DNA and social or behavioural outcomes. They are a tool like others. (Mendelian randomization is hardly a panacea - how credible are those exclusion restrictions, usually?)

I agree, holding up GWAS as some kind of lesson about scientific progress seems pretty silly. I don't think the 'sober experts' deserve shame and humiliation considering that PGSes have had minimal impact on human health.

Unless something has drastically changed in the past 5 odd years, Actual Geneticists (tm) happily use GWAS in many investigations of complex traits. They're often an early step in the overall pipeline in which they trawl for potential candidates for that target of interest, after which they go on to said more complex methods.

Why? You've already said so yourself. Those are expensive and tedious, and searching across the entire pool of human genetics with them is an exercise in futility.

The important part in your post is

>They're often an early step

I think we're in agreement here, I'm just arguing that "woah, look at all those correlations" isn't a breakthrough or 'genomic revolution' in any sense of the word as far as our understanding of human genetics is concerned.

Seemingly, most of the complex traits for which GWA is an effective methodology (a few SNPs of large effect size) have already been discovered. More and more these days, I’m seeing association studies that fail to yield any hits. Whether this is due to sample size, polygenicity, or some other model failure remains to be seen.

Can you post the link about the GDP/A. thaliana correlation? That sounds hilarious, and Google is failing me to find it myself.

Found it:

https://mobile.twitter.com/pastramimachine/status/1263937115...

Edit: if it doesn't work try this i guess? https://www.google.com/url?q=https://mobile.twitter.com/past...

Throws a Twitter error for me.

GWAS have been shown to be practically meaningless. How should this paragraph be interpreted, that transformers can make sense of DNA patterns?

Who cares how genes work? The important part is the real predictive power of the PGS.

'Predictive' is just a glorified word for 'correlating'. Staring at a bunch of correlations only gets you so far and doing only that certainly isn't science. If you have no idea how genes work you're going to be dismayed when all your fancy correlations don't work anymore, as people gathering genomic data from outside the European biobanks are starting to find out.

Nope. You're wrong. Where do you think your MRs or your sibling comparisons are coming from? You're not getting anywhere with weak instruments from a few significant hits. You're also wrong that the only thing of value is inferring "how genes work", and that is the sort of extremely blinkered mechanism-centric attitude which blinded people to GWASes working, because gosh, it would be awful if polygenicity was true, because how would we build any 'scientific theories' on this? (Cue Turkheimer.) And yes, those PGSes are useful for all sorts of things like selective sweeps, enrichments, and clinical instruments, because of incremental validity. (By the way, what does '15% heritability' refer to? I sure hope that, since you're claiming to be an expert here, you aren't confusing heritability with PGS power, like so many supposed human geneticists insist on doing...)

I didn't say GWASes were useless, just that it's absurd to consider them to be a 'revolution'. The actual revolution would be second- and third-generation sequencing which enabled GWASes and a bunch of much more useful things. GWAS is, in effect, just a bunch of correlations. It's just the very starting point to an actual scientific analysis, because 'you have to start somewhere'. If you don't go beyond and investigate, you're not doing science. Everyone in the genomics community agrees to this, and literally every paper that investigates the causes of genetic diseases goes in the introduction like 'GWASes sure look nice but we still have no idea how things work with them so in this paper I present a method to do...' I notice you failed to address many of the spurious correlations drawn by the GWAS bot or the A. thaliana vs. GDP prediction. That it doesn't raise any red flag to you doesn't speak well as to your ability to approach the field of genomics.

>You're also wrong that the only thing of value is inferring "how genes work"

Yes it is, that's literally what genetics is about. Otherwise you're back to making a bunch of correlations. If you want a deep understanding of disease, design effective drugs, or even do proper gene editing, you have to understand what genes do. It seems ridiculous to have to say it.

>By the way, what does '15% heritability' refer to?

It is what we in the less-rationalist circles refer to as a 'joke'.