Really nice work. Short thread follows https://twitter.com/snaqvi1990/status/1253088329014206464

This paper shows that even for the most proximal molecular traits - expression levels of a number of proteins - where a few high-effect variants exist, the majority of the genetic variation is STILL distributed across the genome

I think this reflects a simple truth: cells are integrated complex systems. Messing with anything anywhere will affect that local element a lot, but have ripple effects across the genetic and biochemical networks of the whole cell

Since distant elements vastly outnumber local ones for any individual "trait", the number of indirect genetic effects will vastly outnumber the direct ones.

This is highly inconvenient, if we want to use genetics to understand "the biology" of some trait or process. In model organism genetics, this can be done through secondary screens and hypothesis-testing experiments.

That's obviously not possible in humans, where all we can are observational analyses. That doesn't mean analyses like GWAS are not useful - it's just very hard to take the next step by studying humans alone...

These ideas are discussed further here: Lessons for human genetics from genetic screens in model organisms http://www.wiringthebrain.com/2018/02/lessons-for-human-genetics-from-genetic.html?spref=tw