The genetic inheritance patterns of complex traits can be summarized with just just two numbers per complex trait.
Genome-wide association studies have revealed that the genetic architectures of complex traits vary widely, including in terms of the numbers, effect sizes, and allele frequencies of significant hits. However, at present we lack a principled way of understanding the similarities and differences among traits. Here, we describe a probabilistic model that combines mutation, drift, and stabilizing selection at individual sites with a genome-scale model of phenotypic variation. In this model, the architecture of a trait arises from the distribution of selection coefficients of mutations and from two scaling parameters. We fit this model for 95 diverse, highly polygenic quantitative traits from the UK Biobank. Notably, we infer similar distributions of selection coefficients across all these traits. This shared distribution implies that differences in architectures of highly polygenic traits arise mainly from the two scaling parameters: the mutational target size and heritability per site, which vary by orders of magnitude across traits. When these two scale factors are accounted for, the architectures of all 95 traits are nearly identical.
Yuval B. Simons, et al., "Simple scaling laws control the genetic architectures of human complex traits" bioRxiv 2022.10.04.509926 (October 7, 2022).
Ok... I read it. It sure feels like there is something here. I notice that Graham Coop's lab is mentioned in the credits and the Stanford folk are well known for mathy work. Those are plusses. I need more of the picture colored in to grasp the implications. Sigh...
ReplyDelete@Guy
ReplyDeleteThe nice thing about this is that it is very useful in terms of practical applications even though it is basically just a data fit rather than a result derived from first principles reasoning.
If you look at inheritance patterns for a complex trait, you can estimate these two parameters and can be almost certain that you fully understand how it is inherited as a practical matter, which is something that is actionable, without having to worry about myriad theoretically possible alternatives. It shows that you can almost always ignore the more complicated genetic inheritance models when a complex trait is involved (simple genetic traits might be another matter).
The first parameter, crudely speaking, is an estimate of how many genes are involved in the trait (at least relative to other complex traits as opposed to an actual fixed number of genes).
The second parameter, crudely speaking, is an estimate of the average amount of influence each gene involved has on the complex trait.
At one extreme are single gene dominant traits in so called Mendelian conditions where the complex trait is entirely the product of a single gene. A lot of inherited mental retardation genes (I'm using old school non-PC language for clarity), fit this model.
At the other extreme are complex traits with huge numbers of relevant genes all of which have only a slight incremental effect. The classic example of traits that fits this model are IQ and education.
In between are complex traits with moderate numbers of relevant genes which have intermediate level effects. Classic examples of traits that fit this model are eye color, skin color, height, and lactose intolerance.
But, the two parameters aren't functionally related to each other. Instead, they are independent degrees of freedom.
This is because when you add up the number of relevant genes and their average effect sizes, this doesn't have to explain 100% of the variation in the complex trait. Instead, complex traits can be influenced by factors other than genes as well.
For example, a lot of variation in height in the populations of Japan and Korea are functions not just of genetic propensity to be tall or short measured through the proxy of the heights of a person's close relatives. A lot of the variation in height in Japan and Korea is also strongly a function of a person's date of birth which is a good proxy for dramatically improving nutrition and health care in those countries on a very equitable basis in a very short period of time. These environmental factors unlock someone's full genetic potential height when it is present, but obviously, aren't themselves genetic.
ReplyDeleteWhile this result isn't obvious there is also good reason to think it is a reasonable result if you could do the analysis rigorously.
The "normal" Bell curve probability distribution also called a "Gaussian" probability distribution is the probability of an end result of the sum of an infinite number of yes or no coin toss type events with the same probability for each event. But, it turns out (and is possible to prove with surprising weak assumptions about the underlying events) that usually, the the probability that the sum of any large number of random events none of which is profoundly dominant will have a particular value is well approximated by a Gaussian distribution. This is sometimes known as the "central limit theorem." https://en.wikipedia.org/wiki/Normal_distribution
Essentially what this result shows empirically, since we don't understand the underlying mechanism of the biochemical effects of genes to prove it rigorously, is that something similar is true in the genetics of complex traits. Even though we don't think that complex traits are really due to something as simple as the sum of X genes which each have an identical effect Y on the complex trait, the way genetic inheritance works in complex traits is close enough to this oversimplified model that the deviations from it in the more accurate precise reality average out enough to make it a very good approximation over a very large share of the kind of genetic inheritance patterns that actually produce complex traits.
It is intuitively and heuristically reasonable to expect that something similar to the central limit theorem applies to the inheritance of complex genetic traits which usually mostly involve lots of genes with small effect sizes working together.