Tuesday, March 03, 2020

Pleiotropy: Myths and Reality

The conventional view on pleiotropy is captured by this excerpt from H. Judson's The Eighth Day of Creation (PDF), p.609:
Genes act in concert. In the innumerable interactions of the developmental process, any single gene may affect several characters. The technical term is pleiotropy. And any character will be influenced by many genes. The term is polygeny. ...

[Thomas Hunt] Morgan's precepts establish a crucial and absolute distinction. While only a single allele, or very few, are involved in any given one of a high proportion of genetically related diseases, general heritable qualities of body or mind, our longevity, our intelligence, our particular talents, are the product of many genes, in exquisite balance among themselves and with the environment. This is why positive eugenics, breeding people for the enhancement of general qualities, cannot work, and why the extreme of positive eugenics, direct inter­vention in the germ line to improve such qualities, is a forbidden experiment: almost certainly you would upset the exquisite balance and engender a new human who would be seriously defective. All genes work in concert.
You will note that this description of pleiotropy lacks quantitative precision. In the time of Morgan: genes were entirely theoretical constructs, the importance and nature of DNA still a mystery, and the idea that there might be billions of distinct genetic variants totally beyond comprehension (perhaps to everyone but Fisher). The concept of pleiotropy was formulated before the notion of high dimensional spaces of variation became familiar.

Leaving aside ethical issues related to genetic enhancement, one can ask the practical scientific question: Is it possible?

In our recent paper (see post Live Long and Prosper: Genetic Architecture of Complex Traits and Disease Risk Predictors), we looked at the extent to which SNPs used in polygenic predictors of risk are correlated across pairs of disease conditions. We found relatively low correlations, as depicted in the table below from the paper.

In the conclusions we wrote:
III. The DNA regions used in disease risk predictors so far constructed seem to be largely disjoint (with a few interesting exceptions), suggesting that individual genetic disease risks are largely uncorrelated.

Observation III has interesting implications for pleiotropy [63–65]. We found that genetic risks are largely uncorrelated for different conditions. This suggests that there can exist individuals with, e.g., low risk simultaneously in each of multiple conditions, for essentially any combination of conditions. There is no trade-off required between different disease risks ... One could speculate that a lucky individual with exceptionally low risk across multiple conditions might have an unusually long life expectancy.
We can formulate this concretely (operationally?) as follows:

1. Regions of DNA correlated to different disease risks are largely disjoint.

2. It is plausible that causal genetic variants lie in these regions. For example, the predictor SNPs themselves could be causal, or they could tag (be highly correlated in state with) nearby causal variants.

3. Hypothetically, one could edit these causal variants independently, making the beneficiary simultaneously low risk for many conditions. The number of standard deviations of effect size in the polygenic score for each disease that can be modified independently (i.e., without affecting other disease risks or traits) is large and can be directly estimated from our results.

As the figure below (source) makes clear, a few SD change (e.g., ~5 SD, from 99th percentile to 1st percentile) in polygenic score for a given disease risk can lead to a 10x or possibly 100x decrease in absolute probability in having the condition. Our results suggest that the amount of variance available for engineering is much greater than this.

No comments:

Blog Archive