Jocelyn Kaiser of sciencemag.org writes:
For height, DNA is largely destiny. Studies of identical and fraternal twins suggest up to 80% of variation in height is genetic. But the genes responsible have largely eluded researchers. Now, by amassing genome data for 4 million people—the largest such study ever—geneticists have accounted for a major share of this “missing heritability,” at least for people of European ancestry. In this group, they’ve identified nearly 10,000 DNA markers that appear to fully explain the influence of common genetic variants over height…
…By 2018, Visscher’s team and other members of a global consortium called GIANT had pooled DNA data for 700,000 people and found 3300 common markers that explained 25% of the variation in height. Now, by looking across DNA from 201 GWA studies with 4.1 million participants, GIANT has brought the total to roughly 9900 common markers, accounting for 40% of the variation. Other markers located nearby and likely inherited together account for another 10% of height variability.
That’s still short of the 80% predicted by twin studies. But last year, Visscher’s group drew on whole-genome sequencing data of a smaller number of people to demonstrate that rare variants—those carried by fewer than one in 100 people—should explain another 30% of height’s variation. (The result was released in a March 2019 preprint that the team is revising.)
Some geneticists say they aren’t surprised that heritability gaps can be filled once enough people had their DNA scanned. “It was expected,” says Aravinda Chakravarti of New York University. The problem remains that few of the height-linked DNA markers have been tied to specific genes that clearly alter the trait. “It’s mostly all still ‘missing’ in a biological sense,” says David Goldstein of Columbia University.
This is exciting news. Even though they still can’t say what genetic markers cause height, they will soon be able to predict it with incredible accuracy from just your DNA (at least for modern Westerners). If genomic markers explain 80% of the variation, the correlation between height and DNA would be 0.89! (the square root of 80%). In practical terms that means you could guess someone’s height just from DNA alone and be within 2.6 inches 95% of the time (at least for whites living in the West and eventually for all major populations).
Can we expect similar results for IQ? In the Minnesota study of twins reared apart, the square root of IQ’s heritability was about 92% as large as height’s. So if polygenic scores will one day correlate 0.89 with height, then perhaps a correlation of 0.89(0.92) = 0.82 can be expected for IQ (assuming a mental measure like IQ is analogous to a physical one like height).
This is about as high as the very best IQ tests correlate with each other, and implies they will be able to guess someone’s IQ based on DNA alone, and 95% of the time, those guesses will fall within 17 points of the correct IQ. In theory the precision could be increased if they predicted one’s composite IQ score on multiple high quality tests administered across the life span (thus cancelling out measurement error)
Of course none of this proves DNA causes most of the IQ variation but there is enormous practical value is predicting IQ, regardless of causation, even if said predictions are largely limited to specific nations and generations. Though if enough truly causal variants can be found, they will predict one’s cognitive rank in any society.