Is our future written in our genes? Recent developments in genetics give the impression of a rapidly increasing ability to predict our individual traits based on genomic information, and even to manipulate those traits through technologies such as CRISPR-based genome editing. For some physical traits, like eye color, and for genetically simple diseases, like cystic fibrosis or sickle-cell anemia, this impression is pretty accurate—we really can predict those things from a person’s genetic profile and we really could alter them in embryos with one or a couple of judicious edits.
But could we do the same for more complex traits, including psychological ones like intelligence? Again, recent stories suggest it is possible, at least in principle.
King’s College London geneticist Robert Plomin, in his new book, Blueprint, presents DNA as a “fortune-teller” that is “100 percent reliable” and that can “predict your future from birth.” He also argues that the “only systematic, stable and long-lasting source of who we are is DNA.” A U.S. company, Genomic Prediction, recently said it will offer embryo selection based on polygenic scores for intelligence. And the announcement from China that human babies had been born with CRISPR-edited genomes immediately prompted visions of its use for the creation of “designer babies.” with alterations of genes affecting all kinds of traits, including intelligence.
The prospect of either genetic selection or genome editing for intelligence or other traits has led to an outcry over the lack of consideration that has so far been given to the associated ethical concerns, and rightly so. But it is interesting to note that the backdrop to many of these discussions is the implicit assumption that, even if our ability to predict intelligence from our genome is currently imperfect, it is only a matter of time before it becomes much more accurate.
Indeed, researchers are finding more and more genetic variants associated with intelligence all the time, as sample sizes in these studies continue to increase. And we are used in more general terms to the application of machine learning in dredging these kinds of high-dimensional data for meaningful patterns that give almost godlike powers of classification and prediction in other areas. It seems reasonable to assume that the power of genetic information will continue to increase, even for complex traits like intelligence, which involves variations in thousands of genes.
But these discussions have overlooked a much more fundamental limit in our ability to predict or control our psychological traits. Most such traits are only partly heritable—that is, only a certain proportion of the variation we see in the trait across the population can be attributed to genetic differences between people. For intelligence, the heritability is about 50 percent, at least in developed nations and in samples with relatively uniform socioeconomic status. The rest of the variation is nongenetic.
The common inference is that if it’s not genes making us different from each other, it must be something in the environment. If this were true, and we could identify the causal environmental factors, then maybe we could control those too. And we do know of many environmental factors that do affect intelligence, such as maternal and infant health, nutrition and education.
But even in situations where the variation in these factors is very low, there is still substantial nongenetic variance in the trait that remains unexplained.
That third source of variation may simply be chance. Our psychological traits derive from differences in the physical structure and chemical makeup of our brains. The wiring of the brain is astonishingly complex and its almost miraculous self-assembly relies on a huge number of cellular and developmental processes, involving the actions of thousands of genes. It is variation in precisely those kinds of genes that has been implicated in intelligence.
Those genes encode a program of development but they do not encode the precise outcome. All they encode is a set of mindless rules governing the biochemical interactions of millions of protein molecules, determining which genes get turned on and off in each cell of a developing embryo. Complex sets of feedback and feed-forward interactions ensure that different organs develop in the right place, that all the different types of cells differentiate, and, in the brain, that all the nerve cells and brain regions get connected to each other in the right way.
However, all those processes are subject to noise or inherent randomness at a molecular level. The genes can set the rules, but the outcome will vary—sometimes substantially—from run to run of the program. This is especially true in the brain, due to the nonlinear, self-organizing nature of development, where small differences at one stage can have cascading consequences and be amplified across subsequent stages of development.
Even for genetically identical twins, whose brains are very structurally similar to each other, there is thus still substantial variation between them. This is reflected in differences in their psychological traits like intelligence or personality. The key point is that this variation is not due to environmental factors or anything outside the individual—it is intrinsic to the processes of development themselves. By the time we are born, our brains and minds are already unique—not just due to our genetics, but as the result of a never-to-be-repeated sequence of developmental events.
This places a firm limit—in principle, not just in practice—on the level of precision associated with genomic prediction of psychological traits. We can certainly use genetics to look at statistical effects across the population, but this will give at best very fuzzy predictors for individuals. No matter how good our understanding of the genetics of intelligence gets, we will never be able to predict intelligence of individuals with accuracy from genomic information.