Can we predict intelligence from genetic data?

We all know that intelligence has a genetic component. Twin studies have reported that identical twins, who share their entire genome, have higher concordance in IQ tests than fraternal twins who only share half of their genome. In fact, the genetic component of IQ is as strong as the genetic component of height: about 80% of the differences between people are attributed to genetic polymorphisms. However, twin studies only tell us about the magnitude of the genetic component relative to the environment. They do not tell us anything about which SNPs in the genome are responsible for shaping IQ differences.

A few week ago, the New York Times reported an “Enormous Success” in finding the SNPs associated with intelligence. The article states that 52 new genes were identified as part of a massive analysis of nearly 80,000 genomes. But the article also mentions that these 52 genes have a “minuscule” effect on IQ. We were excited to read about this study, and we saw that it was discussed in several Facebook groups of genetic genealogists, where people asked if it would be possible to predict someone’s IQ from DNA. So we decided to use DNA.Land to see what is really meant by minuscule.

To this end, we took the SNPs from the IQ study that the New York Times covered and put them in the DNA.Land trait prediction engine. Mathematically, the process is fairly simple. Genetic studies of complex traits, called Genome-wide Association Studies (of GWAS), are just classic simple linear models: Y = a*x. “Y” is a trait measurement (IQ); “a” is the effect of the SNPs, meaning by how many points your IQ changes for each variation; and x is the number of copies of the associated allele in the genome of interest. Here is an example to clarify how it works:

Say that the study found that the “A” allele in SNP #rs123 increases IQ by 1 point and that your genotype is “AA” at position #rs123. Since you have two copies of “A” in this position, then the IQ prediction is Y = 1 * 2 = 2 IQ points higher due to these SNPs.

To summarize, the “a” values come from the New York Times-covered study and the “x” values come from each individual genome. Last, we present the IQ prediction of each participant as a deviation: positive scores mean a higher chance for increased IQ, and negative scores mean a lower chance for increased IQ.

Disclaimer: as usual, we give participants the autonomy to decide whether they want to see their prediction. We do not show it automatically to you and your prediction is not shared with anyone.

What we found was quite striking. Most individuals received a score of around zero. This means that for most people, these SNPs did not influence their intelligence in any particular direction! In fact, no individual receives a difference of more than 5 IQ points! For those of you who got a negative score — this doesn’t mean you are not intelligent. Case in point: Dr. Yaniv Erlich, accomplished geneticist, Columbia University Professor, and creator of DNA.Land, has a negative score:

As of 2017, the predictive ability of IQ studies is surprisingly limited even when breakthroughs are claimed. One of the best ways to improve the power of these studies is to increase the sample size — that means collecting genomes from more individuals, the primary purpose of DNA.Land!