Finally, Some Explanation on the Genetic Basis of Intelligence

RCEd Commentary

It’s a commonplace observation that intelligence is, at least in part, inherited from one’s parents.

Much of the research on this subject in past decades has focused on attempts to get a rough estimate of just how heritable it is, and estimates have typically been in the range of 30 to 70 percent (with the expected caveat that intelligence is really a product of both genes and environment, and you have to account for how they interact, and this stuff is really complicated, etc., etc.)  Meanwhile, finding specific genes associated with intelligence has largely been unsuccessful, even a decade after the Human Genome Project. But a recent study using a new method has proved successful where others have failed.

This work is the ultimate needle-in-a-haystack challenge. There are a million candidate single-nucleotide-polymorphisms (SNPs) -- identifiable genetic variations among people. So where to start, in trying to find which SNPs are associated with variations in intelligence?

One strategy would be to try to investigate all of them; just crunch the numbers, and see which SNPs are associated with better or worse scores on an intelligence test. (For simplicity, I’ll stick with “score on an intelligence test” as synonymous with intelligence.) But if you evaluate a million associations, some of them will be high, purely by chance. There are two ways to protect against that possibility. Researchers can increase the number of subjects in the analysis, which gives you more statistical power to separate the real effects from the spurious. But huge databases that include genetic information and reliable measures of intelligence are hard to come by.

A second way to protect against the errors resulting from multiple comparisons is to make fewer comparisons. Researchers can select a subset of the SNPs for evaluation, analyzing only those that, on the basis of other work, could plausibly have some role in intelligence (e.g., they were in genes that coded for neurotransmitters receptors).

That’s the way other groups have identified candidate genetic contributions to intelligence. But attempts to replicate these findings have flopped (Chabris et al., 2012). One out of 32 findings replicated. The researchers suggested that the sample sizes in the original studies were too small, which led false positives.

A new study (Rietveld et al, 2014) sought to combine the best of both methods: they used a large sample of subjects to pare down the number of candidate SNPs to be investigated.

First, researchers did an agnostic, genome-wide sweep in a large sample of 106,73 subjects, looking for associations of SNPs with intelligence -- only they didn’t use a measure of intelligence. As noted above, it’s hard to find really large samples of subjects that include both genetic information and a measure of intelligence. Instead, researchers used a proxy -- namely, how far people got in school. This first pass reduced the candidate SNPs from several hundred thousand to 927, which was then further trimmed to 69 SNPs via other statistical criteria meant to eliminate correlations among the SNPs themselves.

Researchers then used a second, independent sample of 24,189 subjects for whom they had measures of intelligence. Of the 69 SNPs identified by the education analysis, 14 showed clear relationship with intelligence in the second analysis; three in particular showed a very robust relationship. Twelve of the 14 are close to genes thought to be involved in the nervous system.

There are two important findings from this research. First, the method is new, and it’s a clever way around a difficult problem. We don’t have the measure we care about (intelligence) in enough people; so the authors did a first-pass with the measure they did have to narrow the spots on the genome that they should look.

Second, the association of intelligence with each SNP variant, while significant, was very small -- roughly 0.3 IQ points. That’s consistent with the idea that many, many genes are associated with intelligence, each making a modest contribution. That may be true because intelligence really is rooted in the activity of a huge number of genes, or it could be that any trait as abstract and removed from the actual work of genetic activity would have footings in a similarly large number of genes. We’ll have to wait for clarification, but it’s nice -- indeed, it’s almost a relief -- to have a start on describing the genetic basis of intelligence.

References:

Chabris, C. F., Hebert, B. M., Benjamin, D. J., Beauchamp, J., Cesarini, D., van der Loos, M., ... & Laibson, D. (2012). Most reported genetic associations with general intelligence are probably false positives. Psychological science, 0956797611435528.

Rietveld, C. A. et al (2014) Common genetic variants associated with cognitive performance identified using the proxy-phenotype method. Proceedings of the National Academy of Sciences,    doi: 10.1073/pnas.1404623111