6
$\begingroup$

Assuming that intelligence has a genetic component,

• do we know which genes contribute to it?

and, if so,

• can we predict intelligence from genomic analysis?

$\endgroup$
  • $\begingroup$ I'm voting to close this question as off-topic because it concerns a construct — Genetic IQ — that has no meaning, certainly no biological meaning. $\endgroup$ – David Dec 4 '18 at 15:19
  • 1
    $\begingroup$ @David Questions about predicting phenotypes from genotypes are perfectly valid and on-topic questions about biology. $\endgroup$ – Eff Dec 5 '18 at 8:33
  • $\begingroup$ @Eff — Remind me what "phenotype" the poster was asking about. $\endgroup$ – David Dec 5 '18 at 8:51
  • 2
    $\begingroup$ @David IQ. If you read this question--with respect for the fact that laymen will have trouble formulating their questions in very precise terms--OP is asking whether we can predict a person's IQ from their genotype. This is a problem actual biologists (specifically, geneticists) work on, so clearly it is a problem that is relevant to biologists. $\endgroup$ – Eff Dec 5 '18 at 9:14
  • 1
    $\begingroup$ @Eff — This is not a site for laymen, but for "biology researchers, academics, and students". My whole point was that the "phenotype" in the question was something referred to as "Genetic IQ" which is meaningless. Further IQ is not a phenotype and by its nature cannot be measured at birth. Intelligence is different. You appear to think that the question means "Assuming that intelligence has a genetic component, do we know which genes contribute to it, and can we predict intelligence from genomic analysis?", so I have edited it, but it is a duplicate. $\endgroup$ – David Dec 5 '18 at 10:42
8
$\begingroup$

A few words on genomic prediction

No complex trait is 100% heritable, hence no prediction based entirely on DNA would ever be perfect. With that said, predictive genomics is progressing at a quite amazing rate right now. So while predictions can be nowhere near perfect, it is getting possible to make DNA predictions that correlate substantially with observed values.

The genes that correlate with phenotypic differences are found using genome-wide association studies (GWAS). The total effects are then aggregated up to what is called a polygenic score.

A success story -- Height

Height is highly heritable, meaning that much of the phenotypic variance (although not 100%) is due to genetic differences. Height is also highly polygenic -- it's a trait influenced by many genes, each of small effect.

Many variants that correlate with height have now been found. Polygenic scores of height correlate higher than ~0.60 with observed height. DNA predictions are generally within a few centimeters of true height (Lello, Avery, Tellier, Vazquez, de los Campos, Hsu, 2018). According to an interview with Stephen Hsu, one of the authors of this paper and co-founder of the company Genomic Prediction, they can correctly predict the height ordering of siblings within the same family 80-90% of the time (source). So based purely on DNA, can we say which of two siblings are going to be taller than the other? Not perfectly, but with decent reliably, yes.

This is about as good as predictions of complex traits can be based on current methods. There are a few principal reasons that correlations are not higher (even though it is quite impressive already in its own right). First, as I said, complex traits are not 100% heritable and therefore predictions are not expected to ever be perfect. Second, current GWAS are typically based on SNPs which are common gene variants. Rarer gene variants are expected to contribute to the variance and their effects are still needed to be uncovered. Third, only additive effects are taken into account, not gene-by-gene interactions.

Genetic IQ Prediction?

IQ, like height, is a heritable trait and is highly polygenic.

There are several approaches to finding gene variants correlated with IQ, however they can broadly be categorized into two. The most obvious approach is to simply give people intelligence tests and get their DNA. The problem with this is that it is difficult to get large samples with good intelligence tests. When this approach is used, usually a very short (say, 2 min) intelligence test is used. The second approach is to use a proxy phenotype. With this approach, the variable years of educational attainment has been used to good success, and has been shown to have a high genetic correlation with intelligence ($r_g \approx 0.7$).

Many variants that are associated with intelligence or educational attainment have been found, see e.g. (Lee et al, 2018; Savage et al, 2018). While many variants are known, IQ is not as well understood as height, and current SNP predictors correlate about ~0.3 with observed IQ (See e.g. Allegrini et al, 2018). This correlation is bound to increase in the coming few years.

Further Information

If you want a simple introduction to what can be read from the genome so far, there is a TED talk on this subject, TED2016: How to read a genome and build a human being. Although, major improvements have already been made in the years since 2016. For anyone that is more interested in the mathematical techniques used for prediction and the underlying theory, I recommend this talk by Stephen Hsu if you're further interested in genomic prediction of complex traits. I also recommend reading this review by Robert Plomin and Sophie von Stumm to get an easily understandable overview of the current state of knowledge on the subject.

References

Lello, Avery, Tellier, Vazquez, de los Campus, Hsu (2018). Accurate Genomic Prediction of Human Height. DOI: https://doi.org/10.1534/genetics.118.301267

Marty Nemko (2018). The Future of In-Vitro Fertilization and Gene Editing. Psychology Today Link.

Lee, ..., Cesarini (2018). Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals. DOI: https://doi.org/10.1038/s41588-018-0147-3

Savage, ..., Posthuma (2018). Genome-wide association meta-analysis in 269,867 individuals identifies new genetic and functional links to intelligence. DOI: https://doi.org/10.1038/s41588-018-0152-6

Allegrini, Selzam, Rimfeld, von Stumm, Pingault, Plomin (2018). Genomic prediction of cognitive traits in childhood and adolescence. DOI: https://doi.org/10.1101/418210

Sabatini (2016). How to read a genome and build a human being. TED2016 Link to Talk.

Hsu (2018). Genomic Prediction of Complex Traits. Youtube Link to Talk.

Plomin, von Stumm (2018). The new genetics of intelligence. DOI: https://doi.org/10.1038/nrg.2017.104

$\endgroup$
  • $\begingroup$ This answer is clearly better than mine! +1 $\endgroup$ – Remi.b May 30 '18 at 23:57
3
$\begingroup$

No, because the trait you describe does not exist

Your question betrays a common misunderstanding of how genetics and the environment interact in order to produce complex phenotypes. In fact, every biological trait is 100% genetic and 100% environmental. Don't believe me? Try teaching algebra to your cat, or see what height someone is after you've dropped them into the sun.

The only sense in which you have an "IQ you were born with" is the measurable IQ of a newborn child which I'm guessing this comes out as 0. Instead your IQ is a result of continuous interaction between genes and environment. You can say that at the end of process, across the population, 70% (say) of observed variation is explicable by genetic variation but this does not mean that you got 70 points of IQ from your genes and 30 points from the environment. The whole score is attributable to an interaction between the two. There is no "null" environment in which you would observe pure genetics, nor does it follow that increasing the range of environments encountered across your sample will increase the proportion of variation explicable by the environment. A gene that has no effect on IQ in one environment may have a marked effect in a different environment and you'd only discover this by varying the environment so both are encountered.

So, there is no trait "genetic IQ" to predict from someone's DNA, even if we had perfect knowledge of the link between genes and intelligence. Which we don't.

$\endgroup$
  • $\begingroup$ What I meant by "genetic IQ" was "What a person's IQ would be assuming optimal environmental conditions". $\endgroup$ – Thomas Jun 3 '18 at 22:24
  • $\begingroup$ @Thomas That's not a meaningful concept. $\endgroup$ – Jack Aidley Jun 4 '18 at 6:41
  • $\begingroup$ @Thomas To expand on that: there is no way of knowing what the optimal environmental conditions would be, or that there actually exists a peak environment, and if it exists it will almost certainly be completely different for different backgrounds. $\endgroup$ – Jack Aidley Jun 4 '18 at 8:17
2
$\begingroup$

IQ typically has a relatively high heritability (see this post). Please have a look at this post to understand the concept of heritability.

In order to know what loci (QTL as the trait of interest is quantitative) are associated with a specific trait of interest, one must perform a GWAS The studies listed in the first pots linked above do not involve GWAS but only some kind of parent-offspring regression or twin studies to estimate heritability. From these studies, one cannot infer anything about a person's IQ from its genome.

GWAS studies on IQ do exist though (reviewed in Pfiffer 2015). It would be possible from one to make predictions and compute confidence intervals as well of someone's intelligence based on their genome, yes. I doubt anyone ever wrote such algorithm though. Without having such algorithm in front of one's eyes, it is impossible to tell, whether the confidence intervals are going to be very wide or very narrow though.

$\endgroup$
  • 1
    $\begingroup$ Why do you say you would need Bayesian statistics to get predictions? $\endgroup$ – kmm May 28 '18 at 14:34
  • $\begingroup$ To get a confidence interval at least. Or maybe one could use resampling but then you need to re compute the GWAS plenty of time which should take forever. I did not think this through though but it does not sound like easy statistics to me. The details of how to compute this confidence interval is not really on topic with the question but if you have some thoughts, I'll be happy to hear them for curiosity. $\endgroup$ – Remi.b May 28 '18 at 14:50
  • $\begingroup$ I was mostly just curious. I didn't think that Bayesian GWAS was done that much, because of the slowness in MCMC sampling vs. fast matrix algebra for "frequentist" GWAS. The traditional way will yield intervals as well, but probably too wide. $\endgroup$ – kmm May 28 '18 at 15:01
  • $\begingroup$ The Pfiffer paper compares IQ and genetic traits between distinct countries and ethnicities rather than IQ for single people as outlined in question above. $\endgroup$ – tsttst May 28 '18 at 19:18
  • $\begingroup$ In order to make predictions for a single individual, you need to know the association between loci and the trait of interest through a GWAS. None of the cited paper offer an algorithm to make such prediction from a single person but one who would want to develop such algorithms will need access to this type of GWAS. $\endgroup$ – Remi.b May 28 '18 at 19:22
0
$\begingroup$

IQ is pretty much like muscle strength, in the sense that you may have the right combination of genes that would give you extra strong muscles but the environmental factors do play an essential role as well.

Nutrition, training etc will have a profound effect on the development of your muscles. You cannot predict the effect a priori.

For IQ, it is the same thing. On one extreme you may have genetically-linked developmental diseases, physically impairing the normal brain functions, i/e/ negatively affects your IQ. In most of the cases however, the effects of genetics on IQ is shadowed by the actual "brain training" you get during your life, by nutrition, and by other environmental factors.

So, yes, IQ is influenced by the underlying genetics.

Yes, in some extreme cases you can predict a very strong detrimental effect of some genes/mutations on the IQ.

For most of the cases, it's very hard to predict and genetics alone would not be enough to do so.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.