Skip to content

NATURAL SELECTION ON EDUCATIONAL ATTAINMENT DID NOT PRODUCE ALLELIC DIFFERENTIATION BETWEEN POPULATIONS

Only some days after the article of Wong (2018), and in line with this article, a study of Guo (2018) found that natural selection on „educational attainment” did not produce higher allelic differentiation than explained by drift, between AFR, EUR and EAS populations. Unfortunately, Guo (2018) did not test the allelic differentiation for IQ. Wong (2018) found a much higher SNP differentiation for IQ than for EDU in 1000 GENOMES.

Between 10 complex traits, Guo (2018) found SNP differentiation only on height, waist-to-hip ratio and schizophrenia.

In a recent article, Curtis (2018) sustained that polygenic risk score for schizophrenia could be more strongly associated with ancestry than with schizophrenia. Differences of POLY_SCZ between Africans and Europeans are ten fold higher than differences of POLY_SCZ between European schizophrenics and European healthy controls. But Wong (2018) found the lowest allelic differentiation for SCZ.

 

REFERENCES

Curtis, D. et al (2018) Polygenic risk score for schizophrenia is more strongly associated with ancestry than
with schizophrenia. bioRxiv doi: http://dx.doi.org/10.1101/287136 .

Guo, J. et al (2018) Global genetic differentiation of complex traits shaped by natural selection in humans. Nature Communications. DOI: 10.1038/s41467-018-04191-y

Wong, E.S. et al (2018) Evidence for stabilizing selection at pleiotropic loci for human complex traits. bioRxiv doi: http://dx.doi.org/10.1101/126888 .

 

Reclame

SELECTION FOR EDUCATIONAL ATTAINMENT DOES NOT EQUATE WITH SELECTION FOR INTELLIGENCE

Recently, Emily Wong (2018) published a study on the allelic differentiation of 1000 GENOMES populations and subpopulations.

In Figure S4 (Supplementary material) we can find the magnitude of the genetic differentiation between populations for complex trait loci.

It is extremely  interesting that differentiation for educational attainment is significantly lower than differentiation for IQ, openness, consciousness.

It means all populations were roughly equal selected for the complex phenotype „educational attainment”, but some populations were more selected for high IQ, and other populations were more selected for high openness or consciousness or other complex traits related to „self-domestication”. I sustained on this blog all civilizations selected for domestication-increasing variants, but not for IQ-increasing variants.

It is interesting too Wong (2018) found also an intermediate genetic differentiation for head circumference, higher than for educational attainment, but lower (and closer) than for IQ. The decrease of the brain size during Holocene produced worldwide, and its significance is the increase of the human self-domestication. I expect the decrease of the brain size parallels the increase of the POLY_EDU, but the decrease of the POLY_IQ.

REFERENCE

Wong, E.S. et al (2018) Evidence for stabilizing selection at pleiotropic loci for human complex traits. bioRxiv doi: http://dx.doi.org/10.1101/126888 .

THE POSITIVE SELECTION AND THE INCREASE OF THE POLYGENIC SCORE DO NOT EQUATE WITH THE INCREASE OF THE COMPLEX TRAIT (IV)

A very recent study (Aris-Brosou, 2018) found mutational load of Europeans constantly increased since Mesolithic time. It means, even if a polygenic score for a complex trait increased during last 10,000 years, it is possible this complex trait decreased during this period, if the raise of the polygenic score did not (over)compensated the decrease produced by the raise of the mutational load.

The higher polygenic score on height of Mesolithic hunter gatherers and even of Neolithic farmers than Early Upper Paleolithic Europeans (Berg, 2017) despite their shorter stature than Upper Paleolithic hunter gatherers (Formicola, 1999) could be partially explained by this increase of the mutational load.

Also, Beiter (2017) found positive selection for increased total intracranial volume during last 2,000 years, despite the intracranial capacity significantly decreased during this period (Henneberg, 1988). This decrease could be explained by the increase of the mutational load.

Aris-Brosou (2018) sustains the increase of the mutational load in Europe is due of range expansions (migrations), in line with the observations of Peischl (2013). But, if we compare the results of Aris-Brosou (2018) with the results of Henn (2016), we can see today Sub-Saharan Africans have higher mutational load than Mesolithic Europeans. Hence, the mutational load of Sub-Saharan Africans increased too, at least during last 10,000 years, and this increase can not be explained by range expansion, but rather by the population growth, and possibly by other effects of civilization.

 

REFERENCES

Aris-Brosou, S. (2018) Evidence of a nonadaptive buildup of mutational load in human populations over the past 40,000 years. bioRxiv doi: http://dx.doi.org/10.1101/307058 .

Beiter, E.R. et al. (2017) Polygenic selection underlies evolution of human brain structure and behavioral traits. bioRxiv doi: http://dx.doi.org/10.1101/164707.

Berg, J.J. et al. (2017) Polygenic Adaptation has Impacted Multiple Anthropometric Traits. bioRxiv doi: http://dx.doi.org/10.1101/167551 .

Formicola, V. & Giannecchini, M. (1999) Evolutionary trends of stature in Upper Palaeolithic and Mesolithic Europe. J. Hum.
Evol., 36: 319–333.

Gazave, E., Chang, D., Clark, A.G. & Keinan, A. (2013). Population growth inflates the per-individual number of deleterious mutations and reduces their mean effect. Genetics 195: 969-978.

Henn, B. et al (2016) Distance from sub-Saharan Africa predicts mutational load in diverse human genomes. PNAS 113(4): E440-449 doi:  10.1073/pnas.1510805112

Henneberg, M. (1988). Decrease of human skull size in the Holocene. Human Biology 60: 395-405.

Peischl, S., Dupanloup, I., Kirkpatrick, M. & Excoffier, L. (2013). On the accumulation of deleterious mutations during range expansions. Molecular Ecology 22: 5972-5982.

SOME OBSERVATIONS ON THE EVOLUTION OF GENOTYPIC INTELLIGENCE IN HUMANS

FIRST SCENARIO

The earliest known stone tools were made 3.3 million years ago (110,000 generations ago) by Pliocene hominins living in highlands of East Africa. These people probably had an intracranial capacity slightly higher than a chimpanzee, and an IQ of at least 40 (like a chimpanzee). Today Africans living in the same region have an IQ of 70. White Americans living a century ago had an IQ of 70. During 100,000 generations, the evolution of humans produced an increase of 30 IQ points, equating the Lynn-Flynn effect for the last 100 years (three generations): today White Americans have an IQ of 100. But certainly some Lynn-Flynn effect operated on today Africans, and on White Americans living a century ago too, hence the difference between these populations and our extinct ancestors could be even lower than 30 IQ points. Although, this increase of 30 IQ points matches the 900 at 950 cm3 increase of the brain size during this 3.3 mya period: Lefebvre (2015) found an increase of 1 IQ point for every 31 cm3 of the brain volume in today humans (for a correlation of 0.23 between IQ and brain size).  

The GWAS on IQ and EDU found hundreds of alleles with small effect on IQ/EDU. But the effect of each of these alleles discovered by GWAS is not smaller than 0.2 IQ point. An increase of 150 alleles per individual (genome) equates with an increase of 30 IQ points. If this increase of 30 IQ points produced during 100,000 generations, it means an raise of 0.0015 IQ-increasing allele by generation (equating with an increase of one allele per genome every 667 generations).

Studies on correlation between IQ and fertility, but also the count of alleles related on educational attainment found a dysgenic fertility reflected by a decrease of 0.5 at 1 IQ point by generation since the demographic transition (a decrease of 4 at 8 IQ points during last eight generations). It means a decrease of 20 at 40 of the number of IQ-increasing alleles per individual, equating a decrease of 2.5 at 5 alleles by generation.

SECOND SCENARIO

Homo Sapiens emerged 300,000 years ago (10,000 generations ago), and he had an IQ of 40 (like a chimpanzee). White Americans living a century ago had an IQ of 70. It means an increase 150 alleles per genome, equating with a raise of 0.015 IQ-increasing allele per individual (genome) by generation during last 10,000 generations (an increase of one allele by 67 generations).

THIRD SCENARIO

First Eurasians, just after Out-of-Africa, (60,000 years ago; 2,000 generations ago) had an IQ of 50 (lower than today Pygmies), and White Americans living a century ago had an IQ of 70. It means a raise of 0.5 IQ-increasing allele per individual (genome) by generation, and an increase of IQ of 0.01 point by generation.

SOME OBSERVATIONS

There are scholars that sustain the most accelerated increase of the genotypic intelligence in human lineage produced especially during last 10,000 years, or even last 5,000 years (during past 150 at 300 generations). During last 100,000 generations the brain volume had a three fold increase. The human brain is not larger today than 10,000 generations ago. The brain size peaked 1,000 generations ago, in Cro-Magnons. Probably the genotypic IQ peaked in Cro-Magnons too. During past 300 generations brain size decreased at least 10%, and during last 600 generations this decrease was of 15% at 20%. Probably the genotypic IQ decreased during these periods too. The brain volume increase produced against very important energetic and obstetrical constraints. It was a costly process that could not produce without the important benefit of the raise of intelligence. The brain size decrease could be explained by energetic and obstetric advantages, if the selection on intelligence relaxed.

We can notice that the increase of genotypic IQ produced very slow, but the decrease of genotypic IQ (since the demographic transition) was very fast. We can think the mutation pressure (of the 70 de novo mutations by generation) is quite significant, and this do not allow a fast increase of the genotypic intelligence, despite a selection on intelligence. Even in the more ‘optimistic’ scenario (the third), the increase of only one IQ-allele by generation reflects that the highest endeavour of natural selection on intelligence is the purifying selection of deleterious de novo mutations. Although, I can not see why a high intelligence was a higher advantage for surviving and reproduction in Pre-industrial England than in Ice Age North-Asia. It is probably true that inside the upper class of Pre-industrial England there was selection on intelligence and there was also a downward social mobility. But I do not believe there was selection on intelligence inside lower class of Pre-Industrial England, hence those that dropped from upper class had no advantage in reproduction after they changed their social status.

I think the increase of genotypic IQ during the last 10,000 or 5,000 or 1,000 years is not the most parsimonious explanation for the advent of more complex civilizations. Behavioral and cultural changes, also the increase of population number and density, and the chance of some Lynn-Flynn effect (even a smaller effect of 5 IQ points increase) could explain the emergence of civilizations better than 500 generations of selection on intelligence.

THE GENOTYPIC INTELLIGENCE OF EUROPEANS DECREASED SINCE BRONZE AGE (IV)

Michael Woodley & Davide Piffer (2017) published a very interesting study, comparing polygenic scores on educational attainment of today and Bronze Age samples. Woodley & Piffer used a polygenic score on 130 common SNP resulted from the GWAS of Okbay (2016). They found a polygenic score of 3,298.5 : (3,298.5 + 3,997.5) = 45.21% for Bronze Age samples and 61666 : (61,666 + 69,114) = 47.15% for today Europeans. There is an increase of 2%, and this increase equates with 1,000 more EDU-increasing SNP per genome, if we assume POLY_EDU has 50,000 causal SNP. But POLY_EDU could have only 10,000 alleles too. I will use 50,000 SNP to maximize the strength of selection on EDU since Bronze Age.

The average effect of each SNP is an increase of 0.02 SD of educational attainment (Okbay, 2016, Extended Data Figure 2). But the GWAS of Okbay found only the SNP with the largest effect on EDU by the 50,000 SNP of POLY_EDU. The average effect of a SNP of POLY_EDU must be much lower than 0.01 SD, and also I expect a lower frequency increase of alleles with smaller effect on EDU than of SNP with larger effect. But I will estimate the strength of selection using an average effect of 0.01 SD per allele, and an average increase of 2% for all the 50,000 SNP of POLY_EDU.  But 1 SD of EDU equates with 3.8 IQ points (Kong, 2017), hence the mean effect of each SNP is 0.01 x 3.8 = 0.038 IQ points. The total raise of the IQ due of the 1,000 common SNP increase per genome since Bronze Age is 38 IQ points.

The average age of Bronze Age samples is 3,440 years, equating with 3,440 : 30 = 114.66 generations. I will assume the increase of 38 IQ points produced during only 100 generations, because the genotypic IQ of Europeans decreased during last 7 generations due of dysgenic fertility. The increase of genotypic IQ of Europeans could be 0.38 IQ points by generation, if mutational pressure was zero.

Woodley (2015) estimated the 70 de novo mutations of a newborn produce on average a 2.94 IQ points decrease, but Woodley & Fernandes (2016) re-estimated this generational decrease of the IQ by de novo mutations at only 0.15 points. I will try here another estimation.

A recent study of Lodato (2017) found the somatic mutations in neurons of brain linearly increase with age. At 20 years the number of mutations per neuron in prefrontal cortex is 1,000, and at 80 years there are 2,000 mutations per neuron. Another study (Kaufman & Horn, 1996), found an average decrease of fluid IQ with 25 points between ages of 25 and 80. If only 6 IQ points of this decrease is due of the supplementary 1,000 somatic mutations in neurons of older people, the average effect of a mutation is a 0.006 point decrease of IQ. The somatic mutations arose during embryogenesis and neurogenesis, hence each of these affects only a cluster of neurons. I expect each of the 70 de novo mutations of a newborn have a higher average deleterious effect than a somatic mutation, because each de novo mutation affects all the neurons of the brain, not only a cluster. But I will use the minimal estimation of 0.006 IQ points decrease by a de novo mutation. The 70 de novo mutations of a newborn will produce a generational 70 x 0.006 = 0.42 IQ points decrease. This value is too high to be compensate by the strength of selection found by Woodley & Piffer, of 0.38 IQ points increase per generation.

 

REFERENCES

Kaufman, A.S. & Horn, J.L. (1996) Age Changes on Tests of Fluid and Crystallized Ability for Women an Men on the Kaufman Adolescent and Adult Intelligence Test (KAIT) at Ages 17-94 Years. Archives of Clinical Neuropsychology 11(2): 97-121

Kong, A. et al. (2017) Selection against variants in the genome associated with educational attainment. PNAS 114(5): E727-E732. doi: 10.1073/pnas.1612113114.

Lodato, M.A. et al (2017) Aging and neurodegeneration are associated with increased mutations in single human neurons. bioRxiv doi: http://dx.doi.org/10.1101/221960 .

Okbay, A. et al. (2016). Genome-wide association study identifies 74 loci associated with educational attainment. Nature 533: 539-542.

Woodley, M.A. (2015) How fragile is our intellect? Estimating losses in general intelligence due to both selection and mutation accumulation. Personality and Individual Differences 75: 80–84

Woodley, M.A. & Fernandes, H.B.F. (2016) The secular decline in general intelligence from decreasing developmental stability: Theoretical and empirical considerations. Personality and Individual Differences 92: 194-199

Woodley, M.A. et al. (2017) Holocene selection for variants associated with cognitive ability: Comparing ancient and modern genomes. bioRxiv http://dx.doi.org/10.1101/109678

 

 

 

THE POSITIVE SELECTION AND THE INCREASE OF THE POLYGENIC SCORE DO NOT EQUATE WITH THE INCREASE OF THE COMPLEX TRAIT (III)

An extremely interesting study of Zhang & Huang (2017) demonstrated that only 29% of the human genome can freely accommodate mutations. It means on average 50 of the 70 de novo mutations (SNP) of a new-born are deleterious.

Another study (Hernandez, 2017) found that singletons alone contribute ~23% of all cis-heritability across genes, and 50.9% of cis-heritability is contributed by globally very rare variants (MAF<0.1%).

A recent study of Young (2017) found that MAF > 0.1% explain only 17% of educational attainment. Although, the heritability of educational attainment is estimated at 35-40%. It means very rare variants (MAF<0.1%) explain at least 50% of this heritability. Also Hill (2017) found 64% of the heritability of educational attainment and 58% of the heritability of intelligence are explained by rare variants (MAF<1%).

All these results demonstrate that the mutational pressure on complex traits and diseases is much higher than predicted by the neutral theory. If more than two third of de novo mutations are deleterious, it is necessary that polygenic score on common polymorphism increases to maintain the actual level of a complex trait. It means the increase of a polygenic score does not equate with the increase of the complex trait. If the increase of the polygenic score does not (over)compensate the accumulation of de novo mutations, the complex trait will even decrease, despite a positive selection on the trait and on the polygenic score.

Furthermore, not only de novo mutations and rare variants are deleterious for complex traits, but probably even many SNP that reach the frequency of common polymorphism are deleterious too. There are studies that demonstrated the total number of MAF > 1% (common variants) in the individual genome positively correlates with the risk of schizophrenia (He, 2017), Parkinson disease (Zhu, 2015) and type 2 diabetes (Lei, 2017). I expect the total number of MAF > 1% in the individual genome negatively correlates with complex traits too.

 

REFERENCES

He, P. et al (2017) Accumulation of minor alleles and risk prediction in schizophrenia. Sci Rep 7(1): 11661.

Hernandez, R.D. et al (2017) Singleton Variants Dominate the Genetic Architecture of Human Gene Expression. bioRxiv doi: http://dx.doi.org/10.1101/219238 .

Hill, D.W. et al. (2017) Genomic analysis of family data reveals additional genetic effects on intelligence and personality. bioRxiv http://dx.doi.org/10.1101/106203

Lei, X. & Huang, S. (2017) Enrichment of minor allele of SNPs and genetic prediction of type 2 diabetes risk in British population. PLoS One 12(11): e0187644.

Young, A.I. et al (2017) Estimating heritability without environmental bias. bioRxiv doi: http://dx.doi.org/10.1101/218883 .

Zhang, Y. & Huang, S. (2017) De novo mutations in autism spectrum disorders and an empirical test of the neutral DNA model. bioRxiv doi: http://dx.doi.org/10.1101/231944 .

Zhu, Z. et al (2015) Enrichment of Minor Alleles of Common SNPs and Improved Risk Prediction for Parkinson’s Disease. PLoS One 10(7): e0133421.

THE POSITIVE SELECTION AND THE INCREASE OF THE POLYGENIC SCORE DO NOT EQUATE WITH THE INCREASE OF THE COMPLEX TRAIT (II)

A recent meta-analysis found that individuals with an additional 1 Mb of copy-altered interval (be it duplication or deletion) was associated with a 0.132 cm shorter stature, but for each Mb of total deletion burden the decrease of height was 0.41 cm (Mace, 2017). Also, Mace (2017) found an increase of BMI and WHR by the total burden on CNV. I expect there is even a higher effect on intelligence and educational attainment of the total burden on CNV, because the mutational target of both of them must be much higher than the mutational target of height.

De novo structural changes affect on average 4.1kbp of genomic sequence per generation (Kloosterman, 2015). It means 4.1 Mb by 1,000 generations. The effect on genotypic height is 0.54 cm decrease during 25,000-30,000 years. Also Kloosterman (2015) found 66% of structural variants are deletions, and it means 2.71 Mb deleted by 1,000 generations, equating with 1,11 cm shorter stature after 25,000-30,000 years.

Another study found the total burden on CNV in a Han Chinese sample is 44.86 Mb, and the total burden on CNV of an Yoruban sample is 37.75 Mb (Chaisson, 2017). It means a difference of 7.11 MB between Han Chinese and Yoruba produced during 60,000 years, since the Out of Africa. This 7.1 Mb higher burden equates with a 0.94 cm relative decrease of Han Chinese by rapport to Yoruba. The burden on deletions is 25.47 Mb for Han Chinese, and 21.16 Mb for Yoruba (Chaisson, 2017). The difference of 4.31 Mb equates with a 1.77 cm shorter stature of Han Chinese. Differences of total burden of CNV between Han and Yoruba demonstrates that mutational load was not be efficiently eliminated by purifying selection, at least during last 60,000 years in Eurasians. Also, Mace (2017) found rare CNV with large effects on height (>2.4 cm), weight (>5 kg), and body mass index (BMI) (>3.5 kg/m 2 ), that were not eliminated by the purifying selection in Europeans, despite the positive selection on height and BMI found by some studies. The higher burden on CNV of Han Chinese than Yorubans is in line with the study of Henn (2016), that found the mutational load increases with distance from Africa. Also, the mutational load increases during population growth (Gazave, 2013) and during range expansion (Peischl, 2013).

The higher polygenic score on height of Mesolithic hunter gatherers and even of Neolithic farmers than Early Upper Paleolithic Europeans (Berg, 2017) despite their shorter stature than Upper Paleolithic hunter gatherers (Formicola, 1999) could be partially explained by the accumulation of height-decreasing rare CNV, and by the increase of the total burden on CNV after Upper Paleolithic period.

The evolution of height and of POLY_HEIGHT since Early Upper Paleolithic demonstrates than a genotypic trait could decrease even if there is positive selection on this trait, reflected by the increase of the polygenic score on common polymorphism.

 

REFERENCES

Berg, J.J. et al. (2017) Polygenic Adaptation has Impacted Multiple Anthropometric Traits. bioRxiv doi: http://dx.doi.org/10.1101/167551 .

Chaisson, M.J.P. et al. (2017) Multi-platform discovery of haplotype-resolved structural variation in human genomes. bioRxiv doi: http://dx.doi.org/10.1101/193144 .

Formicola, V. & Giannecchini, M. (1999) Evolutionary trends of stature in Upper Palaeolithic and Mesolithic Europe. J. Hum. Evol., 36: 319–333.

Gazave, E., Chang, D., Clark, A.G. & Keinan, A. (2013). Population growth inflates the per-individual number of deleterious mutations and reduces their mean effect. Genetics 195: 969-978.

Henn, B. et al (2016) Distance from sub-Saharan Africa predicts mutational load in diverse human genomes. PNAS 113(4): E440-449 doi:  10.1073/pnas.1510805112

Kloosterman, W.P., Francioli, L.C., Hormozdiari, F., Marschall, T., & Guryev, V. (2015). Characteristics of de novo structural changes in the human genome. Genome Research 25: 792-801.

Mace, A. et al. (2017) CNV-association meta-analysis in 191,161 European adults reveals new loci associated with anthropometric traits. Nature Communications DOI: 10.1038/s41467-017-00556-x

Peischl, S., Dupanloup, I., Kirkpatrick, M. & Excoffier, L. (2013). On the accumulation of deleterious mutations during range expansions. Molecular Ecology 22: 5972-5982.