Open pre-computed polygenic scores. My PGS-computing code and the 1000 Genomes files were made publicly available on OSF (https://osf.io/g9yx8/). They were converted to .csv format so that anyone can load them on R without having to convert them from VCF format or PLINk-like formats (i.e. .bed). I will compute correlations between the latest polygenic score, as inferred from the largest GWAS of Educational attainment (EDU) to date and the polygenic score inferred from previous GWAS of Educational attainment and IQ (Piffer, 2015).

PS_IQ=read.csv(file.choose(),header=TRUE)#Lee_PGS_1kg - IQ_EDU.csv
PS_IQ$Leeetal_PGS=scale(PS_IQ$Leeetal2018,center = TRUE, scale= TRUE)
PS_IQ$Piffer2015_PGS=scale(PS_IQ$Piffer2015,center=TRUE, scale=TRUE)
library(ggplot2)
library(ggrepel)

Show standardized PGS from latest EDU GWAS (Lee et al., 2018) and population IQs

PS_IQ[-c(2,3)]
##             Population Piffer_2017 X1180_PGS Piffer.2015.IQ  Leeetal_PGS
## 1     Afr.Car.Barbados       0.224 0.2695543           83.0 -1.299572787
## 2            US Blacks       0.259 0.2546149           85.0 -1.272171263
## 3   Bengali Bangladesh       0.353 0.2329476           81.0 -0.194490832
## 4          Chinese Dai       0.425 0.2417140             NA  1.011634319
## 5          Utah Whites       0.384 0.2316542           99.0  0.686818769
## 6      Chinese, Bejing       0.481 0.2421085          105.0  1.512552558
## 7       Chinese, South       0.462 0.2460318          105.0  1.321665571
## 8            Colombian       0.343 0.2409787           83.5  0.183599380
## 9        Esan, Nigeria       0.227 0.2668967           71.0 -1.300152503
## 10             Finland       0.442 0.2236820          101.0  1.021133946
## 11         British, GB       0.416 0.2329586          100.0  0.523879885
## 12 Gujarati Indian, Tx       0.389 0.2301143             NA  0.179386510
## 13             Gambian       0.218 0.2753084           62.0 -1.299641371
## 14      Iberian, Spain       0.396 0.2605518           97.0  0.893035210
## 15   Indian Telegu, UK       0.365 0.2305200             NA  0.201000256
## 16               Japan       0.458 0.1893100          105.0  1.443193070
## 17             Vietnam       0.450 0.2227933           99.4  1.086033831
## 18        Luhya, Kenya       0.231 0.2558614           74.0 -1.370743979
## 19 Mende, Sierra Leone       0.231 0.2509300           64.0 -1.517689441
## 20     Mexican in L.A.       0.335 0.2525660           88.0 -0.318167144
## 21      Peruvian, Lima       0.299 0.2355947           85.0 -0.996212200
## 22   Punjabi, Pakistan       0.378 0.2375246           84.0 -0.056823359
## 23        Puerto Rican       0.345 0.2387469           83.5  0.058261889
## 24      Sri Lankan, UK       0.361 0.2374865           79.0  0.005171638
## 25      Toscani, Italy       0.396 0.2214031           99.0  0.711938535
## 26     Yoruba, Nigeria       0.231 0.2649196           71.0 -1.213640486
##    Piffer2015_PGS
## 1     -1.25437693
## 2     -1.04680466
## 3      0.02962202
## 4      0.83495705
## 5      0.52870288
## 6      1.53140171
## 7      1.32496372
## 8      0.02281637
## 9     -1.38935562
## 10     0.64439890
## 11     0.67502432
## 12     0.33360764
## 13    -1.56630247
## 14     0.52416578
## 15     0.26555116
## 16     1.41116859
## 17     1.22741609
## 18    -1.69220696
## 19    -1.49257461
## 20     0.04436759
## 21     0.19749467
## 22     0.16346643
## 23    -0.04524011
## 24    -0.08834255
## 25     0.45043793
## 26    -1.63435895

Display bar chart for the more visually inclined:

ggplot(PS_IQ, aes(Leeetal_PGS, Population)) +
  geom_segment(aes(x = 0, y = Population, xend = Leeetal_PGS, yend = Population), color = "grey50") +
  geom_point()+
  labs(title = "EDU PGS for 26 1KG populations",
       caption = "Lee et al., 2018 GWAS") +
  scale_x_continuous(name="EDU_3")

Correlaton between 2018 EDU PGS and IQ

ggplot(PS_IQ, aes(x=Leeetal_PGS, y=Piffer.2015.IQ)) +
  geom_point(shape=1) +    # Use hollow circles
  geom_smooth(method=lm)+
geom_text_repel(aes(label=Population), size=3)+
  scale_x_continuous(name="EDU_3")+
  scale_y_continuous(name="Population IQ")

Correlation between 2018 EDU PGS and 2015 Piffer PGS

ggplot(PS_IQ, aes(x=Leeetal_PGS, y=Piffer2015_PGS)) +
  geom_point(shape=1) +    # Use hollow circles
  geom_smooth(method=lm)+
geom_text_repel(aes(label=Population), size=3)+
scale_x_continuous(name="EDU_3")+
scale_y_continuous(name="Piffer 2015 PGS")

Visualize correlation matrix

##             Var1        Var2 value
## 1    Piffer_2017 Piffer_2017  1.00
## 2      X1180_PGS Piffer_2017 -0.72
## 3 Piffer.2015.IQ Piffer_2017  0.90
## 4    Leeetal_PGS Piffer_2017  0.98
## 5 Piffer2015_PGS Piffer_2017  0.97
## 6    Piffer_2017   X1180_PGS -0.72
##                Piffer_2017 X1180_PGS Piffer.2015.IQ Leeetal_PGS
## Piffer_2017              1     -0.72           0.90        0.98
## X1180_PGS               NA      1.00          -0.65       -0.67
## Piffer.2015.IQ          NA        NA           1.00        0.90
## Leeetal_PGS             NA        NA             NA        1.00
## Piffer2015_PGS          NA        NA             NA          NA
##                Piffer2015_PGS
## Piffer_2017              0.97
## X1180_PGS               -0.72
## Piffer.2015.IQ           0.91
## Leeetal_PGS              0.95
## Piffer2015_PGS           1.00

The correlation between the 2015 and the 2018 estimate of polygenic intelligence are almost identical and they predict population IQ equally well, demonstrating the high replicability of my results.

References:

Lee et al. (2018). Gene discovery and polygenic prediction from a 1.1-million-person GWAS of educational attainment“. Nature Genetics

Piffer, D. (2015). A review of intelligence GWAS hits: Their relationship to country IQ and the issue of spatial autocorrelation. Intelligence, 53: 43-50.

Piffer, D. (2018, May 25). 1000 Genomes frequencies. Retrieved from osf.io/g9yx8