Correlation between PGS and environmental variables
Davide Piffer
31/03/2018
The purpose of this study is to find inter-correlations between polygenic scores and their correlations with environmental variables, with the goal to shed light on selective pressure on these phenotypes. Principal components for climate were obtained from Hancock et al. (2008), and were extracted from the following variables: precipitation rate, relative humidity, minimum temperature, maximum temperature, mean temperature and short wave radiation flux. The 3 main hypotheses are: 1) Positive correlation of eveningness with latitude and distance from East Africa. According to this theory, a longer circadian period was adaptive for entrainment to the photoperiod that changed with seasons farther away from the equator(Pittendrigh, 1993). Latitudinal clines in circadian period have been found in insects and plants (Hut et al., 2013) 2)Positive correlation between latitude and height PGS, due to Bergmann’s rule. 3)Positive correlation between latitude and PGS EDU (a proxy for general cognitive ability and long term planning), as latitude is correlated with harsher climate and survival challenges of building shelter, storing food and hunting large game. Another research question is whether the positive correlation between eveningness and cognitive ability within populations holds also at the between-population level.
File to upload: “HGDP_PGS.csv” (download from: https://osf.io/ewxqj/)
Create bar charts for PGS_Height, PGS_EDU, PGS_Chron
HGDP_PGS=read.csv(file.choose(),header=TRUE)
#Create bar charts
library(ggplot2)## Warning: package 'ggplot2' was built under R version 3.4.4df_PGS=HGDP_PGS
#Height
ggplot(df_PGS, aes(PGS_Height_Z, Population)) +
geom_segment(aes(x = 0, y = Population, xend = PGS_Height_Z, yend = Population), color = "grey50") +
geom_point()+
labs(title = "Height PGS for 52 HGDP populations",
caption = "Wood et al., 2018 GWAS") 
#EDU
ggplot(df_PGS, aes(PGS_EDU_Z, Population)) +
geom_segment(aes(x = 0, y = Population, xend = PGS_EDU_Z, yend = Population), color = "grey50") +
geom_point()+
labs(title = "EDU PGS for 52 HGDP populations",
caption = "Lee et al., 2018 GWAS") 
#Chronotype
ggplot(df_PGS, aes(PGS_Chron_Z, Population)) +
geom_segment(aes(x = 0, y = Population, xend = PGS_Chron_Z, yend = Population), color = "grey50") +
geom_point()+
labs(title = "Chronotype (eveningness) PGS for 52 HGDP populations",
caption = "Lee et al., 2018 GWAS") 
Compute correlation matrix
## PGS_Chron_Z PGS_Height_Z PGS_EDU_Z Distance.AA Latitude
## PGS_Chron_Z 1.00 0.47 0.37 0.40 0.50
## PGS_Height_Z 0.47 1.00 0.04 0.06 0.37
## PGS_EDU_Z 0.37 0.04 1.00 -0.29 0.57
## Distance.AA 0.40 0.06 -0.29 1.00 -0.23
## Latitude 0.50 0.37 0.57 -0.23 1.00
## B.C_Skin.PGS -0.26 -0.77 -0.08 0.45 -0.49
## B.C_Skin.PGS B.C_Height_PGS Biasutti WinPC1
## PGS_Chron_Z -0.26 0.75 0.73 -0.45
## PGS_Height_Z -0.77 0.86 0.39 -0.15
## PGS_EDU_Z -0.08 0.21 0.59 -0.64
## Distance.AA 0.45 0.30 0.02 0.20
## Latitude -0.49 0.67 0.80 -0.93
## B.C_Skin.PGS 1.00 -0.76 -0.37 0.26## PGS_Chron_Z PGS_Height_Z PGS_EDU_Z Distance.AA Latitude
## PGS_Chron_Z 1 0.47 0.37 0.40 0.50
## PGS_Height_Z NA 1.00 0.04 0.06 0.37
## PGS_EDU_Z NA NA 1.00 -0.29 0.57
## Distance.AA NA NA NA 1.00 -0.23
## Latitude NA NA NA NA 1.00
## B.C_Skin.PGS NA NA NA NA NA
## B.C_Height_PGS NA NA NA NA NA
## Biasutti NA NA NA NA NA
## WinPC1 NA NA NA NA NA
## B.C_Skin.PGS B.C_Height_PGS Biasutti WinPC1
## PGS_Chron_Z -0.26 0.75 0.73 -0.45
## PGS_Height_Z -0.77 0.86 0.39 -0.15
## PGS_EDU_Z -0.08 0.21 0.59 -0.64
## Distance.AA 0.45 0.30 0.02 0.20
## Latitude -0.49 0.67 0.80 -0.93
## B.C_Skin.PGS 1.00 -0.76 -0.37 0.26
## B.C_Height_PGS NA 1.00 0.79 -0.55
## Biasutti NA NA 1.00 -0.74
## WinPC1 NA NA NA 1.00
reorder_cormat <- function(cormat){
# Use correlation between variables as distance
dd <- as.dist((1-cormat)/2)
hc <- hclust(dd)
cormat <-cormat[hc$order, hc$order]
}
# Reorder the correlation matrix
cormat <- reorder_cormat(cormat)
upper_tri <- get_upper_tri(cormat)
# Melt the correlation matrix
melted_cormat <- melt(upper_tri, na.rm = TRUE)
# Create a ggheatmap
ggheatmap <- ggplot(melted_cormat, aes(Var2, Var1, fill = value))+
geom_tile(color = "white")+
scale_fill_gradient2(low = "blue", high = "red", mid = "white",
midpoint = 0, limit = c(-1,1), space = "Lab",
name="Pearson\nCorrelation") +
theme_minimal()+ # minimal theme
theme(axis.text.x = element_text(angle = 45, vjust = 1,
size = 12, hjust = 1))+
coord_fixed()
ggheatmap +
geom_text(aes(Var2, Var1, label = value), color = "black", size = 4) +
theme(
axis.title.x = element_blank(),
axis.title.y = element_blank(),
panel.grid.major = element_blank(),
panel.border = element_blank(),
panel.background = element_blank(),
axis.ticks = element_blank(),
legend.justification = c(1, 0),
legend.position = c(0.6, 0.7),
legend.direction = "horizontal")+
guides(fill = guide_colorbar(barwidth = 7, barheight = 1,
title.position = "top", title.hjust = 0.5))
This preliminary analysis found support for the three main hypotheses. 1) The correlation between eveningness (PGS_Chron_Z) and latitude, distance from AA are r= +0.50, + 0.40, suggesting that eveningness is an adaptation to an environment with a photoperiod that changed with seasons farther away from the equator. The strong correlation (0.93) between latitude and Winter PC (WinPC1) and the slightly higher predictive power of latitude (r=0.5 vs - 0.45) suggests that latitude is the driving force behind the association, as it influences variation in daylight length across the year, which in turn affects the photoperiod.
2)There is a positive correlation (r= 0.37) between latitude and height, supporting Bergmann’s rule.
3)Positive correlation between latitude and polygenic score for Educational attainment (r=0.57), providing support for cold winters theory (Lynn, 1991). Opposite to what was found for chronotype, the strong correlation (0.93) between latitude and Winter PC and the slightly higher predictive power of the winter PC (r=- 0.64 vs 0.57)suggests that latitude is a proxy for winter temperatures.
However, the negative correlation between PGS EDU and distance from East Africa (-0.29) does not support the “Savanna principle” that selection pressure on intelligence increases with distance from the “ancestral environment” (where modern humans putatively evolved) of Eastern Africa (Kanazawa, 2008).
The Height PGS is also highly correlated (r=0.86) to the PGS computed by Berg and Coop (2014), suggesting that the results are robust to different GWAS and PGS computation methods (e.g. different significance thresholds, LD clumping, etc.). The skin pigmentation index (Biasutti) is correlated to PGS EDU, PGS Chronotype and PGS Height (r=0.6, 0.73, 0.39 respectively). There was also a weak positive correlation (0.38) between eveningness and PGS EDU, matching the within-population pattern.
In a follow up study I will take into account the potential confounding factors of linkage disequilibrium and population structure.
References:
Berg, J.J & Coop, G. (2014). A Population Genetic Signal of Polygenic Adaptation.PLoS Genet 10(8): e1004412. https://doi.org/10.1371/journal.pgen.1004412
Hancock AM, Witonsky DB, Gordon AS, Eshel G, Pritchard JK, Coop G, et al. (2008) Adaptations to Climate in Candidate Genes for Common Metabolic Disorders. PLoS Genet 4(2): e32. https://doi.org/10.1371/journal.pgen.0040032
Lynn, R. (1991). The evolution of race differences in intelligence. Mankind Quarterly, 32, 99–173.
Kanazawa, S. (2008). Temperature and evolutionary novelty as forces behind the evolution of general intelligence. Intelligence, 36: 99-108 https://doi.org/10.1016/j.intell.2007.04.001
Pittendrigh CS. 1993 Temporal organisation: reflections of a Darwinian clock-watcher. Annu. Rev. Physiol. 55, 17–54. (doi:10.1146/annurev.ph.55.030193.000313)