Table 1 Estimates of α in simulations

From: Quantification of frequency-dependent genetic architectures in 25 UK Biobank traits reveals action of negative selection

α

\({\mathrm{h}}_{\mathrm{g}}^{2}\)

Sample size

Polygenicity (%)

Imputation noise

LD dependent effects

Mean \(\hat{\mathbf{\alpha}}\)

Mean \(\hat{\mathbf{\alpha}} _{\mathrm{noLD}}\)

−0.3

0.4

5000

1

Yes

Yes

−0.276 ± 0.017

−0.192 ± 0.019

0.0

0.4

5000

1

Yes

Yes

0.021 ± 0.020

0.120 ± 0.017

−0.6

0.4

5000

1

Yes

Yes

−0.573 ± 0.014

−0.471 ± 0.015

−0.3

0.2

5000

1

Yes

Yes

−0.260 ± 0.024

−0.148 ± 0.024

−0.3

0.4

5000

100

Yes

Yes

−0.308 ± 0.012

−0.195 ± 0.013

−0.3

0.4

5000

1

No

Yes

−0.304 ± 0.016

−0.191 ± 0.017

−0.3

0.4

5000

1

Yes

No

−0.373 ± 0.017

−0.284 ± 0.017

−0.3

0.4

2500

1

Yes

Yes

−0.269 ± 0.026

−0.157 ± 0.025

−0.3

0.2

2500

1

Yes

Yes

−0.266 ± 0.052

−0.160 ± 0.034

  1. We simulated phenotypes using imputed UK Biobank genotypes and applied our method to infer α. In each line we show results from phenotypes that were simulated using various values of α, \(h_g^2\), sample size, and the proportion of causal SNPs. In most simulations, imputation noise and LD dependent SNP effects were included in the simulated phenotypes. In each case we report the mean estimated α and standard error of the mean, using our estimation method either with LD correction \(\left( {\hat \alpha } \right)\) or without LD correction \(\left( {\hat \alpha _{{\mathrm{noLD}}}} \right)\).
Back to article page