Per SNP Heritability

GWASLab provides a simple function to calculate the variance explained by each SNP.

Available since v3.4.20

.get_per_snp_r2()

mysumstats.get_per_snp_r2()

It needs effect size BETA and effect allele frequency EAF for calculation. If N is available, it will also calculate the F-statistics.

Options

`.get_per_snp_r2()` options	DataType	Description	Default
`mode`	`q` or `b`	`q`: quantitative trait; `b`: binary trait	`q`

For quantitative traits (mode="q"), GWASLab will use BETA, EAF to calculate SNPR2.

`.get_per_snp_r2()` options	DataType	Description	Default
`vary`	`float` or `se`	Var(Y); if `vary="se"`,Var(Y) will be estimated using `SE`	1
`k`	`int`	k for calculating F	1

For SNP i

When vary=1 and k=1:

h_{i}^{2} = 2 \times β_{i}^{2} \times E A F \times (1 - E A F)

F = h_{i}^{2} \times (n - 2) / (1 - h_{i}^{2})

For binary traits (mode="b"), ncase, ncontrol and prevalence are needed to estimate the variance of liability explained by variants:

`.get_per_snp_r2()` options	DataType	Description	Default
`ncase`	`int`	number of cases	-
`ncontrol`	`int`	number of controls	-
`prevalence`	`float`	prevalence in general population	-

Quote

Equation 10 in Lee, S. H., Goddard, M. E., Wray, N. R., & Visscher, P. M. (2012). A better coefficient of determination for genetic profile analysis. Genetic epidemiology, 36(3), 214-224.

Implementation adopted from TwoSampleMR https://rdrr.io/github/MRCIEU/TwoSampleMR/src/R/add_rsq.r

Examples

Example

mysumstats.get_per_snp_r2()

See Data conversion