Constructs the SNP additive genomic relationship matrix (G) following VanRaden (2008) Method 1. The resulting matrix captures additive genetic relationships among individuals based on SNP marker data, and can be used directly in masreml() or gwablup() via the G argument.
build_G_snp(W, ref_W = NULL)
W
|
numeric matrix (n x m) of raw genotype codes, where n is the number of individuals and m is the number of SNP markers. Values must be 0, 1, or 2 representing the number of copies of the reference allele. Row names are used as individual IDs. Example format: SNP1 SNP2 SNP3 ind1 0 1 2 ind2 1 0 1 ind3 2 1 0 |
SNP additive GRM (VanRaden 2008). The centred genotype matrix \(W\) is first constructed by:
\(W_{ij} = X_{ij} - 2p_j\)
where \(X_{ij} \in \{0,1,2\}\) is the allele dosage of individual \(i\) at SNP \(j\) and \(p_j\) is the allele frequency. The additive GRM is then:
\(G = \frac{WW^\top}{2\sum_j p_j(1-p_j)}\)
The denominator \(2\sum_j p_j(1-p_j)\) scales \(G\) so that diagonal elements average approximately 1 (i.e., similar to the pedigree-based numerator relationship matrix).
numeric matrix (n x n) of genomic relationships. Diagonal elements approximate 1 + inbreeding coefficient. Off-diagonal elements represent genomic relationships between pairs of individuals.
VanRaden, P. M. (2008) Efficient methods to compute genomic predictions. J. Dairy Sci. 91:4414-4423. doi:10.3168/jds.2007-0980
build_G_mh, build_D_snp, masreml
library("masreml")
d <- load_data("small")
G <- build_G_snp(d$snp)
# Diagonal should be ~1 + inbreeding coefficient
round(summary(diag(G)), 3)
# Train-only G with leakage-safe full-set G for prediction
G_train <- build_G_snp(d$snp[d$train_idx, ])
G_full <- build_G_snp(d$snp, ref_W = d$snp[d$train_idx, ])