Build an LdMatrix

Wrap an LD correlation matrix, or combine several into a mixture, for use as an LD reference. See LdMatrix-class.

One panel's LD. Pass the matrix and the variants it covers. The variants may be given as a GRanges (metadata columns A1, A2, and optionally SNP and AF), as a data frame with columns chrom, pos, A1, A2 (and optionally SNP, AF), or omitted when the matrix's dimnames are chrom:pos:A2:A1 ids, from which they are parsed. The correlations must be signed by the dosage of A1.

A mixture. Pass a list of LdMatrix objects and their mixtureWeights. The result is their weighted average \(\sum_k w_k R_k\), computed once here, over the variants every component covers; a variant some component lacks is left out, with a message. Components are matched by position and alleles and put into the first component's allele coding, so they need not share variant ids or allele orientation. The mixture's AF is \(\sum_k w_k AF_k\) when every component records AF, and its nSamples is \(1 / \sum_k w_k^2 / n_k\) when every component records a size.

Usage

LdMatrix(
  correlation,
  variants = NULL,
  nSamples = NA_integer_,
  mixtureWeights = NULL
)

Arguments

correlation

A numeric correlation matrix, or a list of LdMatrix objects to combine into a mixture.

variants

The variants the matrix covers, in its row order: a GRanges or a data frame (see Description), or NULL to parse them from the matrix's dimnames. Not used for a mixture.

nSamples

The size of the panel the LD was estimated from, or NA when unknown. Not used for a mixture.

mixtureWeights

For a mixture, one non-negative weight per component, summing to 1.

Value

An LdMatrix.

Examples

data(eqtlRegionExample)
X <- eqtlRegionExample$X[, 1:8]
ld <- LdMatrix(cor(X), nSamples = nrow(X))
ld
#> LdMatrix: 8 variants, panel size 415
#>   no allele frequencies
ldMatrix(ld)[1:3, 1:3]
#>                    chr22:32119788:T:C chr22:32119867:T:G chr22:32119961:T:G
#> chr22:32119788:T:C          1.0000000          1.0000000         -0.3896071
#> chr22:32119867:T:G          1.0000000          1.0000000         -0.3896071
#> chr22:32119961:T:G         -0.3896071         -0.3896071          1.0000000

# A mixture of two panels' LD, weighted 3:1.
half <- seq_len(nrow(X) / 2)
ldA <- LdMatrix(cor(X[half, ]), nSamples = length(half))
ldB <- LdMatrix(cor(X[-half, ]), nSamples = nrow(X) - length(half))
LdMatrix(list(ldA, ldB), mixtureWeights = c(0.75, 0.25))
#> LdMatrix: 8 variants, panel size 331
#>   mixture of 2 panels, weights 0.75, 0.25
#>   no allele frequencies