identityCoefs: Omnibus function for identity coefficients

In magnusdv/ribd: Pedigree-based Relatedness Coefficients

Omnibus function for identity coefficients

Description

This function calculates the pairwise identity coefficients described by Jacquard (1974). Unlike the previous condensedIdentity() (which will continue to exist), this function also computes the 15 detailed identity coefficients. The implementation supports pedigrees with inbred founders, and X-chromosomal coefficients.

Usage

identityCoefs(
  x,
  ids = labels(x),
  detailed = FALSE,
  Xchrom = FALSE,
  self = FALSE,
  simplify = TRUE,
  method = c("auto", "K", "WL", "LS", "GC", "idcoefs", "identity", "merlin"),
  verbose = FALSE,
  ...
)

detailed2condensed(d)

Arguments

x	A pedigree in the form of a pedtools::ped object.
ids	A vector of two ID labels.
detailed	A logical. If FALSE (default), the 9 condensed coefficients are computed; otherwise the 15 detailed identity coefficients.
Xchrom	A logical, by default FALSE.
self	A logical indicating if self-relationships (i.e., between a pedigree member and itself) should be included. FALSE by default.
simplify	Simplify the output (to a numeric of length 9) if ids has length 2. Default: TRUE.
method	Either "auto", "K", "WL", "LS", "GC", "idcoefs", "identity" or "merlin". By default ("auto") a suitable algorithm is chosen automatically.
verbose	A logical.
...	Further arguments.
d	Either a numeric vector of length 15, or a data frame with 17 columns.

Details

Both the condensed and detailed coefficients are given in the orders used by Jacquard (1974). The function detailed2condensed() converts from detailed coefficients (d1, ..., d15) to condensed ones (D1, ..., D9) using the following relations:

• D1 = d1

• D2 = d6

• D3 = d2 + d3

• D4 = d7

• D5 = d4 + d5

• D6 = d8

• D7 = d9 + d12

• D8 = d10 + d11 + d13 + d14

• D9 = d15

Algorithms for computing identity coefficients

The following is a brief overview of various algorithms for computing (single-locus) condensed and/or detailed identity coefficients. This topic is closely linked to that of generalised kinship coefficients, which is further described in the documentation of gKinship().

For each algorithm below, it is indicated in brackets how to enforce it in identityCoefs().

• Karigl (1981) gave the first recursive algorithm for the 9 condensed identity coefficients. [method = "K"]

• Weeks & Lange (1988) suggested a broader and more natural generalisation of kinship coefficients, leading to a slightly different algorithm for condensed coefficients. [method = "WL"]

• Lange & Sinsheimer (1992) described an even further generalisation of kinship coefficients, allowing a mix of deterministic and random sampling of alleles. They used this to give (i) an alternative algorithm for the 9 condensed identity coefficients, and (ii) an algorithm for the 15 detailed coefficients. [method = "LS"]

• The C program IdCoefs (version 2.1.1) by Mark Abney (2009) uses a graph model to obtain very fast computation of condensed identity coefficients. This requires IdCoefs to be installed on the computer (see link under References) and available on the system search path. The function then writes the necessary files to disk and calls IdCoefs via system(). [method = "idcoefs"]

• The R package identity provides an R interface for IdCoefs, avoiding calls to system(). [method = "identity"]

• The MERLIN software (Abecasis et al, 2002) offers an option "–extended" for computing detailed identity coefficients. This option requires MERLIN to be installed on the system. The function then writes the necessary files to disk and calls MERLIN via system(). If detailed = FALSE, the coefficients are transformed with detailed2condensed() before returning. Note: MERLIN rounds all numbers to 3 decimal places. Since this rounding is done on the detailed coefficients, rounding errors may happen when converting to the condensed ones. [method = "merlin"]

Value

A data frame with L + 2 columns, where L is either 9 or 15 (if detailed = TRUE).

If simplify = TRUE and length(ids) = 2: A numeric vector of length L.

References

• Jacquard, A. (1974). The Genetic Structure of Populations. Springer.

• Karigl, G. (1981). A recursive algorithm for the calculation of identity coefficients. Ann. Hum. Genet.

• Weeks, D.E. & Lange, K. (1988). The affected-pedigree-member method of linkage analysis. Am. J. Hum. Genet

• Lange, K. & Sinsheimer, J.s. (1992). Calculation of genetic identity coefficients. Ann. Hum. Genet.

• Abney, M. (2009). A graphical algorithm for fast computation of identity coefficients and generalized kinship coefficients. Bioinformatics, 25, 1561-1563. https://home.uchicago.edu/~abney/abney_web/Software.html

See Also

condensedIdentity(), gKinship()

Examples

x = fullSibMating(1)

### Condensed coefficients
j1 = identityCoefs(x, method = "K")
j2 = identityCoefs(x, method = "WL")
j3 = identityCoefs(x, method = "LS")
j4 = identityCoefs(x, method = "GC")
j5 = condensedIdentity(x, ids = 1:6) # legacy version

stopifnot(all.equal(j1,j2), all.equal(j1,j3), all.equal(j1,j4), all.equal(j1,j5))

### Detailed coefficients
jdet1 = identityCoefs(x, detailed = TRUE, method = "LS")
jdet2 = identityCoefs(x, detailed = TRUE, method = "GC")

stopifnot(all.equal(jdet1,jdet2))

### X-chromosomal coefficients
jx1 = identityCoefs(x, Xchrom = TRUE, method = "K")
jx2 = identityCoefs(x, Xchrom = TRUE, method = "GC")
jx3 = condensedIdentityX(x, ids = 1:6)  # legacy version

stopifnot(all.equal(jx1,jx2), all.equal(jx1,jx3))

### Detailed X-chromosomal coefficients
jdx = identityCoefs(x, detailed = TRUE, Xchrom = TRUE, method = "GC")

stopifnot(all.equal(detailed2condensed(jdx), jx1))

magnusdv/ribd documentation built on March 29, 2024, 5:20 a.m.