haplo.score: Score Statistics for Association of Traits with Haplotypes

Description

Compute score statistics to evaluate the association of a trait with haplotypes, when linkage phase is unknown and diploid marker phenotypes are observed among unrelated subjects. For now, only autosomal loci are considered.

Usage

haplo.score(y, geno, trait.type="gaussian", offset = NA, x.adj = NA,
            min.count=5, skip.haplo=min.count/(2*nrow(geno)),
            locus.label=NA, miss.val=c(0,NA), haplo.effect="additive",
            eps.svd=1e-5, simulate=FALSE, sim.control=score.sim.control(),
            em.control=haplo.em.control())

Value

List with the following components:

score.global: Global statistic to test association of trait with haplotypes that have frequencies >= skip.haplo.
df: Degrees of freedom for score.global.
score.global.p: P-value of score.global based on chi-square distribution, with degrees of freedom equal to df.
score.global.p.sim: P-value of score.global based on simulations (set equal to NA when simulate=F).
score.haplo: Vector of score statistics for individual haplotypes that have frequencies >= skip.haplo.
score.haplo.p: Vector of p-values for score.haplo, based on a chi-square distribution with 1 df.
score.haplo.p.sim: Vector of p-values for score.haplo, based on simulations (set equal to NA when simulate=F).
score.max.p.sim: Simulated p-value indicating for simulations the number of times a maximum score.haplo value exceeds the maximum score.haplo from the original data (equal to NA when simulate=F).
haplotype: Matrix of hapoltypes analyzed. The ith row of haplotype corresponds to the ith item of score.haplo, score.haplo.p, and score.haplo.p.sim.
hap.prob: Vector of haplotype probabilies, corresponding to the haplotypes in the matrix haplotype.
locus.label: Vector of labels for loci, of length K (same as input argument).
call: The call to the haplo.score function; useful for recalling what parameters were used.
haplo.effect: The haplotype effect model parameter that was selected for haplo.score.
simulate: Same as function input parameter. If [T]rue, simulation results are included in the haplo.score object.
n.val.global: Vector containing the number of valid simulations used in the global score statistic simulation. The number of valid simulations can be less than the number of simulations requested (by sim.control) if simulated data sets produce unstable variances of the score statistics.
n.val.haplo: Vector containing the number of valid simulations used in the p-value simulations for maximum-score statistic and scores for the individual haplotypes.

Arguments

y: Vector of trait values. For trait.type = "binomial", y must have values of 1 for event, 0 for no event.
geno: Matrix of alleles, such that each locus has a pair of adjacent columns of alleles, and the order of columns corresponds to the order of loci on a chromosome. If there are K loci, then ncol(geno) = 2*K. Rows represent alleles for each subject.
trait.type: Character string defining type of trait, with values of "gaussian", "binomial", "poisson", "ordinal".
offset: Vector of offset when trait.type = "poisson"
x.adj: Matrix of non-genetic covariates used to adjust the score statistics. Note that intercept should not be included, as it will be added in this function.
min.count: The minimum number of counts for a haplotype to be included in the model. First, the haplotypes selected to score are chosen by minimum frequency greater than skip.haplo (based on min.count, by default). It is also used when haplo.effect is either dominant or recessive. This is explained best in the recessive instance, where only subjects who are homozygous for a haplotype will contribute information to the score for that haplotype. If fewer than min.count subjects are estimated to be affected by that haplotype, it is not scored. A warning is issued if no haplotypes can be scored.
skip.haplo: Minimum haplotype frequency for which haplotypes are scored in the model. By default, the frequency is based on "min.count" divided by the 2*N total haplotype occurrences in the sample.
locus.label: Vector of labels for loci, of length K (see definition of geno matrix)
miss.val: Vector of codes for missing values of alleles
haplo.effect: the "effect" of a haplotypes, which determines the covariate (x) coding of haplotypes. Valid options are "additive" (causing x = 0, 1, or 2, the count of a particular haplotype), "dominant" (causing x = 1 if heterozygous or homozygous carrier of a particular haplotype; x = 0 otherwise), and "recessive" (causing x = 1 if homozygous for a particular haplotype; x = 0 otherwise).
eps.svd: epsilon value for singular value cutoff; to be used in the generalized inverse calculation on the variance matrix of the score vector (see help(Ginv) for details).
simulate: Logical: if FALSE, no empirical p-values are computed; if TRUE, simulations are performed. Specific simulation parameters can be controlled in the sim.control parameter list.
sim.control: A list of control parameters to determine how simulations are performed for simulated p-values. The list is created by the function score.sim.control and the default values of this function can be changed as desired. See score.sim.control for details.
em.control: A list of control parameters to determine how to perform the EM algorithm for estimating haplotype frequencies when phase is unknown. The list is created by the function haplo.em.control - see this function for more details.

Details

Compute the maximum likelihood estimates of the haplotype frequencies and the posterior probabilities of the pairs of haplotypes for each subject using an EM algorithm. The algorithm begins with haplotypes from a subset of the loci and progressively discards those with low frequency before inserting more loci. The process is repeated until haplotypes for all loci are established. The posterior probabilities are used to compute the score statistics for the association of (ambiguous) haplotypes with traits. The glm function is used to compute residuals of the regression of the trait on the non-genetic covariates.

References

Schaid DJ, Rowland CM, Tines DE, Jacobson RM, Poland GA. "Score tests for association of traits with haplotypes when linkage phase is ambiguous." Amer J Hum Genet. 70 (2002): 425-434.

Examples

Run this code

  # establish all hla.demo data, 
  # remove genotypes with missing alleles just so haplo.score runs faster 
  # with missing values included, this example takes 2-4 minutes
  # FOR REGULAR USAGE, DO NOT DISCARD GENOTYPES WITH MISSING VALUES

  data(hla.demo)
  geno <- as.matrix(hla.demo[,c(17,18,21:24)])
  keep <- !apply(is.na(geno) | geno==0, 1, any)
  hla.demo <- hla.demo[keep,]
  geno <- geno[keep,]
  attach(hla.demo)
  label <- c("DQB","DRB","B")
 
# For quantitative, normally distributed trait:

  score.gaus <- haplo.score(resp, geno, locus.label=label, 
                            trait.type = "gaussian")
  print(score.gaus)

# For ordinal trait:
  y.ord <- as.numeric(resp.cat)
  score.ord <- haplo.score(y.ord, geno, locus.label=label,
                           trait.type="ordinal")
  print(score.ord)

# For a  binary trait and simulations,
# limit simulations to 500 in score.sim.control, default is 20000
  y.bin <-ifelse(y.ord==1,1,0)
  score.bin.sim <- haplo.score(y.bin, geno, trait.type = "binomial",
                     locus.label=label, simulate=TRUE,
                     sim.control=score.sim.control(min.sim=200,max.sim=500))

  print(score.bin.sim)

# For a binary trait, adjusted for sex and age:

  x <- cbind(male, age)
  score.bin.adj <- haplo.score(y.bin, geno, trait.type = "binomial", 
                               locus.label=label, x.adj=x)
  print(score.bin.adj)

Run the code above in your browser using DataLab