mpIM: (Composite) Interval Mapping for QTL detection in multi-parent crosses

Description

Interval mapping in multi-parent crosses with options for single-stage mixed model approach; multi-stage approach using predicted means; multi-stage approach including cofactors (CIM)

Usage

mpIM(baseModel, object, pheno, idname = "id", threshold = 0.001, chr, step = 0, responsename = "predmn", ncov = 1000, window = 10, mrkpos = FALSE, ...)

Arguments

baseModel

Base phenotypic model for analysis

object

Object of class mpcross

pheno

Phenotypic object

idname

The idname in phenotypic data for which to output predicted means. Should match rownames of the object$finals

threshold

Significance threshold for QTL p-values

chr

Subset of chromosomes for which to compute QTL profile

step

Step size at which to compute the QTL profile. See mpprob for further description of default values

responsename

Optional input of response name to look for in object$pheno

ncov

Number of marker covariates to search for - default is to search for as many as possible using stepAIC (forward/backward selection)

window

Window of cM on each side of markers where we exclude covariates in CIM

mrkpos

Flag for whether to consider both marker positions and step positions or just steps. Is overridden if step=0

...

Additional arguments

Value

pheno: Input phenotype data
pvalue: Each component contains estimated p-values at each position on a given chromosome
wald: Each component contains Wald statistics at each position on a given chromosome
fndrfx: Each component contains founder effects estimated at each position on a given chromosome
qtl: Each component contains the position and effects of a detected QTL
call: Input arguments to function

Details

Depending on the options selected, different models will be fit for QTL detection. If the baseModel input does not include a term matching the idname input, it will be assumed that a single-stage QTL mapping approach is desired. In this case, no covariates will be added (ncov will be set to 0); all models will be fitted in asreml; and all phenotypic covariates and design factors specified in the baseModel will be fitted along with genetic covariates in mixed model interval mapping.

If the baseModel input does include a term matching the idname, then it will be assumed that a two-stage QTL mapping approach is desired. In this case, the baseModel will be fit using asreml and predicted means will be output to be used as a response in linear model interval mapping. If ncov>0 additional marker cofactors will be fit; otherwise simple interval mapping will be run. All phenotypic covariates and design factors specified in the baseModel will be fit in the first stage.

Note that no weights are used in the second stage of analysis which may result in a loss of efficiency compared to a one-stage approach.

If no baseModel is input, it will be assumed that predicted means have been included in object as a phenotypic variable named predmn. In this case pheno is not required and asreml does not need to be used. (composite) Interval mapping will proceed as in the two-stage case depending on the value of ncov.

Examples

Run this code

sim.map <- sim.map(len=rep(100, 2), n.mar=11, include.x=FALSE, eq.spacing=TRUE)
sim.ped <- sim.mpped(4, 1, 500, 6, 1)
sim.dat <- sim.mpcross(map=sim.map, pedigree=sim.ped, qtl=matrix(data=c(1, 10, .4, 0, 0, 0, 1, 70, 0, .35, 0, 0), nrow=2, ncol=6, byrow=TRUE), seed=1)
mpp.dat <- mpprob(sim.dat, program="qtl")
## Two-stage simple interval mapping
mpq.dat <- mpIM(object=mpp.dat, ncov=0, responsename="pheno")