EpiTest.fit {EpiTest}

Inference and Test

Description

This function fits the EpiTest model on a bi-parental population dataset, then outputs the estimates of the fixed and random effects, along with their associated test results.

Usage

EpiTest.fit(Ancestry, Pheno, Trait, ParentName, Parents = NULL, Weight = FALSE)

Arguments

Ancestry

a numeric matrix with 0/1 entries representing the allele ancestries of a bi-parental population at a set of markers (0: homozygous for parent A alleles and 1: homozygous for parent B alleles)

Pheno

a data.frame with a Genotype character column indicating the names of individuals, a numeric column for each trait to be analyzed, and possibly a Weight numeric column indicating the number of observation for each individual. Missing values are discarded if present for a trait

Trait

a string indicating the name of the phenotypic trait

ParentName

a character vector of length 2 providing the parent names where the first element should be named P0 (parent whose allele ancestries are coded 0) and the second element should be named P1 (parent whose allele ancestries are coded 1)

Parents

(optional) a data.frame with a Genotype character column indicating the names of parents, a Family character column indicating the names of families/populations, a numeric column for each trait to be analyzed, and possibly a Weight numeric column indicating the number of observation for each parent (see Details)

Weight

a boolean indicating whether weights indicating the number of observations should be used for the inference. If TRUE, a column Weight should be found in Pheno and Parents (see Details)

Details

Depending on cases, the data may be organized in two different ways.

In Case 1, the Ancestry matrix and the Pheno data.frame respectively provide the whole ancestry and phenotypic data, including the parental data.

In Case 2, the phenotypic data are split into two parts, the parental data being included in a separate Parents data.frame.

In both cases, one may provide a set of weights associated to each individual by including a Weight column in the Pheno and Parents data.frames. These weights correspond to the number of observations that where used to compute the parental BLUPS/mean values. These weights must be provided as a Weight column in the Parents and Pheno data.frame.

The fitted EpiTest model includes three variance components: the segregation variance, the (segregation x segregation) variance and the error variance. For each genetic variance a likelihood ratio test is performed.

The fitted EpiTest model includes three fixed parameters: the intercept, the linear regression coefficient (beta) on parent proportions and the quadratic regression coefficient (delta) on parent proportions that only involves epistatic effects. Each fixed parameter is tested using a Wald test.

Additionally, epistasis can be tested by testing that both the (segregation x segregation) variance component and the quadratic mean component (delta) are null through a likelihood ratio test. Note that in this case the LRT is based on the full (i.e. unrestricted) likelihoods.

The function outputs a list of five items: Beta is the vector of estimated fixed effects, Sigma2 is the vector of estimated variances, Test_full, Test_fixed, Text_random are 3 data.table that provide the results of the tests for no epistasis (jointly tested on fixed and variance components), and for the fixed and random effects, respectively.

Value

A list of four items: Beta, Sigma2, Test_fixed, Test_random

Examples

## One bi-parental population, no weighting and no parental phenotypes
data(Pheno.list)
data(Ancestry.list)
Ancestry <- Ancestry.list[[1]]
Pheno <- Pheno.list[[1]]
ParentName <- c(P0 = 'B73',P1 = 'B97')
ETest.1 <- EpiTest.fit(Ancestry = Ancestry,
                       Pheno = Pheno,
                       ParentName = ParentName,
                       Trait = "GDD_DTA")

## One bi-parental population, with weights and parental phenotypes
data(Parents)
ETest.2 <- EpiTest.fit(Ancestry = Ancestry,
                       Pheno = Pheno,
                       ParentName = ParentName,
                       Trait = "GDD_DTA",
                       Parents = Parents,
                       Weight=TRUE)

## Full NAM analysis, with weights and parental phenotypes
Parent.list <- Parents$Genotype[-1]
names(Parent.list) <- Parents$Family[-1]
ETest.nam <- purrr::imap(Parent.list, ~ EpiTest.fit(Ancestry = Ancestry.list[[.y]],
                                                    Pheno = Pheno.list[[.y]],
                                                    ParentName=c(P0 = 'B73',P1 = .x),
                                                    Parents = Parents,
                                                    Trait = 'GDD_DTA',
                                                    Weight = TRUE))