R: P-values adjusted for latent factors computed by 'lfmm2'.

lfmm2.test {LEA}

R Documentation

P-values adjusted for latent factors computed by `lfmm2`.

Description

The function returns a vector of p-values for association between loci and environmental variables adjusted for latent factors computed by lfmm2. It takes an object of class lfmm2Class with the data that were used to adjust the LFMM.

Usage

lfmm2.test (object, input, env, genomic.control, linear, family)

Arguments

`object`	An object of class lfmm2Class.
`input`	A genotypic matrix or a character string containing a path to the input file. The genotypic matrix must be in the `lfmm{lfmm_format}` format without missing values (9 or `NA`). See `impute` for completion based on nonnegative matrix factorization and consider R packages for reading large matrices.
`env`	A matrix of environmental covariates or a character string containing a path to the environmental file. The environment matrix must be in the `env` format without missing values. Variables must be encoded as `numeric`.
`genomic.control`	A logical value. If `TRUE`, the p-values are recalibrated by using genomic control after correction for confounding.
`linear`	A logical value indicating whether linear or generalized linear models should be used to perform the association tests. If `FALSE`, `family` should be provided in the next argument.
`family`	a `family` for generalized linear models used in the association tests. The default is binomial(link = "logit")), which requires that y is between 0 and 1.

Value

`pvalues`	A matrix of p-values for each locus and each environmental variable.
`zscores`	A matrix of z-scores for each locus and each environmental variable.
`gif`	Genomic inflation factors computed for each environmental variable.

Author(s)

Olivier Francois

References

Caye K, Jumentier B, Lepeule J, Francois O. (2019). LFMM 2: fast and accurate inference of gene-environment associations in genome-wide studies. Molecular biology and evolution, 36(4), 852-860.

Examples

### Example of analysis using lfmm2 ###

# Simulation with 10 target loci, with effect sizes ranging between -10 an 10 
# n = 100 individuals and L = 1000 loci

X <- as.matrix(rnorm(100)) # environmental variable
B <- rep(0, 1000) 
target <- sample(1:1000, 10) # target loci
B[target] <- runif(10, -10, +10) # effect sizes

# Creating hidden factors and loadings

U <- t(tcrossprod(as.matrix(c(-1,0.5,1.5)), X)) + matrix(rnorm(300), ncol = 3)
V <- matrix(rnorm(3000), ncol = 3)

# Simulating a binarized matrix containing haploid genotypes 
# Simulation performed with the generative LFMM

Y <- tcrossprod(as.matrix(X), B) + tcrossprod(U, V) + matrix(rnorm(100000, sd = .5), nrow = 100)
Y <- matrix(as.numeric(Y > 0), ncol = 1000)

######################################
# Fitting an LFMM with K = 3 factors #
######################################

mod2 <- lfmm2(input = Y, env = X, K = 3)

# Computing P-values and plotting their minus log10 values 
# Target loci are highlighted

pv <- lfmm2.test(object = mod2, input = Y, env = X, linear = TRUE)
plot(-log10(pv$pvalues), col = "grey", cex = .4, pch = 19)
points(target, -log10(pv$pvalues[target]), col = "red")

[Package LEA version 2.99.0 Index]

P-values adjusted for latent factors computed by lfmm2.