Title:	Meta-Analysis for Detecting Genotype x Environment Associations
Version:	1.2.2
Maintainer:	Annaïg De Walsche <annaig.de-walsche@inrae.fr>
Description:	Provides functions to perform all steps of genome-wide association meta-analysis for studying Genotype x Environment interactions, from collecting the data to the manhattan plot. The procedure accounts for the potential correlation between studies. In addition to the Fixed and Random models, one can investigate the relationship between QTL effects and some qualitative or quantitative covariate via the test of contrast and the meta-regression, respectively. The methodology is available from: (De Walsche, A., et al. (2025) \doi{10.1371/journal.pgen.1011553}).
License:	GPL-3
Depends:	R (≥ 3.0.2)
Imports:	corrplot, data.table, dplyr, emdbook, furrr, future, ggplot2, ggrepel, gplots, graphics, grDevices, ks, purrr, qqman, Rfast, stats, stringr, tibble, tidyr, utils, viridis
Suggests:	DT, knitr, rmarkdown, tidyverse
VignetteBuilder:	knitr
Encoding:	UTF-8
LazyData:	true
RoxygenNote:	7.3.2
NeedsCompilation:	no
Packaged:	2025-07-21 09:21:10 UTC; Annaig
Author:	Annaïg De Walsche [aut, cre], Tristan Mary-Huard [aut]
Repository:	CRAN
Date/Publication:	2025-07-21 12:30:07 UTC

metaGE: Meta-Analysis for Detecting Genotype x Environment Associations

Description

Provides functions to perform all steps of genome-wide association meta-analysis for studying Genotype x Environment interactions, from collecting the data to the manhattan plot. The procedure accounts for the potential correlation between studies. In addition to the Fixed and Random models, one can investigate the relationship between QTL effects and some qualitative or quantitative covariate via the test of contrast and the meta-regression, respectively. The methodology is available from: (De Walsche, A., et al. (2025) doi:10.1371/journal.pgen.1011553).

Details

The main functions of the package corresponding to the steps of a genome-wide association meta-analysis are:

• metaGE.collect: Collect the results of genome-wide association studies data from different files.

• metaGE.cor: Infer the correlation between studies.

• metaGE.fit: Perform global meta-analysis test procedure for quantitative trait loci detection (using a Fixed or Random effect model)

• metaGE.test: Perform tests of contrast or meta-regression investigate the relationship between marker effects and some qualitative or quantitative covariate, respectively.

• metaGE.lscore: Apply the local score approach as multiple tests correction.

• metaGE.manhattan: Draw a Manhattan plot.

Author(s)

Maintainer: Annaïg De Walsche annaig.de-walsche@inrae.fr (ORCID)

Authors:

• Tristan Mary-Huard tristan.mary-huard@agroparistech.fr (ORCID)

Check and reformat the matrix of contrast

Description

The function CheckContrast check and reformat the matrix of contrast.

Usage

CheckContrast(Contrast, ContrastName)

Arguments

Value

The matrix of contrast in the right format.

Check and reformat the matrix of incidence

Description

The function CheckIncidence check and reformat the matrix of incidence.

Usage

CheckIncidence(Incidence, IncidenceName)

Arguments

Value

The matrix of incidence in the right format.

Compute the statistic of the contrast test.

Description

The function ContrastStatTest compute the statistic of the contrast test.

Usage

ContrastStatTest(Incidence, Contrast = NULL, Zmat, MatCorr, IncidenceName)

Arguments

Value

A dataset of two columns containing the pvalue of the test of contrast and the minimum number of environment per group of all markers.

Compute the statistic of the contrast test in presence of missing values

Description

The function ContrastStatTest compute the statistic of the contrast test.

Usage

ContrastStatTest.NA(
  Incidence,
  Contrast = NULL,
  Zmat,
  MatCorr,
  Data,
  Configs.list,
  IncidenceName
)

Arguments

Value

A dataset of two columns containing the pvalue of the test of contrast and the minimum number of environment per group of all markers.

FastKerFdr

Description

Computes H1 posteriors of the Z-scores.

Usage

FastKerFdr(
  Z,
  p0,
  plotting = FALSE,
  NbKnot = 1e+05,
  tol = 1e-05,
  max_iter = 10000
)

Arguments

Value

tau is the vector of H1 posteriors

GetH0Items

Description

This function give the index of the markers which seems not significant (under H0)

Usage

GetH0Items(Zmat, Threshold = 0.8, plotting = FALSE, Cores = NULL)

Arguments

Value

A vector of index of markers which seems not significant (under H0)

LLikelihoodT_vect

Description

This function compute the values of loglikelihood for all markers.

Usage

LLikelihoodT_vect(Zmat, Delta, P, Mu, Tau)

Arguments

Value

A vector containing the value of the Log-Likelihood of all markers

Drawing a QQplot

Description

The function MakeQQplot displays the QQplot of the -log10(pvalues).

Usage

MakeQQplot(Pvalues, Name = NULL, Xrange = NULL, Yrange = NULL)

Arguments

ReadData

Description

This function read the one file, select interesting columns and rename them.

Usage

ReadData(ListN, FileN, VarN, MinFreq = 0)

Arguments

Value

A tibble with the interesting columns selected and renamed.

Compute the pvalue of the meta-regression test.

Description

The function RegressionStatTest compute the statistic and the pvalue of the regression test.

Usage

RegressionStatTest(Covariate, CovName, Zmat, MatCorr)

Arguments

Value

A dataset of two columns containing the pvalue of the meta-regression test and the number of environment used to perform the test of all markers.

Compute the pvalue of the regression test in presence of missing values.

Description

The function RegressionStatTest compute the statistic and the pvalue of the regression test.

Usage

RegressionStatTestNA(Covariate, CovName, Zmat, MatCorr, Data, Configs.list)

Arguments

Value

A dataset of two columns containing the pvalue of the meta-regression test and the number of environment used to perform the test of all markers.

Computation of the autocorrelation

Description

The function autocor computes the autocorrelation.(function from localscore)

Usage

autocor(x)

Arguments

Value

the autocorrelation.

Description of the environments.

Description

A dataset containing variables describing the 22 environments.

Usage

envDesc

Format

A data frame with 22 rows and 3 variables:

• FileName: environment name

• Temp: temperature

• Water: water condition

Computation of the lindley process from scores.

Description

The function lindley computes the lindley process from scores.(function from localscore)

Usage

lindley(scores)

Arguments

Value

the lindley.

Results of different GWAS.

Description

A dataset containing the results of 10 different genetic association studies testing the association between a set of 25,436 markers and the grain yield. The data are extrated from: Drops Amaizing available on the https://doi.org/10.15454/6TL2N4 website. This dataset were obtained thanks to the metaGE.collect function.

Usage

metaData

Format

A data frame with 25,436 rows and 35 variables:

• CHR: chromosome of the marker

• POS: position of the marker

• MARKER: name of the marker

• FREQ.env: maf of the marker in the environment env

• EFFECT.env: regression coefficient of the marker in the environment env

• EFFECT_SE.env: standard error of the regression coefficient of the marker in the environment env

• PVAL.env: pvalue of the marker in the environment env

• WEIGHT.env: weight of the marker in the environment env

• ALLELE0: allele0

• ALLELE1: allele1

Collect the results of GWAS data from different files

Description

This function merges files containing the summary statistics of GWAS in different environments (one file per environment).

Usage

metaGE.collect(
  FileNames,
  VariableNames,
  MinFreq = 0,
  DropDuplicates = TRUE,
  Verbose = FALSE,
  NA.rmv = TRUE
)

Arguments

Details

Each file MUST contain the variables below:

• MARKER: the marker name

• CHR: the chromosome

• POS: the position of the marker

• EFFECT: the mean effect of the marker

• PVAL: the pvalue

Each file might contain the variables:

• FREQ: MAF

• ALLELE0: Allele coding for allele 0

• ALLELE1: Allele coding for allele 1

Value

A list with the following elements:

Examples

require(dplyr)
require(tibble)
require(stringr)
RepData <- system.file("extdata", package = "metaGE")
# Get the complete list of association files
File.list <- list.files(RepData ,full.names = TRUE) %>%
            tibble(Names = .) %>%
            mutate(ShortNames = Names %>%
                  str_remove(pattern = paste0(RepData,"/")) %>%
                  str_remove(pattern = "_DF.txt"))  %>%
            select(ShortNames,Names) %>%
            deframe
###Build the dataset
## First provide the list of variable names
Names.list <- list(MARKER="Marker_Name",
                  CHR="Chromosome",
                  POS="Marker_Position",
                  FREQ="Maf",
                  EFFECT="SNP_Weight",
                  PVAL="Pvalue",
                  ALLELE0="Allele1",
                  ALLELE1="Allele2")

MinFreq <- 0.07

## Now collect
metaData <- metaGE.collect(File.list, Names.list,MinFreq = MinFreq)

Infer inter-environment correlation matrix

Description

This function infer the inter-environment correlation matrix from the z-scores after filtering markers with high probability of being under H1.

Usage

metaGE.cor(Data, Threshold = 0.6, NA.omit = TRUE, Cores = NULL)

Arguments

Value

The inter-environment correlation matrix

Examples

require(corrplot)
data("metaData")
Threshold <- 0.8
matCorr <- metaGE.cor(metaData, Threshold = Threshold)
corrplot(matCorr,order = "hclust")

Meta-analysis procedure: Fixed or Random effect.

Description

Quantitative trait loci detection via Fixed or Random effect meta-analysis GWAS procedure.

Usage

metaGE.fit(Data, MatCorr, Method, NA.omit = TRUE, DropZScores = FALSE)

Arguments

Details

Different tests may be performed:

• Fixed Effect (Fe), to identify markers with a stable effect across environments.

• Random Effect (Re), to identify markers whose effects may be unstable across environments.

Value

The dataset Data with supplementary columns:

• PVALUE: The pvalue of the MA test,

• Mu: Estimate of the mean marker effect,

• Tau: Estimate of the variance of the marker effect, for the Random model only,

• the Zscores for each environment if DropZScores = FALSE.

Examples

require(dplyr)
# Import the data
data("metaData")

# Compute the inter-environment correlation matrix
matCorr <- metaGE.cor(metaData, Threshold = 0.8)

# Fixed Effect
FeDF <- metaGE.fit(metaData, matCorr, Method = "Fe")
head(FeDF %>% select(CHR, POS, MARKER, Mu, Tau, PVALUE))

# Random Effect
ReDF <- metaGE.fit(metaData, matCorr, Method = "Re")
head(ReDF %>% select(CHR, POS, MARKER, Mu, Tau, PVALUE))

Draw the heatmap to see markers effects across environments.

Description

The function metaGE.heatmap displays the heatmap of the zscores, the estimated marker effects or the pvalues of each markers (in rows) in each environments (in columns).

Usage

metaGE.heatmap(
  Data,
  Prefix = "Z.",
  EnvGroups = NULL,
  QTLsVarName = NULL,
  RowOrder = TRUE,
  ColOrder = TRUE,
  ShowDendrogram = FALSE,
  Colors = c("red", "black", "green"),
  Main = ""
)

Arguments

Value

The heatmap

Examples

require(dplyr)
# Import the data
data("metaData")

# Compute the inter-environment correlation matrix
matCorr <- metaGE.cor(metaData, Threshold = 0.8)

# Fit the Fixed Effect model
FeDF <- metaGE.fit(metaData, matCorr, Method = "Fe")

# Control the FDR (here Benjamini-Hochberg)
Alpha <- 0.05
Signif <-  FeDF$PVALUE %>% p.adjust(method = "BH") %>% `<`(Alpha) %>% which

# Draw the z-scores heatmap of the significant markers
heatmap <- metaGE.heatmap(Data = FeDF[Signif,],
                         Prefix = "Z.")
                        

Create the matrix of incidence.

Description

The function metaGE.incidence convert categorical variable describing the environments into a matrix of dummy variables with in rows the levels of the variables and in columns the environment.

Usage

metaGE.incidence(VarName, Covariate, EnvName, Data, AtLeast = 1)

Arguments

Details

The names of the environment must be the same as used in the Data dataset.

Value

A binary matrix containing indicator variables with in rows the levels of the variables and in columns the environment.

Examples

# Import the data
data("metaData")
data("envDesc")

# Build the matrix of incidence
(Incidence.Temp <- metaGE.incidence(VarName = "Temp",Covariate = envDesc,
                                    EnvName = "ShortName", Data = metaData))

Compute the local score from a set of pvalues.

Description

The function metaGE.lscore computes the local score and the significant regions from a set of pvalues.

Usage

metaGE.lscore(Data, PvalName, xi)

Arguments

Details

This function is directly inherited from the scorelocalfunctions.R R code file of Fariello MI, Boitard S, Mercier S, et al., as available on the https://forge-dga.jouy.inra.fr/projects/local-score website. The technical details of the computation can be found in Fariello MI, Boitard S, Mercier S, et al. Accounting for linkage disequilibrium in genome scans for selection without individual genotypes: The local score approach. doi:10.1111/mec.14141. The function computes a local score for the detection of significant regions based on the hypothesis that the H0 distribution of the pvalues is uniform. Under this hypothesis the local score follows a Gumbel distribution (under H0) whose parameters depend on the threshold xi and on the autocorrelation between pvalues within each chromosome. The threshold has to be selected in 1,2,3,4 and the autocorrelation is computed internally.

Value

A list with the following elements:

Examples

require(dplyr)
# Import the data
data("metaData")

# Compute the inter-environment correlation matrix
matCorr <- metaGE.cor(metaData, Threshold = 0.8)

# Fit the Fixed Effect model
FeDF <- metaGE.fit(metaData, matCorr, Method = "Fe")

# Compute the score local
xi <- 2
FeScore <- metaGE.lscore(FeDF,"PVALUE", xi)
#FeScore$SigZones

Draw the Manhattan plot.

Description

The function metaGE.manhattan displays the Manhattan plot of the -log10(p-value) or the local score of each marker along the genome.

Usage

metaGE.manhattan(
  Data,
  VarName,
  Threshold = NULL,
  SigZones = NULL,
  Score = FALSE,
  AnnotateMarkers = NULL,
  Main = "",
  col = c("grey", "black"),
  colSigZones = "blue",
  Ylim = NULL
)

Arguments

Value

The Manhattan plot.

Examples

require(dplyr)
# Import the data
data("metaData")

# Compute the inter-environment correlation matrix
matCorr <- metaGE.cor(metaData, Threshold = 0.8)

# Fit the Fixed Effect model
FeDF <- metaGE.fit(metaData, matCorr, Method = "Fe")

# Control the FDR (here Benjamini-Hochberg)
Alpha <- 0.05
Signif <-  FeDF$PVALUE %>% p.adjust(method = "BH") %>% `<`(Alpha) %>% which

# Draw the corresponding manhattan plot
PvalThresholdFe <- FeDF[Signif,]$PVALUE%>% max %>% max(.,0)
manhattan_pval <- metaGE.manhattan(Data = FeDF,VarName = 'PVALUE',
                             Threshold = PvalThresholdFe,
                              Main = '-log10(Pval) alongside the chromosome Fe method')


# Compute the score local
xi <- 2
FeScore <- metaGE.lscore(FeDF,"PVALUE", xi)

# Draw the corresponding manhattan plot
manhattan_lscore <- metaGE.manhattan(Data = FeScore$Data,VarName = 'SCORE',
                                     SigZones = FeScore$SigZones, Score = TRUE,
                                     Main = 'Local score alongside the chromosome Fe method')

Display visual checks of pvalues.

Description

The function metaGE.pvalplot displays the pvalue distribution and the QQplot of the -log10(pvalues).

Usage

metaGE.pvalplot(Pvalues, Main = "")

Arguments

Value

No return value, the plot is displayed in the active graphics window.

Examples

# Import the data
data("metaData")

# Compute the inter-environment correlation matrix
matCorr <- metaGE.cor(metaData, Threshold = 0.8)

# Fit the Fixed Effect model
FeDF <- metaGE.fit(metaData, matCorr, Method = "Fe")

# Check the pvalues
metaGE.pvalplot(Pvalues = FeDF$PVALUE, Main= "Pvalue Fe")

Plot the z-score of a marker according to a covariate.

Description

The function metaGE.regplot displays the graph of the z-scores of a marker according to a covariate.

Usage

metaGE.regplot(
  Data,
  Covariate,
  EnvName,
  MarkerName,
  VarName,
  Zscore = FALSE,
  aesCol = NULL,
  Main = ""
)

Arguments

Value

The plot

Examples

data("metaData")
data("envDesc")
metaGE.regplot(Data = metaData, Covariate = envDesc, EnvName = "ShortName",
               MarkerName = "AX-91369217", VarName = "Tnight.mean", aesCol = "Classification")

Meta-analysis test for Genotype x Environment interactions: Contrast or Regression.

Description

The function metaGE.test compute meta-analysis contrast or regression test.

Usage

metaGE.test(
  Data,
  MatCorr,
  Incidence = NULL,
  Contrast = NULL,
  Covariate = NULL,
  EnvName = NULL,
  NA.omit = TRUE,
  DropZScores = FALSE
)

Arguments

Details

If Incidence is provided, the function will perform all the corresponding tests of contrast. If Covariate is provided, the function will perform all the corresponding meta-regression tests. The Contrast can be NULL, in this case the identity matrix is used.

Value

The dataset Data with supplementary columns containing the PVALUE of each test performed.

Examples

require(dplyr)

 # Import the data
data("metaData")
data("envDesc")

# Compute the inter-environment correlation matrix
matCorr <- metaGE.cor(metaData, Threshold = 0.8)

#### Contrast test
# Build the matrix of incidence
Incidence.Water <- metaGE.incidence(VarName = "Water",Covariate = envDesc,
                                    EnvName = "ShortName", Data = metaData)

# Perform the contrast test
ContrastDF <- metaGE.test(metaData, matCorr,Incidence = Incidence.Water, Contrast = NULL)
head(ContrastDF %>% select(CHR, POS, MARKER, PVALUE.Contrast1))

#### Regression test
RegressionDF <- metaGE.test(metaData,matCorr, Covariate = envDesc[,c(1,5)],EnvName = "ShortName" )
head(RegressionDF %>% select(CHR, POS, MARKER, PVALUE.Tnight.mean))

Computation of the significative regions

Description

The function sig_sl computes the significative regions from a lindley process given a significance threshold.(function from localscore)

Usage

sig_sl(lind, pos, th)

Arguments

Value

the significance threshold.

Computation of the significance threshold

Description

The function thresUnif computes the significance threshold.(function from localscore)

Usage

thresUnif(L, cor, xi, alpha = 0.05)

Arguments

Details

The distribution of the p-values is uniform, the local score follows a Gumbel distribution under the null.

Value

the significance threshold.