Skip to contents

GXwas: Running genome-wide association study (GWAS) and X-chromosome-wide association study (XWAS) models.

Source: R/GXwasR_main_functions.R
GXwas.Rd

This function runs GWAS models in autosomes with several alternative XWAS models. Models such as "FMcombx01","FMcombx02",and "FMstratified" can be applied to both binary and quantitative traits, while "GWAcxci" can only be applied to a binary trait.

For binary and quantitative features, this function uses logistic and linear regression, allowing for multiple covariates and the interactions with those covariates in a multiple-regression approach. These models are all run using the additive effects of SNPs, and each additional minor allele's influence is represented by the direction of the regression coefficient.

This function attempts to identify the multi-collinearity among predictors by displaying NA for the test statistic and a p-value for all terms in the model. The more terms you add, the more likely you are to run into issues.

For details about the different XWAS model, please follow the associated publication.

Usage

GXwas(
  DataDir,
  ResultDir,
  finput,
  trait = c("binary", "quantitative"),
  standard_beta = TRUE,
  xmodel = c("FMcombx01", "FMcombx02", "FMstratified", "GWAScxci"),
  sex = FALSE,
  xsex = FALSE,
  covarfile = NULL,
  interaction = FALSE,
  covartest = c("ALL"),
  Inphenocov = c("ALL"),
  combtest = c("fisher.method", "fisher.method.perm", "stouffer.method"),
  MF.zero.sub = 1e-05,
  B = 10000,
  MF.mc.cores = 1,
  MF.na.rm = FALSE,
  MF.p.corr = "none",
  plot.jpeg = FALSE,
  plotname = "GXwas.plot",
  snp_pval = 1e-08,
  annotateTopSnp = FALSE,
  suggestiveline = 5,
  genomewideline = 7.3,
  ncores = 0
)

Arguments

DataDir

Character string for the file path of the input PLINK binary files.

ResultDir

Character string for the folder path where the outputs will be saved.

finput

Character string, specifying the prefix of the input PLINK binary files with both male and female samples. This file needs to be in DataDir.

Note: Case/control phenotypes are expected to be encoded as 1=unaffected (control), 2=affected (case); 0 is accepted as an alternate missing value encoding. The missing case/control or quantitative phenotypes are expected to be encoded as 'NA'/'nan' (any capitalization) or -9.

trait

Boolean value, 'binary' or 'quantitative' for the phenotype i.e. the trait.

standard_beta

Boolean value, TRUE or FALSE in case of quantitative trait for standardizing the trait or phenotype values (mean 0, unit variance), so the resulting coefficients will be standardized. The default is TRUE.

xmodel

Models "FMcombx01","FMcombx02",and "FMstratified" can be chosen for both binary and quantitative traits while "GWAcxci" can only apply to the binary trait. These models take care of the X-chromosomal marker. Three female genotypes are coded by 0, 1, and 2 in FM01 and FM02. The two genotypes of males that follow the X-chromosome inactivation (XCI) pattern as random (XCI-R) in the FM01 model are coded by 0 and 1, while the two genotypes that follow the XCI is escaped (XCI-E) in the FM02 model are coded by 0 and 1. To reflect the dose compensation connection between the sexes, FM02 treats men as homozygous females.

In the "FMstratified" associations are tested separately for males and females, and then the combined p values are computed the Fisher's method, Fisher's method with permutation, or Stouffer's method(1,3-7]. An X-chromosome inactivation (XCI) pattern, or coding technique for X-chromosomal genotypes between sexes, is not required for the XCGA. By simultaneously accounting for four distinct XCI patterns, namely XCI-R, XCI-E, XCI-SN (XCI fully toward normal allele), and XCI-SR (XCI fully toward risk allele), this model may maintain a respectably high power (Su et al. 2022) .

Note: sex shouldn't be provided as a covariate in the XCGA model.

sex

Boolean value, TRUE or FALSE for using sex as covariate in association test. It is applicable genome-wide.

The default is FALSE.

xsex

Boolean value, TRUE or FALSE for using sex as covariate in association test for X-chromosomal SNPs. The default is FALSE. This will overwrite 'sex' argument for X-chromosome.

covarfile

Character string for the full name of the covariate file in .txt format. This file should be placed in DataDir.

Note about the covariate file: The first column of this file should be FID, the second column should be IID and the other columns should be covariates. The primary header line should be there starting with “FID”, and “IID” followed by covariate names. If an individual is not present in the covariate file, or if the individual has a missing phenotype value (i.e. -9 by default) for the covariate, then that individual is set to missing (i.e. will be excluded from association analysis). It is important to note that for stratified GWAS model, if PCs are included as covar then it should be generated separately for each cohort and then included in the covarfile. Use the function DummyCovar to generate a new covariate file with categorical variables down-coded as binary dummy variables for the covariate file with categorical variables. For instance, if a variable has K categories, K-1 new dummy variables are constructed, and the original covariate is now estimated with a coefficient for each category.

interaction

Boolean value, TRUE or FALSE for including SNP x covariate interaction term/terms from the association analysis. The default is FALSE. If a permutation procedure is chosen then the interaction will be automatically FALSE. For the interaction with the two covariates COV1 and COV2, the model will look like: \(Y = b0 + b1.ADD + b2.COV1 + b3.COV2 + b4.ADD x COV1 + b5.ADD x COV2 + e\). When interaction factors are incorporated into the model, the main effects' significance is not always determined simply; rather, it depends on the arbitrary coding of the variables. To put it another way, you should probably just interpret the p-value for the interaction. Also, The p-values for the covariates do not represent the test for the SNP-phenotype association after controlling for the covariate. That is the first row (ADD). Rather, the covariate term is the test associated with the covariate-phenotype association. These p-values might be extremely significant (e.g. if one covaries for smoking in an analysis of heart disease, etc) but this does not mean that the SNP has a highly significant effect necessarily. Note that, this feature is not valid for XCGA model for XWAS part.

covartest

Vector value with NULL,"ALL" or covariate name/names to be included in the test. The default is NULL. For instance, the user can choose “AGE” and “SEX” as covartest = c(“AGE”, “SEX”) or all the covariates as covartest = c(“ALL”).

Inphenocov

Vector of integer values starting from 1 to extract the terms which user wants from the above model: \(Y = b0 + b1.ADD + b2.COV1 + b3.COV2 + b4.ADDxCOV1 + b5.ADDxCOV2 + e\). The terms will appear in order as (Purcell et al. 2007) for ADD, (Su et al. 2022) for COV1, (Rhodes 2002) for ADD x COV1, and (Moreau and others 2003) for ADD x COV2. If the user wants to extract the terms for COV1 and ADD x COV1, they need to specify it as c(2,4). The default is c(“ALL”).

Note: This feature is not valid for the XCGA model for the XWAS part.

combtest

Character vector specifying method for combining p-values after stratified GWAS/XWAS models. Choices are “stouffer.method”, "fisher.method" and "fisher.method.perm". For fisher.method the function for combining p-values uses a statistic, \(S = -2 ∑^k /log p\), which follows a \(χ^2\) distribution with 2k degrees of freedom (Fisher 1925) .

For fisher.method.perm, using p-values from stratified tests, the summary statistic for combining p-values is \(S = -2 ∑ /log p\). A p-value for this statistic can be derived by randomly generating summary statistics (Rhodes 2002) . Therefore, a p-value is randomly sampled from each contributing study, and a random statistic is calculated. The fraction of random statistics greater or equal to S then gives the final p-value.

For stouffer.method ,the function applies Stouffer’s method (Stouffer et al. 1949) to the p-values assuming that the p-values to be combined are independent. Letting p1, p2, . . . , pk denote the individual (one- or two-sided) p-values of the k hypothesis tests to be combined, the test statistic is then computed with \($z = ∑^k_{1}frac{z_{i}}{sqrt(k)}$\) where \($z_{i}$ = Φ−1 (1 – $p_{i}$)\) and \(Φ −1 (·)\) denotes the inverse of the cumulative distribution function of a standard normal distribution. Under the joint null hypothesis, the test statistic follows a standard normal distribution which is used to compute the combined p-value. This functionality is taken from the R package poolr (Cinar and Viechtbauer 2022) .

Note that only p-values between 0 and 1 are allowed to be passed to these methods.

Note: Though this parameter is enabled for both autosome GWAS and XWAS, the combining pvalue after sex-stratified test is recommended to ChrX only.

MF.zero.sub

Small numeric value for substituting p-values of 0 in in stratified GWAS with FM01comb and FM02comb XWAS models. The default is 0.00001. As log(0) results in Inf this replaces p-values of 0 by default with a small float.

B

Integer value specifying the number of permutation in case of using fisher.method.perm method in stratified GWAS with FM01comb and FM02comb XWAS models. The default is 10000.

MF.mc.cores

Number of cores used for fisher.method.perm in stratified GWAS with FM01comb and FM02comb XWAS models.

MF.na.rm

Boolean value, TRUE or FALSE for removing p-values of NA in stratified GWAS with FM01comb and FM02comb XWAS in case of using Fisher’s and Stouffer’s methods. The default is FALSE.

MF.p.corr

Character vector specifying method for correcting the summary p-values for FMfcomb and FMscomb models. Choices are "bonferroni", "BH" and "none" for Bonferroni, Benjamini-Hochberg and none, respectively. The default is "none".

plot.jpeg

Boolean value, TRUE or FALSE for saving the plots in .jpeg file. The default is TRUE.

plotname

A character string specifying the prefix of the file for plots. This file will be saved in DataDir. The default is "GXwas.plot".

snp_pval

Numeric value as p-value threshold for annotation. SNPs below this p-value will be annotated on the plot. The default is 1e-08.

annotateTopSnp

Boolean value, TRUE or 'FALSE. If TRUE, it only annotates the top hit on each chromosome that is below the snp_pval threshold. The default is FALSE.

suggestiveline

The default is 5 (for p-value 1e-05).

genomewideline

The default is 7.3 (for p-value 5e-08).

ncores

Integer value, specifying the number of cores for parallel processing. The default is 0 (no parallel computation).

Value

A dataframe with GWAS (with XWAS for X-chromosomal variants) along with Manhattan and Q-Q plots. In the case of the stratified test, the return is a list containing three dataframes, namely, FWAS, MWAS, and MFWAS with association results in only female, only male, and both cohorts, respectively. This will be accompanied by Miami and Q-Q plots. The individual manhattan and Q-Q-plots for stratified tests prefixed with xmodel type will be in the DataDir.

References

Cinar O, Viechtbauer W (2022). “The poolr package for combining independent and dependent p values.” Journal of Statistical Software, 101(1), 1–42. doi:10.18637/jss.v101.i01 .

Fisher RA (1925). Statistical Methods for Research Workers. Oliver and Boyd, Edinburgh.

Moreau Y, others (2003). “Comparison and meta-analysis of microarray data: from the bench to the computer desk.” Trends in Genetics, 19(10), 570–577.

Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D, Maller J, Sklar P, de Bakker PIW, Daly MJ, others (2007). “PLINK: A Tool Set for Whole-Genome Association and Population-Based Linkage Analyses.” The American Journal of Human Genetics, 81(3), 559–575. doi:10.1086/519795 .

Rhodes DR (2002). “Meta-analysis of microarrays: interstudy validation of gene expression profiles reveals pathway dysregulation in prostate cancer.” Cancer Research, 62(15), 4427–4433.

Stouffer SA, Suchman EA, DeVinney LC, Star SA, Williams RM (1949). The American Soldier: Adjustment During Army Life, volume 1. Princeton University Press, Princeton, NJ.

Su Y, Hu J, Yin P, Jiang H, Chen S, Dai M, Chen Z, Wang P (2022). “XCMAX4: A Robust X Chromosomal Genetic Association Test Accounting for Covariates.” Genes (Basel), 13(5), 847. doi:10.3390/genes13050847 , http://www.ncbi.nlm.nih.gov/pubmed/35627231.

Author

Banabithi Bose

Examples


DataDir <- GXwasR:::GXwasR_data()
ResultDir <- tempdir()
finput <- "GXwasR_example"
standard_beta <- TRUE
xsex <- FALSE
sex <- TRUE
Inphenocov <- NULL
covartest <- NULL
interaction <- FALSE
MF.na.rm <- FALSE
B <- 10000
MF.zero.sub <- 0.00001
trait <- "binary"
xmodel <- "FMcombx02"
combtest <- "fisher.method"
snp_pval <- 1e-08
covarfile <- NULL
ncores <- 0
MF.mc.cores <- 1
ResultGXwas <- GXwas(
    DataDir = DataDir, ResultDir = ResultDir,
    finput = finput, xmodel = xmodel, trait = trait, covarfile = covarfile,
    sex = sex, xsex = xsex, combtest = combtest, MF.p.corr = "none",
    snp_pval = snp_pval, plot.jpeg = TRUE, suggestiveline = 5, genomewideline = 7.3,
    MF.mc.cores = 1, ncores = ncores
)
#> • Running FMcombx02 model