metsAssoc: Basic metabolomics association analysis.

In buddha2490/BERGMets: Metabolomics data and analysis in the BERG

Description

This function will run a basic association analysis adjusted for covariates for a given list of metabolites. Depending on the outcome, the function will choose a linear or logistic regression model. It will calculate the associations between outcome and each metabolite, merge with the metadata files, and return an organized table.

Usage

metsAssoc(dat, biochem, outcome, metabolites, covariates=NULL, normalize=T)

Arguments

dat	data frame containing your metabolites and covariates
biochem	The biochemical metadata file. Typically this is in the $biochem element of the metabolites list. You can subset this data frame any way you wish, some variables are probably unneeded. Recommend that you subset to only include COMP_ID, BIOCHEMICAL, SUPER_PATHWAY, SUB_PATHWAY.
outcome	Outcome for the analysis. Logistic regression must be coded [0,1]; linear regression must be continuous numeric. Function will identify which is which and will provide an error message if you code it wrong.
compid	Vector of COMP_IDs used for the models
covariates	Optional. Vector of covariate names from the dat data frame.
normalize	Logical. Set to TRUE if you haven't already mean-centered and glog transformed your metabolite data. Default is TRUE.

Details

This function will run simple association models (either linear or logistic regressions). I recommend that you use the metsModels() function which is a wrapper for all sorts of metabolomics models. It will either run metsAssoc() or metsInt() depending on the inputs.

Value

Data frame organizing the associations and p-values for all metabolites.

Author(s)

Brian Carter

See Also

normalizeMets, metsInt, metsModels,

Examples

breast_metabolomics <- getMetabolites("breast_metabolomics")
data(survey)
df <- left_join(survey,breast_metabolomics$metabolites,"ID")
covars<- c("AGE_INT","LASTATE")
comp.id <- names(breast_metabolomics$metabolites)[-1]
biochem <- breast_metabolomics$biochem[,c("COMP_ID","BIOCHEMICAL")]

# Linear regression models
out <- metsAssoc(dat=df,
                       biochem=biochem,
                       outcome="BMI",
                       compid=comp.id[1:10],
                       covariates=covars,
                       normalize=T)  # normalizing data

# Logistic regression models
out <- metsAssoc(dat=df,
                       biochem=biochem,
                       outcome="BMIBIN",
                       compid=comp.id[1:10],
                       covariates=covars,
                       normalize=T)  # normalizing data

buddha2490/BERGMets documentation built on Sept. 6, 2020, 5:11 p.m.