metsModels: Wrapper function that will run association, interaction, or...
In buddha2490/BERGMets: Metabolomics data and analysis in the BERG

Description Usage Arguments Details Value Author(s) See Also Examples

metsModels should do everything you need it to do. If you include a intvar= variable, the function will run interaction or stratified models (depending on intvar class, see details). If the intvar= argument is left NULL, simple association models will be returned.

1 2	metsModels(dat, biochem, outcome, compid,intvar=NULL, covariates=NULL, normalize=T)

`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
`intvar)`	Optional Name of interaction variable. If left NULL, a simple association analysis will be returned. If class(intvar)=="numeric", simple interaction models will be fit coded as metabolite*intvar. If class(intvar)=="factor", the function will return estimates for the metabolites stratified by the intvar.
`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

the metsModels() function is a wrapper function that runs either metsAssocation() for simple association models, or metsInteraction() for interaction/stratified models.

A simple association model includes either linear or logistic regression (depending on the outcome), adjusted for covariates. If the user includes an intvar= variable, the interaction/stratified models will be returned, depending on the intvar class. See documentation files for metsAssociation() and metsInteraction() for details of either sets of models

A data frame containing the results. Results are merged with the biochem-metadata file and estimates/pvalues are nicely formatted.

Simple association models will return estimates, p-values and metabolite metadata

Stratified models will return estimates across each level of the interaction variables.

Interaction models will return only the estimate and p-values associated with the interaction term.

Both will return an overall p-value for the interaction.

Brian Carter

metsAssoc, metsInt, normalizeMets

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")]

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



# Stratified analysis (linear regression)
     metsInt(dat=df,
               biochem=biochem,
               outcome="BMI",
               compid=comp.id,
               intvar="AGE_INT",
               covariates=covars,
               normalize=T)

{
  }