taxa.compare {metamicrobiomeR}

Compare taxa relative abundance

Description

This function compares taxa relative abundance summary tables at all levels between groups using GAMLSS with BEZI or Linear/Linear Mixed Effect models (LM/LMEM) after filtering (using prevalence and relative abundance thresholds).

Usage

taxa.compare(
  taxtab,
  propmed.rel = "gamlss",
  transform = "none",
  zeroreplace.method = "none",
  comvar,
  adjustvar,
  personid = "personid",
  longitudinal = "yes",
  percent.filter = 0.05,
  relabund.filter = 5e-05,
  p.adjust.method = "fdr"
)

Arguments

taxtab

taxa relative abundance table (already merged to mapping file) from phylum to species or any preferred highest taxa level.

propmed.rel

statistical method for comparing relative abundance. Options are "lm" for LM/LMEM or "gamlss" for GAMLSS with BEZI family.

transform

transformation of relative abundance data. Options are "none" for no transformation, "gmpr" for Geometric Mean of Pairwise Ratios (GMPR) normalization, "asin.sqrt" for arcsine transformation, "logit" for logit transformation, "clr" for centered log ratio transformation. Default is "none".

zeroreplace.method

Method for zero replacement implemented in R package *zCompositions*. Options are "none" for no replacement, "multKM" for Multiplicative Kaplan-Meier smoothing spline (KMSS) replacement, "multLN" for Multiplicative lognormal replacement, "multRepl" for Multiplicative simple replacement, "lrEM" for Log-ratio EM algorithm, "lrDA" for Log-ratio DA algorithm. Default is "none".

comvar

main variable for comparison

adjustvar

variables to be adjusted.

personid

name of variable for person id (applicable for longitudinal data)

longitudinal

whether data is longitudinal? Options are "yes" or "no". Default is "yes".

percent.filter

prevalence threshold (the percentage of number of samples the taxa/pathway available). Default is 0.05.

relabund.filter

relative abundance threshold (the minimum of the average relative abundance for a taxa/pathway to be retained). Default is 0.00005.

p.adjust.method

method for multiple testing adjustment. Options are those of the p.adjust.methods of stats:: p.adjust function. Default for this function is "fdr".

Value

matrice of coefficients, standard errors, p-values and multiple testing adjusted p-values of all variables in the models.

Examples

#Load summary tables of bacterial taxa relative abundance from Bangladesh data
data(taxtab6)
tab6<-as.data.frame(taxtab6)
tl<-colnames(taxtab6)[grep("k__bacteria.p__fusobacteria",colnames(taxtab6))]
taxacom.ex<-taxa.compare(taxtab=tab6[,c("personid","x.sampleid","bf","age.sample",tl)],
propmed.rel="gamlss",comvar="bf",adjustvar="age.sample",
longitudinal="yes",p.adjust.method="fdr")