P-value adjustment in MTX_model

Downstream analysis and statistics MaAsLin
1

Hi all,

I use MTX_model package 1.2.4 for differential expression analysis of MTX paired with MGX data in R. I use one fixed effect, one random effect and MGX as covariate. However, I have noted that the qvalues for the fixed effect effect on features differ between the R model on one hand and the files all_results.tsv and significant_results.tsv on the other hand. Can you help me to understand ? Is it possible, the adjustment of p-value is done only on fixed effect in R but done for all p-values (including covariate ones) when files are exported ?

My code looks like this :

MTXmodel(
unstrat_mtx_uniref90[,3:length(unstrat_mtx_uniref90)], mtd, ‘output’,
min_prevalence = 0.2,
fixed_effects = c(‘treatment’),
random_effects = c(‘grp’),
normalization = ‘NONE’,
transform = “LOG”,
standardize = FALSE,
input_dnadata = unstrat_mgx_uniref90[,3:length(unstrat_mgx_uniref90)]
)

Many thanks :slight_smile:

2

Hi there,

Thanks for your interest. There are two types of qvalues reported by MTX_model:

  1. Qvaues that adjust all pvalues from all covariates except of MGX covariate (i.e. only fixed effects that user customize) are exported to all_results.tsv and significant_results.tsv.
  2. Qvalues that adjust all pvalues from all covariates (both MGX covariate and provided fixed effects) are exported to all_results.tsv.tmp and significant_results.tsv.tmp.

Thanks!
Yancong

1 Like