Metabolomics meta-regression

Perform meta-regression for repeated measurements. Return merged timewise and training differential analysis summary statistics, where results for features with multiple measurements are replaced with the meta-regression results when appropriate. This method was used to generate the MotrpacRatTraining6moData::METAB_DA_METAREG differential analysis results, which are the version of results used in the manuscript analyses.

Usage

metab_meta_regression(
  tissue,
  timewise_input = NULL,
  training_input = NULL,
  het_p_threshold = 0.001
)

Arguments

tissue

character, tissue abbreviation, one of MotrpacRatTraining6moData::TISSUE_ABBREV

timewise_input

r data frame, custom input. To see the expected format, look at a table returned by load_metab_da() with type="timewise".

training_input

r data frame, custom input. To see the expected format, look at a table returned by load_metab_da() with type="training".

het_p_threshold

numeric, meta-regression cases with a heterogeneity p-value below this are considered to have high heterogeneity. Default: 0.001

Value

named list where meta_reg_timewise_dea is a data frame with the adjusted timewise results, training_meta_regression is a data frame with the adjusted training results, meta_regression_results is a named list with meta-regression results per redundant metabolite, meta_regression_models is a table of the number of each type of model used for meta-regression, and metabolite_categories is a named list of the RefMet IDs of metabolites corresponding to each category described in the details.

Details

We try multiple models per repeated analyte:

• Model 1:Two random effects factors if platform and is_targeted are not redundant. Default optimization.

• Model 2:Two random effects factors if platform and is_targeted are not redundant with alternative optimization.

• Model 3:platform and is_targeted are redundant. Use a single RE factor with default optimization. Also, use this if QMp is NA, which is an indication of over-parameterization of the model.

• Model 4:platform and is_targeted are redundant. Use a single RE factor with alternative optimization.

• Model 5:If all previous analyses failed, use a simple fixed-effects approach.

After performing meta-regression, we define four categories of metabolites:

1. measured once

2. measured multiple times, high heterogeneity, has a targeted platform

3. measured multiple times, high heterogeneity, no targeted platform

4. measured multiple times, low heterogeneity

For categories 1 and 3 we keep the results as is. For category 2 we take the targeted data only. Finally, for category 4 we take the meta-regression results.

See also

load_metab_da(), metabolite_meta_regression(), MotrpacRatTraining6moData::METAB_DA_METAREG

Examples

# Perform meta-regression for gastrocnemius
res = metab_meta_regression("SKM-GN")
#> Warning: SKM-GN differential analysis results have ref standard results. Excluding them from meta-analysis.
#> Performing meta-regression for 117 redundant metabolites in SKM-GN...
#> Done.
#> Number of models that were fit for SKM-GN:

#>                                                                 metareg_nplatform
#> metareg_calls                                                     2  3  4  5
#>   list(~analysis_group | platform, ~analysis_group | istargeted)  0 22  8  5
#>   x_subset                                                        2  5  1  0
#>   FE                                                             74  0  0  0
#> Total number of cases with high heterogeneity: 6
#> Total number of cases with low heterogeneity: 111
#> 
#> Summary of the number of SKM-GN metabolites in each category:

#>                                 SKM-GN
#> unique_metabs                      933
#> high_het_metabs_targeted             5
#> high_het_metabs_untargeted_only      1
#> meta_anal_metabs                   111
names(res)
#> [1] "timewise_meta_regression" "training_meta_regression"
#> [3] "meta_regression_results"  "meta_regression_models"  
#> [5] "metabolite_categories"