ALDEx2 analysis

Differential abundance analysis using ALDEx2's Monte Carlo log-ratio transformation and native statistical tests. With a centralized model, fits a GLM for each planned contrast and reports its coefficient and adjusted P value.

Usage

step_aldex(
  rec,
  max_significance = 0.05,
  mc.samples = 128,
  denom = "all",
  rarefy = FALSE,
  id = rand_id("aldex"),
  engine_args = list()
)

Arguments

rec

A Recipe object. The step will be added to the sequence of operations for this Recipe.

max_significance

Adjusted P value cutoff for the GLM, or the Benjamini-Hochberg corrected Welch P value in model-free execution.

mc.samples

The number of Monte Carlo instances to use to estimate the underlying distributions. The default is 128.

denom

An any variable (all, iqlr, zero, lvha, median, user) indicating features to use as the denominator for the Geometric Mean calculation The default "all" uses the geometric mean abundance of all features. Using "median" returns the median abundance of all features. Using "iqlr" uses the features that are between the first and third quartile of the variance of the clr values across all samples. Using "zero" uses the non-zero features in each grop as the denominator. This approach is an extreme case where there are many nonzero features in one condition but many zeros in another. Using "lvha" uses features that have low variance (bottom quartile) and high relative abundance (top quartile in every sample). It is also possible to supply a vector of row indices to use as the denominator. Here, the experimentalist is determining a-priori which rows are thought to be invariant. In the case of RNA-seq, this could include ribosomal protein genes and and other house-keeping genes. This should be used with caution because the offsets may be different in the original data and in the data used by the function because features that are 0 in all samples are removed by aldex.clr. Centralized models require "all" or explicit numeric denominator indices.

rarefy

Boolean indicating if OTU counts must be rarefyed. This rarefaction uses the standard R sample function to resample from the abundance values in the otu_table component of the first argument, physeq. Often one of the major goals of this procedure is to achieve parity in total number of counts between samples, as an alternative to other formal normalization procedures, which is why a single value for the sample.size is expected. If 'no_seed', rarefaction is performed without a set seed.

id

Unique, non-empty identifier for this step. DA step IDs must not use the reserved internal names taxa_id, contrast_id, comparison, contrast_type, var, effect, sum, .tax_order, .contrast_order or .effect_order. Other identifiers, including names such as .custom, are retained unchanged.

engine_args

Named lists of advanced arguments for the native clr and fit stages. The former effect stage is not supported: centralized effects now come from the GLM. Arguments managed by dar or exposed above cannot be overridden.

Value

An object of class Recipe

Details

With add_model(), effect is the contrast-oriented coefficient from ALDEx2::aldex.glm(), averaged over Monte Carlo instances, and effect_metric is "aldex_glm_coefficient". It comes from the same GLM as the reported adjusted P value. Under the default denominator, scale and Gaussian identity-link settings, the response is CLR in base 2; this coefficient is not an absolute-abundance log2 fold change or the standardized effect from ALDEx2::aldex.effect(). Advanced native settings can change the response or link scale. Recompute previously prepared centralized ALDEx2 results to obtain these coefficients.

Deprecated model-free execution retains native two-group Welch tests and standardized effects from ALDEx2::aldex.effect(), identified as "standardized_clr_effect" in tidy results.

See also

Other Diff taxa steps: step_ancom(), step_corncob(), step_deseq(), step_lefse(), step_linda(), step_maaslin(), step_wilcox()

Examples

data(metaHIV_phy)

rec <- recipe(metaHIV_phy) |>
  add_model(~ RiskGroup2, targets = "RiskGroup2", tax_level = "Phylum")

rec_default <- step_aldex(rec)

# Increase the number of Monte Carlo instances.
rec_more_instances <- step_aldex(rec, mc.samples = 256)