LinDA differential-abundance analysis

step_linda() fits bias-corrected linear or linear mixed-effects models with MicrobiomeStat::linda(). It requires a centralized model defined by add_model() and executes every contrast in that model's contrast plan.

Usage

step_linda(
  rec,
  prev_filter = 0,
  mean_abund_filter = 0,
  max_abund_filter = 0,
  winsorize = TRUE,
  outlier_pct = 0.03,
  adaptive = TRUE,
  zero_handling = c("pseudo-count", "imputation"),
  pseudo_count = 0.5,
  corr_cut = 0.1,
  p_adj_method = "BH",
  alpha = 0.05,
  n_cpus = 1L,
  rarefy = FALSE,
  id = rand_id("linda"),
  engine_args = list()
)

Arguments

rec

An unprepared Recipe object. Execution requires add_model() before prep(); the step can be configured first.

prev_filter

Minimum feature prevalence retained for analysis.

mean_abund_filter

Minimum mean relative abundance retained.

max_abund_filter

Minimum maximum relative abundance retained.

winsorize

Whether high-abundance outliers are winsorized.

outlier_pct

Expected fraction of outliers used for winsorization.

adaptive

Whether LinDA chooses zero handling from depth-covariate correlations.

zero_handling

Zero treatment used when adaptive = FALSE: either "pseudo-count" or "imputation".

pseudo_count

Positive pseudo-count used by pseudo-count zero handling.

corr_cut

Significance cutoff used by adaptive zero handling.

p_adj_method

Multiple-testing correction accepted by p.adjust().

alpha

Adjusted p-value threshold used to classify significance.

n_cpus

Number of cores used by LinDA mixed-effects models.

rarefy

Whether counts are rarefied for this step.

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 fit stage. Arguments managed by dar or exposed above cannot be overridden.

Value

A Recipe object.

Details

With MicrobiomeStat 1.4, dar uses a local copy of the native function with the adaptive zero-handling capitalization error corrected. This restores the native choice of imputation when sequencing depth is associated with the design. The installed package, model fitting, winsorization and multiple-testing correction are unchanged.

See also

Other Diff taxa steps: step_aldex2(), step_aldex3(), step_ancom(), step_corncob(), step_deseq(), step_lefse(), step_maaslin(), step_rademu(), step_wilcox()

Examples

data(metaHIV_phy)

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

rec_default <- step_linda(rec)

# Retain features present in at least 10% of samples.
rec_filtered <- step_linda(rec, prev_filter = 0.1)