ml_tax_HFE: Hierarchical Feature Selection

View source: R/ml.R

ml_tax_HFER Documentation

Hierarchical Feature Selection

Description

For each clade (defined by tax_level), aggregate species abundances at each taxonomic level up to the user-defined "tax_level", (optionally filter out near-zero features), then filter out taxa that correlate strongly (just one taxon is selected of those that correlate).

Usage

ml_tax_HFE(
  brk,
  tax_level,
  corr_cutoff = 0.7,
  threads = 2,
  freqCut = 95/1,
  uniqueCut = 5,
  quiet = TRUE
)

Arguments

brk

data.table generated by read_bracken(). Columns: Sample, Abundance, Phylum=>Species

tax_level

which taxonmoic level to use?

corr_cutoff

features with >cutoff will be filtered to just one

freqCut

as in caret::nearZeroVar; use NULL to skip

uniqueCut

as in caret::nearZeroVar; use NULL to skip

Value

data.table of filtered features


leylabmpi/LeyLabRMisc documentation built on Nov. 3, 2022, 3:45 p.m.