Identify cell-type niches from NMF latent factors — compute_cells

For each NMF factor, extracts the top-contributing cell groups (by basis weight), maps them to their constituent cell types, and keeps only cell types enriched relative to the background composition. Saves a star-network PDF per factor and returns the weighted cell-type associations.

Usage

compute_cells_niches(
  latent_factors,
  dt,
  cell.groups,
  enrich_thresh = 1.5,
  quantile_cutoff = 0.7,
  cells_extra = NULL,
  return = TRUE,
  file_name = NULL
)

Arguments

latent_factors: A list returned by compute.latent_factors(), containing W (features x factors) and the NMF model object.
dt: A named list of deconvolution subgroup results, used to build the cell-group composition matrix via compute.composition.matrix().
cell.groups: A list of cell group definitions as returned by construct_cell_groups().
enrich_thresh: Numeric. Minimum enrichment ratio (foreground / background frequency) for a cell type to be retained per factor. Default 1.5.
quantile_cutoff: Numeric between 0 and 1. Quantile threshold for selecting top-contributing cell groups per factor. Default 0.7.
cells_extra: Optional character vector of additional cell-type columns to include in the composition matrix.
return: Logical. If TRUE, saves network PDF plots to Results/. Default TRUE.
file_name: Character. Suffix appended to output file names.

Value

A named list (one element per factor) of named numeric vectors giving the enriched cell types and their cumulative NMF edge weights, sorted descending.