Get the top factors — get_top_factors • phenomix

Search a reduction for observations (e.g. traits) that match a given term (case-insensitive substring search). Then get the factors with the highest mean loading for the matching features.

get_top_factors(
  obj,
  metadata = NULL,
  reduction = NULL,
  term,
  search_col = "label_phe",
  n_quantiles = 10,
  select_quantiles = n_quantiles,
  plot_hist = FALSE,
  verbose = TRUE
)

Arguments

obj	Seurat object or dimensionality reduction object.
metadata	Phenotype metadata. Not needed if `obj` is a Seurat object.
reduction	Which reduction to extract from. Only used if `obj` is of class `Seurat`.
term	Term with which to perform substring search of observations.
search_col	Which column in the observation metadata to perform substring search.
n_quantiles	How many quantiles to bin factor loadings into.
select_quantiles	Which quantiles to return. Defaults to the top quantile only.
plot_hist	Whether to plot the distribution of loadings.
verbose	Print messages.

Value

data.table

Examples

data("DEGAS_seurat")
top_factors <- get_top_factors(
    obj = DEGAS_seurat,
    term = "parkinson",
    select_quantiles = 8:10
)
#> Extracting embeddings from Seurat reduction:  contributionGene
#> + 2 matching phenotypes identified:
#>   - HC206
#>   - FH1262