tinyarray (version 3.0.0)

auto_geo_group: auto_geo_group

Description

Infer sample groups from GEO phenotype metadata.

Usage

auto_geo_group(
  x,
  group_col = "title",
  choice = NULL,
  min_group_size = 2L,
  max_groups = 20L,
  min_coverage = 0.8,
  search_other_cols = TRUE,
  stopwords = .tinyarray_geo_stopwords(),
  levels = NULL,
  ref = NULL,
  verbose = TRUE
)

Value

A list with `group_list`, `source_col`, `selected_index`, `candidate_table`, `candidates`, and `pdata`.

Arguments

x

A GEO result list with a `pd` data frame, or a phenotype data frame.

group_col

Column name to try first. The default is `title`.

choice

Candidate index to select. Use `0` to cancel. If `NULL`, the function returns the candidate table without selecting.

min_group_size

Minimum number of samples required for a split label.

max_groups

Maximum number of group labels allowed in one column.

min_coverage

Minimum fraction of samples that must be assigned a label.

search_other_cols

Whether to search other columns when `group_col` and `title` fail.

stopwords

Character vector of generic words to ignore during tokenization.

levels

Optional character vector giving the desired factor level order.

ref

Optional reference level to place first.

verbose

Whether to print messages and candidate tables.

Details

The function first tries the user-specified `group_col`, then `title`. If no stable grouping is found, it scans the remaining columns and returns a table of candidate groupings for manual selection. Both binary and multigroup factors are supported. Rare groups are left as `NA` rather than being relabeled as another group.

Labels that differ only in capitalization are merged, retaining their most common spelling. Shared affixes and trailing sample identifiers are removed when they can be identified consistently.

Examples

pd <- data.frame(
  title = c("Control sample", "Control sample", "Treat sample", "Treat sample"),
  stringsAsFactors = FALSE
)
rownames(pd) <- paste0("s", 1:4)
res <- auto_geo_group(
  pd,
  choice = 1,
  levels = c("Control", "Treat"),
  ref = "Control",
  verbose = FALSE
)
res$source_col
levels(res$group_list)
table(res$group_list)