Prepare SEStats from a list of stat data.frame

Usage

list_to_sestats(
  stats_dfs,
  use_assay_name = "norm",
  p_cutoff = 1,
  adjp_cutoff = 0.05,
  fold_cutoff = 1.5,
  mgm_cutoff = 0,
  hit_pattern = "hit",
  fold_pattern = "fold",
  lfc_pattern = "^logFC|^log2F|^logfold|^log.fold|^lfc",
  p_pattern = "p.{0,1}val(ue|)",
  adjp_pattern = "adj.{0,1}p.{0,1}val(ue|)|p.{0,1}adj|fdr|q.{0,1}val(ue|)",
  mgm_pattern = "mgm|max.{0,1}group.{0,1}mean",
  contrast_pattern = "^.*[ ]([^ ]+-[^ ]+)$",
  verbose = FALSE,
  ...
)

Arguments

stats_dfs

list with one of two formats:

Containing a list named by assay_names (signal), which then contains a list of data.frame objects.
Containing data.frame objects.

Note: The list of data.frame objects can be named by contrast, and these contrast names will be used if the contrast is not already encoded in the colnames the corresponding data.frame.

use_assay_name

character string, default "norm", used when argument 'stats_dfs' is supplied as a list of data.frames (option 2 described for 'stats_dfs'). In this case, the assay name will be 'use_assay_name'.

p_cutoff, adjp_cutoff, fold_cutoff, mgm_cutoff

numeric values used when there is no matching hit column in each data.frame, matching the hit_pattern.

hit_pattern, fold_pattern, lfc_pattern, p_pattern, adjp_pattern, mgm_pattern

character string with regular expression pattern used to match each corresponding column type: Hit column, fold change, log fold change, P-value (unadjusted), adjusted P-value (or FDR or Q-value), max group mean (mgm).

The pattern is assumed to be at the start of each column name string.

contrast_pattern

character string used to match the contrast encoded in any of the cutoff columns for the '_pattern' arguments.

The pattern is essentially any non-whitespace string that contains a hyphen '-' inside it. If it also contains a colon ':' the string is assumed to be a comp (see comp2contrast()) otherwise it is considered a contrast name.
When present, the contrast is extracted from the column name.
When not present, the name from the list of data.frame objects is assumed to be the contrast name.

verbose

logical indicating whether to print verbose output.

...

additional arguments are ignored.

Value

list consistent with se_contrast_stats() output, except limited to elements: 'stats_df', 'hit_array', 'hit_list'. The content should be sufficient for use in heatmap_se() and other related functions.

Details

This function converts a list of stat data.frame objects to a list object as returned by se_contrast_stats(). In future this function will return a S4 SEStats object.

Limitations

This function only populates certain components:

stats_dfs as a list by assay_name (signal), with list by contrast name with data.frame objects.
hit_array as a 3-dimension array with 'Cutoffs', 'Contrasts', and 'Signal' dimensions.
hit_list as a list by assay name (signal), with list by contrast names, containing numeric values named by measurement.

This function does not return certain data which is not described in a list of data.frame objects:

idesign, icontrasts, normgroup cannot be defined or inferred.

Other limitations:

stats_df with the combined stats across contrasts as one wide data.frame, is not created, partly because it does not assume the contrast name is included in each corresponding stat colname in each data.frame. Potential future enhancement.

Rules Used

In each data.frame certain colnames are recognized:

The hit columns is recognized by one or more colnames that begin with "hit ". The hit column should have values of -1, 0, or 1 to indicate whether the row met the statistical criteria. Typically the statistical criteria are also included in the column header, followed by the contrast name.
- If there is no hit column, the arguments '_cutoff' are applied to each row, and a new hit column is created accordingly. If there are multiple cutoff values, multiple hit columns are created.
Fold change column is recognized by a column beginning "fold " (with space), typically followed by the contrast name. For example "fold Dex-Ctrl".
- If no fold change column exists, it searches for log fold column, specifically assumed to be log2 fold change even when "logFC". By convention, edgeR and limma use "logFC" for log2 fold change.
- If neither fold change, nor log fold change, columns are found, this function stops.
- The log2 fold change is converted to normal fold change for filtering.
The raw P-value column should begin "P-Value " or any variation of "pval", "p.val", "pvalue", "p.value", case-insensitive.
The adjusted P-value column should begin "adj-P-Val " or any variation of "adjp", "padj", "pvalue", "p.value", case-insensitive.