filter_contrast_names: Filter contrast names

In jmw86069/jamses: Jam SummarizedExperiment Stats

Filter contrast names

Description

Filter contrast names

Usage

filter_contrast_names(
  contrast_names = NULL,
  sedesign = NULL,
  factor_controls = NULL,
  apply_to_sedesign = FALSE,
  must_be_control = TRUE,
  verbose = FALSE,
  ...
)

Arguments

contrast_names	character vector of contrast names
sedesign	SEDesign object
factor_controls	optional character vector with one value per experiment factor column used in the contrast names. Typically the first observed value in each column is taken as the overall control, using only the baseline control group.
apply_to_sedesign	logical indicating when sedesign is supplied, whether to return an SEDesign object with update contrasts() consistent with the contrast_names generated by this function.
must_be_control	logical indicating whether the control factor level must also be the control in factor comparisons. When must-be_control=TRUE (default) the factor comparison must be in the form (Test-Control). When must_be_control=FALSE it permits the control factor level to be present as the test or control: (Control-Test) or (Test-Control).
verbose	logical indicating whether to print verbose output.
...	additional arguments are ignored.

Details

Utility function to help filter a large set of contrasts to a smaller set of comparisons versus specific factor level controls.

This function was motivated by a design with 5 treatments and 3 genotypes, the "all-by-all" pairwise strategy produced 45 oneway contrasts, and 75 contrasts overall.

After filtering for comparisons versus the first level per experimental factor, it produced 22 oneway contrasts, and 30 overall.

This function is still being tested to determine effective ways to pare down an enormous number of potential contrasts to the minimal set of "useful contrasts". The examples show a few use cases.

See Also

Other jam experiment design: check_sedesign(), contrast2comp(), contrast_colors_by_group(), contrast_names_to_sedesign(), contrasts_to_factors(), contrasts_to_venn_setlists(), draw_oneway_contrast(), draw_twoway_contrast(), groups_to_sedesign(), plot_sedesign(), sedesign_to_factors(), validate_sedesign()

Examples

group_names <- paste0(
   rep(c("UL3", "dH1A", "dH1B"), each=5), "_",
   c("Veh", "DEX", "PMA", "SF", "Ins"))
sedesign <- groups_to_sedesign(group_names)
plot_sedesign(sedesign,
   which_contrasts=jamba::unvigrep("[(]", contrast_names(sedesign)),
   arrow_ex=0.2, twoway_lwd=1, contrast_style="none")
title(main="45 total oneway and twoway contrasts")

new_contrast_names <- filter_contrast_names(contrast_names(sedesign))
new_contrast_names

# apply to sedesign directly
sedesign2 <- filter_contrast_names(sedesign=sedesign, apply_to_sedesign=TRUE)
plot_sedesign(sedesign2,
   which_contrasts=jamba::unvigrep("[(]", contrast_names(sedesign2)),
   arrow_ex=0.2, twoway_lwd=1, contrast_style="none")
title(main="22 oneway contrasts (vs UL3 / vs Veh)")

# subset for contrasts involving certain control values
# SF is the control for factors Ins,SF, so we filter for comparisons
# where SF is the control, using `must_be_control=TRUE` (default)
# but the contrast must involve SF or Ins directly
contrast_names_sf <- filter_contrast_names(
   contrast_names=jamba::vigrep("SF|Ins",
   contrast_names(sedesign)))
data.frame(jamba::rbindList(strsplit(contrast2comp(contrast_names_sf), ":")))

# must_be_control=FALSE allows any orientation involving "UL3" or "SF"
# it therefore also permits "SF-Veh"
contrast_names_sf <- filter_contrast_names(
   contrast_names=jamba::vigrep("SF|Ins", contrast_names(sedesign)),
   must_be_control=FALSE,
   factor_controls=c(Genotype="dH1A", Treatment="SF"))
data.frame(jamba::rbindList(strsplit(contrast2comp(contrast_names_sf), ":")))

# use must_be_control=c(FALSE, TRUE) to enforce only on the second factor
# 1. do not require UL3 to be the control in the comparison
# 2. do require SF to be the control in the comparison
contrast_names_sf <- filter_contrast_names(
   contrast_names=jamba::vigrep("SF|Ins", contrast_names(sedesign)),
   must_be_control=c(FALSE, TRUE),
   factor_controls=c(Genotype="dH1A", Treatment="SF"))
data.frame(jamba::rbindList(strsplit(contrast2comp(contrast_names_sf), ":")))

# for the example below, use this form
contrast_names_sf <- filter_contrast_names(jamba::vigrep("SF|Ins", contrast_names(sedesign)))
data.frame(jamba::rbindList(strsplit(contrast2comp(contrast_names_sf), ":")))

# Veh is the control for all other groups, omit Ins in these comparisons
contrast_names_veh <- filter_contrast_names(jamba::unvigrep("Ins", contrast_names(sedesign)))
data.frame(jamba::rbindList(strsplit(contrast2comp(contrast_names_veh), ":")))

use_contrasts <- unique(c(contrast_names_veh, contrast_names_sf))
sedesign3 <- sedesign;
contrast_names(sedesign3) <- use_contrasts

# show contrasts after filtering
jamba::rbindList(strsplit(contrast2comp(contrast_names(sedesign3)), ":"))

# show all one-way contrasts
plot_sedesign(sedesign3,
   which_contrasts=jamba::unvigrep("[(]", contrast_names(sedesign3)),
   arrow_ex=0.5, twoway_lwd=1, contrast_style="none")
title(main="22 oneway contrasts (filtered)")

# show only the two-way contrasts
plot_sedesign(sedesign3,
   which_contrasts=jamba::vigrep("[(]", contrast_names(sedesign3)),
   arrow_ex=0.5, twoway_lwd=1, contrast_style="none")
title(main="8 twoway contrasts (filtered)")

# same two-way contrasts, showing flipped orientation
plot_sedesign(sedesign3,
   flip_twoway=TRUE,
   which_contrasts=jamba::vigrep("[(]", contrast_names(sedesign3)),
   arrow_ex=0.5, twoway_lwd=1, contrast_style="none")
title(main="8 twoway contrasts (flipped)")

jmw86069/jamses documentation built on Nov. 4, 2024, 9:25 p.m.