The cohort as one picture: populations, samples and every clinical variable

WHAT IT IS. One column per sample, ordered along the clinical variable named in order_by. Below the strips, one row per population, filled by that sample's value expressed as a z-score WITHIN the population, so a row is readable whether the population is 40% of CD45+ or 0.4%. Above them, one strip per clinical variable.

Usage

fig_clinical_landscape(
  freq,
  clin,
  outfile,
  order_by = NULL,
  value_col = NULL,
  max_vars = Inf,
  dpi = 200,
  colors = fcs_colors()
)

Arguments

freq

population_frequencies table.

clin

named list of clinical variables, sample_id -> value.

outfile

path.

order_by

which clinical variable orders the columns; the first numeric one by default.

value_col

frequency column.

max_vars

most clinical strips to draw. Inf by default, so every variable passed to --clinical-columns appears; the figure grows taller with each strip rather than dropping the later ones.

dpi

resolution.

colors

palette.

Details

WHY Z-SCORE WITHIN A ROW AND NOT ACROSS THE MATRIX. Populations differ by two orders of magnitude in abundance. On a shared scale the frequent lineages saturate the palette and every rare one is the same colour, so the figure shows which populations are large – which is already known – instead of which samples are unusual, which is the question.

WHAT IT CANNOT SHOW. Nothing here is a test. A gradient that reads clearly across twelve ordered columns can still be the ordering the eye was given, and the tests in clinical_association.csv are what decide.