Get Spectral Variations for Fluorophores — get.spectral.variants • AutoSpectral

This function cycles through all the fluorophores defined in control.def.file, identifying variations in spectral profiles. It does this by performing SOM clustering on the positive events in the cleaned control data. The output is saved as an .rds file, and figures summarizing the variation are saved, if desired. Note that the .rds file contains all the needed information for downstream processing (per-cell unmixing), so you can just load that using the readRDS function) rather than re-running this process.

Usage

get.spectral.variants(
  control.dir,
  control.def.file,
  asp,
  spectra,
  af.spectra,
  n.cells = 2000,
  som.dim = 10,
  figures = TRUE,
  output.dir = NULL,
  parallel = FALSE,
  verbose = TRUE,
  threads = NULL,
  refine = FALSE,
  problem.quantile = 0.95,
  ...
)

Arguments

control.dir: File path to the single-stained control FCS files.
control.def.file: CSV file defining the single-color control file names, fluorophores they represent, marker names, peak channels, and gating requirements.
asp: The AutoSpectral parameter list. Prepare using get.autospectral.param
spectra: A matrix containing the spectral data. Fluorophores in rows, detectors in columns.
af.spectra: Spectral signatures of autofluorescences, normalized between 0 and 1, with fluorophores in rows and detectors in columns. Prepare using get.af.spectra.
n.cells: Numeric, default 2000. Number of cells to use for defining the variation in spectra. Up to n.cells cells will be selected as positive events in the peak channel for each fluorophore, above the 99.5th percentile level in the unstained sample.
som.dim: Numeric, default 10. Number of x and y dimensions to use in the SOM for clustering the spectral variation. The number of spectra returned for each fluorophore will increase with the quadratic of som.dim, so for 10, you will get up to 100 variants. Somewhere between 4 and 7 appears to be sufficient, but with the pruning of variants implemented in unmix.autospectral() in v1.0.0, this is less important.
figures: Logical, controls whether the variation in spectra for each fluorophore is plotted in output.dir. Default is TRUE.
output.dir: File path to whether the figures and .rds data file will be saved. Default is NULL, in which case asp$variant.dir will be used.
parallel: Logical, default is FALSE, in which case sequential processing will be used. The new parallel processing should always be faster.
verbose: Logical, default is TRUE. Set to FALSE to suppress messages.
threads: Numeric, default is NULL, in which case asp$worker.process.n will be used. asp$worker.process.n is set by default to be one less than the available cores on the machine. Multi-threading is only used if parallel is TRUE.
refine: Logical, default is FALSE. Controls whether to perform a second round of variation measurement on "problem cells", which are those with the highest spillover, as defined by problem.quantile. When FALSE, behavior is identical to versions of AutoSpectral prior to 1.0.0. Setting to TRUE may help reduce spillover spread and unmixing errors. Using refine=TRUE does not impact subsequent unmixing calculation time in any significant way, unlike the same setting in get.af.spectra().
problem.quantile: Numeric, default 0.95. The quantile for determining which cells will be considered "problematic" after unmixing with per-cell AF extraction. Cells in the problem.quantile or above with respect to total signal in the fluorophore (non-AF) channels after per-cell AF extraction will be used to determine additional autofluorescence spectra, using a second round of clustering and modulation of the previously selected autofluroescence spectra. A value of 0.95 means the top 5% of cells, those farthest from zero, will be selected for further investigation.
...: Ignored. Previously used for deprecated arguments such as pos.quantile and sim.threshold, which are now fixed internally and no longer user-settable.

Value

A vector with the indexes of events inside the initial gate.