Unmix Using Poisson Regression — unmix.poisson • AutoSpectral

This function performs unmixing of raw data using Poisson regression, with iterative reweighted least squares (IRLS) and fallback methods for cells that fail to converge.

Usage

unmix.poisson(
  raw.data,
  spectra,
  asp,
  initial.weights = NULL,
  noise.floor = 125,
  parallel = TRUE,
  threads = NULL
)

Arguments

raw.data: Matrix containing raw data to be unmixed.
spectra: Matrix containing spectra information.
asp: The AutoSpectral parameter list. Prepare using get.autospectral.param
initial.weights: Optional numeric vector of weights, one per fluorescent detector. Default is NULL, in which case weighting will be done by channel means.
noise.floor: Numeric, default 125. Lower clamp on the initial WLS weights used to seed each cell's fit. Signal units, same convention as noise.floor elsewhere in the package. Not used inside glm.fit itself; see unmix.poisson.fast() for floor-aware per-cell IRLS reweighting.
parallel: Logical, default is TRUE, which enables parallel processing for per-cell unmixing.
threads: Numeric, default is NULL, in which case asp$worker.process.n will be used if parallel=TRUE.

Value

A matrix containing the unmixed data.

References

Novo D. et al. (2014). "Generalized Unmixing Model for Multispectral Flow Cytometry Utilizing Nonsquare Compensation Matrices" Cytometry Part A, 83(5):508–520. doi:10.1002/cyto.a.22272