CytOpT {CytOpT}

Function to estimate the type cell proportions in an unclassified cytometry data set denoted X_s by using the classification Lab_source from an other cytometry data set X_s. With this function the computation of the estimate of the class proportions is done with a descent ascent or minmax or two algorithms.

Description

Function to estimate the type cell proportions in an unclassified cytometry data set denoted X_s by using the classification Lab_source from an other cytometry data set X_s. With this function the computation of the estimate of the class proportions is done with a descent ascent or minmax or two algorithms.

Usage

CytOpT(
  X_s,
  X_t,
  Lab_source,
  Lab_target = NULL,
  theta_true = NULL,
  method = c("minmax", "desasc", "both"),
  eps = 1e-04,
  n_iter = 10000,
  power = 0.99,
  step_grad = 10,
  step = 5,
  lbd = 1e-04,
  n_out = 5000,
  n_stoc = 10,
  minMaxScaler = TRUE,
  monitoring = FALSE,
  thresholding = TRUE
)

Arguments

X_s

a cytometry dataframe with only d numerical variables for ns observations. The columns correspond to the different biological markers measured. One line corresponds to the cytometry measurements performed on one cell. The classification of this Cytometry data set must be provided with the Lab_source parameters.

X_t

a cytometry dataframe with only d numerical variables for nt observations. The columns correspond to the different biological markers measured. One line corresponds to the cytometry measurements performed on one cell. The CytOpT algorithm targets the cell type proportion in this Cytometry data set

Lab_source

a vector of length ns Classification of the X_s cytometry data set

Lab_target

a vector of length nt Classification of the X_s cytometry data set

theta_true

If available, gold-standard proportions in the target data set X_t derived from manual gating. It allows to assess the gap between the estimate and the gold-standard. Default is NULL, in which case no assessment is performed.

method

a character string indicating which method to use to compute the cytopt, either 'minmax', 'desasc' or 'both' for comparing both Min-max swapping and descent-ascent procedures. Default is 'minmax'.

eps

a float value of regularization parameter of the Wasserstein distance. Default is 1e-04

n_iter

an integer Constant that iterate method select. Default is 10000

power

a float constant the step size policy of the gradient ascent method is step/n^power. Default is 0.99

step_grad

an integer number step size of the gradient descent algorithm of the outer loop. Default is 10

step

an integer constant that multiply the step-size policy. Default is 5

lbd

a float constant that multiply the step-size policy. Default is 1e-04

n_out

an integer number of iterations in the outer loop. This loop corresponds to the gradient descent algorithm to minimize the regularized Wasserstein distance between the source and target data sets. Default is 1000

n_stoc

an integer number of iterations in the inner loop. This loop corresponds to the stochastic algorithm that approximates a maximizer of the semi dual problem. Default is 10

minMaxScaler

a logical flag indicating to whether to scale observations between 0 and 1. Default is TRUE.

monitoring

a logical flag indicating to possibly monitor the gap between the estimated proportions and the manual gold-standard. Default is FALSE.

thresholding

a logical flag indicating whether to threshold negative values. Default is TRUE.

Value

a object of class CytOpt, which is a list of two elements:

• proportions a data.frame with the (optionally true and) estimated proportions for each method

• monitoring a list of estimates over the optimization iterations for each method (listed within)

Examples

if(interactive()){

res <- CytOpT(X_s = HIPC_Stanford_1228_1A, X_t = HIPC_Stanford_1369_1A, 
             Lab_source = HIPC_Stanford_1228_1A_labels,
             method='minmax')
summary(res)
plot(res)

}