InitBinaryFA: This function should be called to initialize input parameters...

In quon-titative-biology/bfa: A dimensionality reduction tool using gene detection pattern to mitigate noisy expression profile of scRNA-seq

Description

This function should be called to initialize input parameters into the main scBFA function

Usage

InitBinaryFA(
  modelEnv,
  GeneExpr,
  numFactors,
  epsilon,
  X = NULL,
  Q = NULL,
  initCellcoef,
  updateCellcoef,
  updateGenecoef,
  NUM_CELLS_PER_CHUNK = min(ncol(GeneExpr), 50000),
  doChunking = (NUM_CELLS_PER_CHUNK < modelEnv$numCells)
)

Arguments

modelEnv	Empty R environment variable to contain following parameters: A,Z,V,U,β,γ,ε
GeneExpr	G by N rawcount matrix, in which rows are genes and columns are cells
numFactors	Numeric value, number of latent dimensions
epsilon	Numeric value, parameter to control the strength of regularization
X	N by C cell-specific covariate matrix(e.g batch effect), in which rows are cells,columns are number of covariates. If no such covariates are available, then X = NULL
Q	G by T gene-specific covariate matrix(e.g quality control measures), in which rows are genes columns are number of covariates, If no such covariates are available, then Q = NULL
initCellcoef	Initialization of C by G gene-specific coefficient matrix as user-defined coefficient β. Such user defined coefficient can be applied to address confounding batch effect
updateCellcoef	Logical value, parameter to decide whether to update C by G gene-specific coefficient matrix. Again, when the cell types are confounded with technical batches or there is no cell level covariate matrix, the user can keep the initialization of coefficients as known estimate.
updateGenecoef	Logical value, parameter to decide whether to update N by T gene-specific coefficient matrix. Again, when there is no gene level covariate matrix, this value should be FALSE by default.
NUM_CELLS_PER_CHUNK	scBFA can run out of memory on large datasets, so we can chunk up computations to avoid this if necessary. NUM_CELLS_PER_CHUNK is the number of cells per 'chunk' computed. Shrink if running out of mem.
doChunking	Use memory-efficient (but slower) chunking. Will do automatically if the chunk size is specified to be smaller than the # of cells in dataset.

Value

A model environment containing the following parameters: A,Z,V,U,β,γ,ε.

quon-titative-biology/bfa documentation built on March 18, 2020, 11:53 a.m.