fitMuDispGeneMM: Optim wrapper for gene-wise models other than mixture model.

In LineagePulse: Differential expression analysis and model fitting for single-cell RNA-seq data

Description

NOT YET SUPPORTED. LINEAGEPULSE CODE WILL BE EXTENDED BY MODULAR FUNCTIONALITIES AND THIS IS ONE INSTANCE OF A PLACEHOLDER USED FOR DEVELOPING. Given a parameter initialisation, this function performs numerical optimisation using BFGS of the likelihood function given the supplied mean and dispersion model. This is the wrapper that calls optim.

Usage

fitMuDispGeneMM(vecCounts, vecMuGuess, lsvecBatchParamGuessMu, lsMuModelGlobal,
  vecDispGuess, lsvecBatchParamGuessDisp, lsDispModelGlobal, matDropoutLinModel,
  vecPiConstPredictors, lsDropModelGlobal, vecPiParam, matWeights, MAXIT,
  RELTOL)

Arguments

vecCounts	(count vector number of cells) Observed read counts, not observed are NA.
vecMuGuess	(numeric vector number of mean model parameters) Initialisation for impulse model.
lsvecBatchParamGuessMu	(list) Object containing initialisation for mean parameter batch correction model.
lsMuModelGlobal	(list) Object containing meta-data of gene-wise mean parameter models.
vecDispGuess	(numeric vector number of dispersion model parameters) Initialisation for dispersion model.
lsvecBatchParamGuessDisp	(list) Object containing initialisation for dispersion parameter batch correction model.
lsDispModelGlobal	(list) Object containing meta-data of gene-wise dispersion parameter models.
matDropoutLinModel	(matrix number of cells x number of predictors) Logistic linear model parameters of the dropout rate as a function of the mean and constant gene-wise coefficients.
vecPiConstPredictors	(numeric vector constant gene-wise coefficients) Constant gene-wise coeffiecients, i.e. predictors which are not the offset and not the mean parameter.
lsDropModelGlobal	(list) Object containing meta-data of cell-wise drop-out parameter models.
vecPiParam	(numeric vector number of observations) Pre-evaluated drop-out model if model is not a function on the mean parameter to be fit.
matWeights	(numeric matrix cells x mixtures) [Default NULL] Assignments of cells to mixtures (for strMuModel="MM").
MAXIT	(scalar) Maximum number of BFGS iterations handed to optim().
RELTOL	(scalar) Cost function convergence criterium handed to optim().

Details

This function performs error handling of the numerical fitting procedure. This function corrects for the likelihood sensitivity bounds used in the cost function.

Value

list

• vecMuModel (numeric vector number of mu model parameters) Contains the mean model parameters according to the used model.

• lsvecBatchModelMu (list) Fit of batch correction models for mean parameter to given gene.

• vecDispModel (numeric vector number of dispersion model parameters) Contains the dispersion model parameters according to the used model.

• lsvecBatchModelDisp (list) Fit of batch correction models for dispersion parameter to given gene.

• scaConvergence (numeric vector number of genes) Convergence status of optim for given gene.

• scaLL (numeric vector number of genes) Likelihood of model fit for given gene.

Author(s)

David Sebastian Fischer

LineagePulse documentation built on Nov. 8, 2020, 7:01 p.m.