fit.gene.k: Build mixture models of gene expression
Description
Models the imputed gene expression values as a mixture of gaussian
distributions. For a two-state model, estimates the probability that a given
cell is in the 'on' or 'off' state for any gene. Followed by a greedy
k-means step where cells are allowed to flip states based on the overall
structure of the data (see Manuscript for details)
Usage
fit.gene.k(object, gene, do.k = 2, num.iter = 1, do.plot = FALSE, genes.use = NULL, start.pct = NULL)
Arguments
do.k
Number of modes for the mixture model (default is 2)
num.iter
Number of 'greedy k-means' iterations (default is 1)
do.plot
Plot mixture model results
genes.use
Genes to use in the greedy k-means step (See manuscript for details)
start.pct
Initial estimates of the percentage of cells in the 'on'
state (usually estimated from the in situ map)
Value
A Seurat object, where the posterior of each cell being in the 'on'
or 'off' state for each gene is stored in object@mix.probs