Re-scale data using multiclass Gaussian Mixture Model. If data is lack of variance, the function will trying to do log normal distribution re-scale. Re-scaled values will be probabilities between 0-1.
fitGMM(
seu,
method = c("Rmixmod", "mclust"),
model = mixmodGaussianModel(family = "general", listModels = "Gaussian_p_Lk_Ck",
free.proportions = FALSE, equal.proportions = TRUE),
max_zero_percentage = 0.1,
...
)A Seurat object
The package used to do the GMM fit. Available tools are 'Rmixmod' and 'mclust'.
If method is Rmixmod, the model will be used to fit the data. see mixmodGaussianModel.
The cutoff of zero percentage. If the percentage is greater than max_zero_percentage, the zeros will be removed by random pick to meet the max percentage. This is used to make sure there is enough variance to fit the model.
Not use.
A Seurat object with a new assay named as 'GMM'. The re-scaled values saved in layer 'data'.