add_deconvolution_training_model_NMF: add_deconvolution_training_model_NMF

In RaikOtto/ArtDeco: Alikeness Based on Regression on Transciptome DECOnvolution

Description

add_deconvolution_training_model_NMF adds a new model

Usage

add_deconvolution_training_model_NMF(
    transcriptome_data,
    model_name,
    subtype_vector,
    rank_estimate,
    exclude_non_interpretable_NMF_components,
    training_nr_marker_genes,
    parallel_processes,
    nrun
)

Arguments

transcriptome_data	Transcriptomic data to be used for training. Has to contain the cell subtypes to which the similarity will be calculated. Note that the row names have to contain the HGNC symbols and the column names the sample names.
model_name	Name of the model.
subtype_vector	Character vector containing the subtype labels of the training data samples (transcriptome_data).
rank_estimate	Rank of the NMF model. Will be set to amount of different subtypes defined in the subtype_vector if not specified manually.
exclude_non_interpretable_NMF_components	Boolean parameter that indicates whether trained NMF components that cannot clearly be associated with either an endocrine or acinar & ductal or hisc subtyp shall be excluded. Default value FALSE.
training_nr_marker_genes	Amount of genes to be utilized as marker genes for each cell type. Default value 100.
parallel_processes	Amount of parallel processes used for training. Warning, RAM utilization increases linearly. Default value 1.
nrun	Amount of times the NMF model will be trained. Default value 10.

Value

Stores a new model in the package directory

Examples

data(Lawlor) # Data from Lawlor et al.
data(meta_data)

# extract the training sample subtype labels
subtype_vector = as.character(meta_data$Subtype) 

add_deconvolution_training_model_NMF(
    transcriptome_data = Lawlor,
    model_name = "my_model",
    subtype_vector = subtype_vector,
    rank_estimate = 0,
    exclude_non_interpretable_NMF_components = FALSE,
    training_nr_marker_genes = 100,
    parallel_processes = 1,
    nrun = 1
)

RaikOtto/ArtDeco documentation built on Oct. 30, 2021, 6:20 p.m.