usage: Deconv [-h] --ref REF --bulk BULK --outdir OUTDIR [--cell-type-key CELL_TYPE_KEY] [--transpose-bulk] [--no-plots] [--figure-fmt {pdf,png}] [--model {nb,gamma,beta,lognormal,static}]
            [--num-genes NUM_GENES] [--layer LAYER] [--fp-hack] [--ref-epochs REF_EPOCHS] [--ref-lr REF_LR] [--ref-lrd REF_LRD] [--ref-dropout {separate,shared,none}]
            [--dec-dropout {auto,true,false}] [--dec-epochs DEC_EPOCHS] [--dec-lr DEC_LR] [--dec-lrd DEC_LRD] [--device {auto,cpu,cuda}] [--ncores NCORES] [--version]

Probabilistic Cell Type Deconvolution of Bulk RNA-seq.

options:
-h, --help            show this help message and exit
--ref REF             Path to Single-cell reference (.h5ad/.csv/.tsv/.txt). If you provide as matrix (csv/.tsv/.txt), make sure the genes are in the columns and cell-types are specified in the first
                        column.
--bulk BULK           Path to bulk RNA-seq data (.csv/.tsv/.txt/.h5ad). Make sure genes are in the columns and samples are in the rows
--outdir OUTDIR       Path to output directory.
--cell-type-key CELL_TYPE_KEY
                        Key in 'adata.obs' that contains cell type labels.
--transpose-bulk      Transpose bulk data.
--no-plots            Do not save plots.
--figure-fmt {pdf,png}
                        Format for saving figures.
--model {nb,gamma,beta,lognormal,static}
                        Model to use for gene expression modelling.
--num-genes NUM_GENES
                        Number of genes to use for deconvolution (top-n variable genes). Default: all genes.
--layer LAYER         Layer with raw counts in AnnData object. Default: 'X'.
--fp-hack             Use the floating point underflow for numerical stability.
--ref-epochs REF_EPOCHS
                        Number of epochs for reference model training.
--ref-lr REF_LR       Learning rate for reference model training.
--ref-lrd REF_LRD     Learning rate decay for reference model training.
--ref-dropout {separate,shared,none}
                        Reference model dropout type. 'separate' - Unique for each cell type; 'shared' - Shared across all cell types; 'none' - No dropout.
--dec-dropout {auto,true,false}
                        Use dropout in the bulk gene expression distribution. Default: 'auto' - Use dropout if 'ref-dropout' is not 'none'.
--dec-epochs DEC_EPOCHS
                        Number of epochs for deconvolution model training.
--dec-lr DEC_LR       Learning rate for deconvolution model.
--dec-lrd DEC_LRD     Learning rate decay for deconvolution model.
--device {auto,cpu,cuda}
                        Device to use for computations. Default: 'cuda' if available.
--ncores NCORES       Number of cores to use, default: no limit (-1).
--version             show program's version number and exit