SCMMIB documentation

User manual of SCMMIB package

paired_latent_metrics

## for all paired integration metrics except for graph output
paired_latent_metrics(adata,
                          method = 'bindSC',
                          cluster = 'louvain',
                          batch = 'batch',
                          label = 'cell_type',
                          outf = './text.txt'
                         )

adata: anndata object for joint embedding. The obs is celll metadata. The input latent embedding stores in adata.X.
method: name of the input algorithm.
cluster: clustering methods, deprecated.
batch: column name of batch used. Do not consider batch effect removal if batch=None.
label: column name of cell type used. Do not consider biological conservation if label=None
outf: output metric filename. Stdout if outf = None.

paired_graph_metrics

## for all paired integration metrics of graph output

paired_graph_metrics(adata,
                          method = 'bindSC',
                          cluster = 'louvain',
                          batch = 'batch',
                          label = 'cell_type',
                          outf = './text.txt'
                         )

adata: anndata object for joint graph. The obs is cell metadata. The input graph stores in adata.obsp[method]
method: name of the input algorithm.
cluster: clustering methods, deprecated.
batch: column name of batch used. Do not consider batch effect removal if batch=None.
label: column name of cell type used. Do not consider biological conservation if label=None
outf: output metric filename. Stdout if outf = None.

unpaired_latent_metrics

## for all unpaired integration metrics
unpaired_latent_metrics(adata,
                            method,
                            cluster = 'louvain',
                            batch = 'batch',
                            label = 'cell_type',
                            mods = ['RNA', 'ATAC'],
                            outf = None,
                            embed_acc = True
                           ):

adata: anndata object. the obs is celll metadata. Two latent embeddings are stored in obsm[‘RNA’] and obsm[‘ATAC’] or obsm[‘ADT’]
method: name of the input algorithm.
cluster: clustering methods, deprecated.
batch: column name of batch used. Do not consider batch effect removal if batch=None.
label: column name of cell type used. Do not consider biological conservation if label=None
mods: the name of obsm data used for evaluation.
outf: output metric filename. Stdout if outf = None.

mosaic_latent_metrics

## for mosaic integration metrics in accuracy and downsample robustness tasks.
mosaic_latent_metrics(latents,metadatas,paired="s1d1", unpaired="s3d10",
                            mod2="atac", batch="batch",label="cell_type",
                            latent_path="", method='sciPENN', writef=True):

latents: python list of np.ndarry or pandas Dataframe object for paired, RNA and the other modality. Detail see tutorials.
metadatas: python list of np.ndarry or pandas Dataframe object for cell metadata of paired, RNA and the other modality.
paired: batch name of paired cells in metadata batch column.
unpaired: batch name of unpaired cells in metadata batch column.
mod2: the second mosaic modality. “atac” or “adt”
batch: column name of batch used. Do not consider batch effect removal if batch=None.
label: column name of cell type used. Do not consider biological conservation if label=None
latent_path: name of input latent, must contain “latent” in filename. And two metrics files will generated in the samme path of different suffix.
method: name of the input algorithm.
writef: write two metrics file in latent_path if True. stdout if FALSE.

mosaic_cnk_latent_metrics

# for mosaic integration metrics in paired sizes robustness tasks. only focus on unpair size
mosaic_cnk_latent_metrics(latents,metadatas,paired="s1d1", unpaired="s3d10",
                            mod2="atac", batch="batch",label="cell_type",latent_path="",
                             method='sciPENN', writef=True):

latents: python list of np.ndarry or pandas Dataframe object for paired, RNA and the other modality.
metadatas: python list of np.ndarry or pandas Dataframe object for cell metadata of paired, RNA and the other modality.
paired: batch name of paired cells in metadata batch column.
unpaired: batch name of unpaired cells in metadata batch column.
mod2: the second mosaic modality. “atac” or “adt”
batch: column name of batch used. Do not consider batch effect removal if batch=None.
label: column name of cell type used. Do not consider biological conservation if label=None
latent_path: name of input latent, must contain “latent” in filename. And two metrics files will generated in the samme path of different suffix.
method: name of the input algorithm.
writef: write two metrics file in latent_path if True. stdout if FALSE.

imputation_pair_rna_atac

# For paired scRNA and scATAC imputation methods
imputation_pair_rna_atac(metadata_path,rna_imp_path, rna_path,atac_imp_path, atac_path, method, outf=None ):

metadata_path: path of cell metadata.
rna_imp_path: path of rna imputation csv file.
rna_path: path of rna raw h5ad file. Used as rna gold standard.
atac_imp_path: path of atac imputation csv file.
atac_path: path of atac raw h5ad file. Used as atac gold standard.
method: name of the input algorithm.
outf: output to given file name or pd.Dataframe() (None).

imputation_mosaic_rna_atac

# For unpaired scRNA and scATAC mosaic imputation methods
imputation_mosaic_rna_atac(metadata_path,rna_imp_path, rna_path,atac_imp_path, atac_path,method,paired="s1d1",unpaired="s3d10",batch="batch", outf=None ):

metadata_path: path of cell metadata.
rna_imp_path: path of rna imputation csv file.
rna_path: path of rna raw h5ad file. Used as rna gold standard.
atac_imp_path: path of atac imputation csv file.
atac_path: path of atac raw h5ad file. Used as atac gold standard.
method: name of the input algorithm.
paired: batch name of paired cells in metadata batch column.
unpaired: batch name of unpaired cells in metadata batch column.
outf: output to given file name or pd.Dataframe() (None).

imputation_mosaic_rna_adt

# For unpaired scRNA and ADT mosaic imputation methods
imputation_mosaic_rna_adt(metadata_path,rna_imp_path=None, rna_path=None,adt_imp_path=None, adt_path=None, method="sciPENN",paired="s3d6", unpaired="s2d1", batch="batch", outf=None ):

metadata_path: path of cell metadata.
rna_imp_path: path of rna imputation csv file.
rna_path: path of rna raw h5ad file. Used as rna gold standard.
adt_imp_path: path of atac imputation csv file.
adt_path: path of atac raw h5ad file. Used as adt gold standard.
method: name of the input algorithm.
paired: batch name of paired cells in metadata batch column.
unpaired: batch name of unpaired cells in metadata batch column.
outf: output to given file name or pd.Dataframe() (None).

imputation_rna

imputation_rna(metadata_path,rna_imp_path, rna_path, method, outf=None ):
#  For scMVAE only. Only evaluate the scRNA modality.

metadata_path: path of cell metadata.
rna_imp_path: path of rna imputation csv file.
rna_path: path of rna raw h5ad file. Used as rna gold standard.
outf: output to given file name or pd.Dataframe() (None).

imputation_stabmap

# For stabmap mosaic scRNA and scATAC only, which omit scATAC knn smoothing for too few scATAC peaks.

imputation_stabmap(metadata_path,rna_imp_path, rna_path,atac_imp_path, atac_path,method,paired="s1d1",unpaired="s3d10",batch="batch", outf=None ):

metadata_path: path of cell metadata.
rna_imp_path: path of rna imputation csv file.
rna_path: path of rna raw h5ad file.
atac_imp_path: path of atac imputation csv file.
atac_path: path of atac raw h5ad file.
method: name of the input algorithm.
paired: batch name of paired cells in metadata batch column.
unpaired: batch name of unpaired cells in metadata batch column.
outf: output to given file name or pd.Dataframe() (None).

mouse_brain_divide

mouse_brain_divide(func, adatas,
                        method,
                        cluster = 'louvain',
                        batch = 'batch',
                        label = 'cell_type',
                        outf = None
                        ):
## A wrap function for mouse brain dataset and all functions above.

#    10X mouse brain datasets calculated batch removal metrics for
#    WT(wild type) and AD brain separately, then took the average
#     of metrics from two disease status group.

# Example
mouse_brain_divide(paired_latent_metrics,[adata1,adata2], method, 'louvain', batch,label, outfile)