Generate synthetic single-cell datasets#

Module to generate synthetic datasets with ambient contamination

Classes:

`scrnaseq`(n_cells, n_celltypes, n_features[, ...])	Generate synthetic single-cell RNAseq data with ambient contamination
`citeseq`(n_cells, n_celltypes, n_features[, ...])	Generate synthetic ADT count data for CITE-seq with ambient contamination
`cropseq`(n_cells, n_celltypes, n_features)	Generate synthetic sgRNA count data for scCRISPRseq with ambient contamination

class scar.main._data_generater.scrnaseq(n_cells, n_celltypes, n_features, n_total_molecules=8000, capture_rate=0.7)#

Generate synthetic single-cell RNAseq data with ambient contamination

Parameters

n_cells (int) – number of cells
n_celltypes (int) – number of cell types
n_features (int) – number of features (mRNA)
n_total_molecules (int, optional) – total molecules per cell, by default 8000
capture_rate (float, optional) – the probability of being captured by beads, by default 0.7

Examples

import numpy as np
from scar import data_generator

n_features = 1000  # 1000 genes, bad visualization with too big number
n_cells = 6000  # cells
n_total_molecules = 20000 # total mRNAs
n_celltypes = 8  # cell types

np.random.seed(8)
scRNAseq = data_generator.scrnaseq(n_cells, n_celltypes, n_features, n_total_molecules=n_total_molecules)
scRNAseq.generate(dirichlet_concentration_hyper=1)
scRNAseq.heatmap(vmax=5)

Attributes:

`n_cells`	int, number of cells
`n_celltypes`	int, number of cell types
`n_features`	int, number of features (mRNA, sgRNA, ADT, tag, CMO, and etc.)
`n_total_molecules`	int, number of total molecules per cell
`capture_rate`	float, the probability of being captured by beads
`obs_count`	vector, observed counts
`ambient_profile`	vector, the probability of occurrence of each ambient transcript
`cell_identity`	matrix, the onehot expression of the identity of cell types
`noise_ratio`	vector, contamination level per cell
`celltype`	vector, the identity of cell types
`ambient_signals`	matrix, the real ambient signals
`native_signals`	matrix, the real native signals
`native_profile`	matrix, the frequencies of the real native signals
`total_counts`	vector, the total observed counts per cell
`empty_droplets`	matrix, synthetic cell-free droplets

Methods:

`generate`([dirichlet_concentration_hyper])	Generate a synthetic scRNAseq dataset.
`heatmap`([feature_type, return_obj, figsize, ...])	Heatmap of synthetic data.

n_cells#: int, number of cells

n_celltypes#: int, number of cell types

n_features#: int, number of features (mRNA, sgRNA, ADT, tag, CMO, and etc.)

n_total_molecules#: int, number of total molecules per cell

capture_rate#: float, the probability of being captured by beads

obs_count#: vector, observed counts

ambient_profile#: vector, the probability of occurrence of each ambient transcript

cell_identity#: matrix, the onehot expression of the identity of cell types

noise_ratio#: vector, contamination level per cell

celltype#: vector, the identity of cell types

ambient_signals#: matrix, the real ambient signals

native_signals#: matrix, the real native signals

native_profile#: matrix, the frequencies of the real native signals

total_counts#: vector, the total observed counts per cell

empty_droplets#: matrix, synthetic cell-free droplets

generate(dirichlet_concentration_hyper=0.05)#

Generate a synthetic scRNAseq dataset.

Parameters: dirichlet_concentration_hyper (None or real, optional) – the concentration hyperparameters of dirichlet distribution. Determining the sparsity of native signals. If None, 1 / n_features, by default 0.005.
Return type: After running, several attributes are added

heatmap(feature_type='mRNA', return_obj=False, figsize=(12, 4), vmin=0, vmax=10)#

Heatmap of synthetic data.

Parameters

feature_type (str, optional) – the feature types, by default “mRNA”
return_obj (bool, optional) – whether to output figure object, by default False
figsize (tuple, optional) – figure size, by default (15, 5)
vmin (int, optional) – colorbar minimum, by default 0
vmax (int, optional) – colorbar maximum, by default 10

Returns

if return_obj, return a fig object

Return type

fig object

class scar.main._data_generater.citeseq(n_cells, n_celltypes, n_features, n_total_molecules=8000, capture_rate=0.7)#

Generate synthetic ADT count data for CITE-seq with ambient contamination

Parameters

n_cells (int) – number of cells
n_celltypes (int) – number of cell types
n_features (int) – number of distinct antibodies (ADTs)
n_total_molecules (int, optional) – number of total molecules, by default 8000
capture_rate (float, optional) – the probabilities of being captured by beads, by default 0.7

Examples

import numpy as np
from scar import data_generator

n_features = 50  # 50 ADTs
n_cells = 6000  # 6000 cells
n_celltypes = 6  # cell types

# generate a synthetic ADT count dataset
np.random.seed(8)
citeseq = data_generator.citeseq(n_cells, n_celltypes, n_features)
citeseq.generate()
citeseq.heatmap()

Methods:

`generate`([dirichlet_concentration_hyper])	Generate a synthetic ADT dataset.
`heatmap`([feature_type, return_obj, figsize, ...])	Heatmap of synthetic data.

generate(dirichlet_concentration_hyper=None)#

Generate a synthetic ADT dataset.

Parameters: dirichlet_concentration_hyper (None or real, optional) – the concentration hyperparameters of dirichlet distribution. If None, 1 / n_features, by default None
Return type: After running, several attributes are added

heatmap(feature_type='ADT', return_obj=False, figsize=(12, 4), vmin=0, vmax=10)#

Heatmap of synthetic data.

Parameters

feature_type (str, optional) – the feature types, by default “ADT”
return_obj (bool, optional) – whether to output figure object, by default False
figsize (tuple, optional) – figure size, by default (15, 5)
vmin (int, optional) – colorbar minimum, by default 0
vmax (int, optional) – colorbar maximum, by default 10

Returns

if return_obj, return a fig object

Return type

fig object

class scar.main._data_generater.cropseq(n_cells, n_celltypes, n_features)#

Generate synthetic sgRNA count data for scCRISPRseq with ambient contamination

Parameters

n_cells (int) – number of cells
n_celltypes (int) – number of cell types
n_features (int) – number of dinstinct sgRNAs
library_pattern (str, optional) –
the pattern of sgRNA libraries, three possibilities:

”uniform” - each sgRNA has equal frequency in the libraries

”pyramid” - a few sgRNAs have significantly higher frequencies in the libraries

”reverse_pyramid” - a few sgRNAs have significantly lower frequencies in the libraries

By default “pyramid”.
noise_ratio (float, optional) – global contamination level, by default 0.005
average_counts_per_cell (int, optional) – average total sgRNA counts per cell, by default 2000
doublet_rate (int, optional) – doublet rate, by default 0
missing_rate (int, optional) – the fraction of droplets which have zero sgRNAs integrated, by default 0

Examples

import numpy as np
from scar import data_generator

n_features = 100  # 100 sgRNAs in the libraries
n_cells = 6000  # 6000 cells
n_celltypes = 1  # single cell line

# generate a synthetic sgRNA count dataset
np.random.seed(8)
cropseq = data_generator.cropseq(n_cells, n_celltypes, n_features)
cropseq.generate(noise_ratio=0.98)
cropseq.heatmap(vmax=6)

Attributes:

sgrna_freq

vector, sgRNA frequencies in the libraries

Methods:

`generate`([dirichlet_concentration_hyper, ...])	Generate a synthetic sgRNA count dataset.
`heatmap`([feature_type, return_obj, figsize, ...])	Heatmap of synthetic data.

sgrna_freq#: vector, sgRNA frequencies in the libraries

generate(dirichlet_concentration_hyper=None, library_pattern='pyramid', noise_ratio=0.96, average_counts_per_cell=2000, doublet_rate=0, missing_rate=0)#

Generate a synthetic sgRNA count dataset.

Parameters

library_pattern (str, optional) – library pattern, by default “pyramid”
noise_ratio (float, optional) – global contamination level, by default 0.005
average_counts_per_cell (int, optional) – average total sgRNA counts per cell, by default 2000
doublet_rate (int, optional) – doublet rate, by default 0
missing_rate (int, optional) – the fraction of droplets which have zero sgRNAs integrated, by default 0

Return type

After running, several attributes are added

heatmap(feature_type='sgRNAs', return_obj=False, figsize=(12, 4), vmin=0, vmax=7)#

Heatmap of synthetic data.

Parameters

feature_type (str, optional) – the feature types, by default “sgRNAs”
return_obj (bool, optional) – whether to output figure object, by default False
figsize (tuple, optional) – figure size, by default (15, 5)
vmin (int, optional) – colorbar minimum, by default 0
vmax (int, optional) – colorbar maximum, by default 10

Returns

if return_obj, return a fig object

Return type

fig object