socube package

Submodules

socube.cube module

class socube.cube.SoCube(data_path: str, home_path: Optional[str] = None, metric: str = 'correlation', dtype: str = 'float32', verbose: int = 1, **kwargs)

Bases: object

Abstract class of SoCube.

Parameters

  • data_path (str) – Path of original scRNA-seq data file

  • home_path (str, optional) – Path of output home directory. If not specified, it will be set to the same directory as data_path.

  • metric (str, optional) – Metric to use for distance calculation. supported: ‘correlation’, ‘cosine’, ‘euclidean’, ‘manhattan’, etc. default is ‘correlation’.

  • dtype (str, optional) – Data type of output data. isupported: ‘float32’, ‘float64’, etc. default is ‘float32’.

  • verbose (int, optional) – Verbose level. default is 1.

static available() → dict

Return available map types

batchTransform(samples: pandas.core.frame.DataFrame, data_dir: Optional[str] = None, use_index: bool = False, **kwargs) → numpy.ndarray

The sample data batch is subjected to feature transformation by socube to obtain the feature 3d matrix in the form of NCHW

Parameters

  • samples (pd.DataFrame) – The sample data to be transformed, row as samples, columns as features

  • data_dir (str) – Path to save samples. If given, it will save samples

  • use_index (bool) – If True, it will save samples with using numerical index as filename. such as 0.npy otherwise, sample name used instead.

Returns

Return type

The generated embedding feature matrix and gene matching

abstract fit(**kwargs)

all subclass must implement it

property grid

Return the grid data

static instance(map_type: str, data_path: str, home_path: Optional[str] = None, metric: str = 'correlation', dtype: str = 'float32', verbose: int = 1, **kwargs) → SoCubeType

SoCube is designed using the factory pattern, and this method is used to create SoCube instance objects

Parameters

  • map_type (str) – Type of map to use. supported: ‘pca’

  • data_path (str) – Path of original scRNA-seq data file

  • home_path (str, optional) – Path of output home directory. If not specified, it will be set to the same directory as data_path.

  • metric (str, optional) – Metric to use for distance calculation. supported: ‘correlation’, ‘cosine’, ‘euclidean’, ‘manhattan’, etc. default is ‘correlation’.

  • dtype (str, optional) – Data type of output data. isupported: ‘float32’, ‘float64’, etc. default is ‘float32’.

  • verbose (int, optional) – Verbose level. default is 1.

Returns

Return type

SoCube instance object

abstract load(cube_path: str) → bool

Load pretrained Socube

property scatter

Return the scatter data

property shape

Return the socube 2d grid’s shape

abstract transform(sample: pandas.core.series.Series, **kwargs) → numpy.ndarray

Transform a droplet sample vector to its embedding form. for batch transform, see batchTransform api.

Parameters

sample (pd.Series) – a droplet sample vector

Returns

Return type

embedding form

Module contents

socube.main(*args: str)

Entrypoint function of socube commandline interface. For parameter details, just run it without any args to print help info. See also at https://docs.idrblab.cn/socube.

Examples

>>> from socube import main
>>> main() # print hellp info