tdm.dataset.PolynomialDataset#

class tdm.dataset.PolynomialDataset(ds: Dataset, degree: int = 2, feature_cell_types: list[str] | None = None, include_bias: bool = True, scale_features: bool = False, log_transform: bool = False)[source]#

Performs feature transformations and constructs polynomial features from a dataset.

__init__(ds: Dataset, degree: int = 2, feature_cell_types: list[str] | None = None, include_bias: bool = True, scale_features: bool = False, log_transform: bool = False) → None[source]#

Performs feature transformations and constructs polynomial features from a dataset.

Parameters:

  • ds (Dataset) – dataset whose features we transform

  • degree (int, optional) – order of polynomial interactions to construct. Defaults to 2.

  • feature_cell_types (list[str] | None, optional) – selects a subset of cell types for features. Defaults to all cell types present.

  • include_bias (bool, optional) – add a bias term to the features. Defaults to True.

  • scale_features (bool, optional) – standardize features (subtract mean and divide by standard deviation). Defaults to False.

  • log_transform (bool, optional) – log2(1+x) transformation of features. Defaults to False.

Examples

>>> # nds is a NeighborsDataset instance
>>> pds = PolynomialDataset(nds, degree=2)
construct_features_from_counts(cell_counts: dict | DataFrame, target_cell: str, **kwargs) → DataFrame[source]#

Constructs features compatible with construct_polynomial_features Cell vals is in raw counts (i.e “64” cells, not the log-value: 6)!

construct_polynomial_features(neighbor_features: DataFrame, scale_features: bool, fit_scaler: bool = False, target_cell: str | None = None) → tuple[DataFrame, StandardScaler][source]#
Parameters:

  • neighbor_features – output of fetch() from a NeighborsDataset

  • scale_features – True performs standard scaling on the resulting polynomial features

  • fit_scaler – True fits a new standard scaler to the resulting polynomial features

  • target_cell – cell we’re modelling, used to fetch the correct StandardScaler

Returns:

(a polynomial features dataframe, a standard scaler)

Return type:

tuple

intercept_idx()[source]#