cellflow.model.CellFlow.predict

CellFlow.predict(adata, covariate_data, sample_rep=None, condition_id_key=None, key_added_prefix=None, rng=None, **kwargs)

Predict perturbation responses.

Parameters:

• adata (AnnData) – An AnnData object with the source representation.

• covariate_data (DataFrame) – Covariate data defining the condition to predict. This DataFrame should have the same columns as obs of cellflow.model.CellFlow.adata, and as registered in cellflow.model.CellFlow.data_manager.

• sample_rep (str | None) – Key in obsm where the sample representation is stored or 'X' to use X. If None, the key is assumed to be the same as for the training data.

• condition_id_key (str | None) – Key in 'covariate_data' defining the condition name.

• key_added_prefix (str | None) – If not None, prefix to store the prediction in obsm. If None, the predictions are not stored, and the predictions are returned as a dict.

• rng (ndarray[tuple[Any, ...], dtype[float64]] | None) – Random number generator. If None and cellflow.model.CellFlow.conditino_mode is 'stochastic', the condition vector will be the mean of the learnt distributions, otherwise samples from the distribution.

• kwargs (Any) – Keyword arguments for the predict function, i.e. cellflow.solvers.OTFlowMatching.predict() or cellflow.solvers.GENOT.predict().

Return type:

dict[str, ndarray[tuple[Any, ...], dtype[float64]]] | None

Returns:

: If 'key_added_prefix' is None, a dict with the predicted sample representation for each perturbation, otherwise stores the predictions in obsm and returns None.