API#

found.find.HiDDEN(x, /, cond_col, control_val, algo=Pipeline(dimr_fn=<function run_pca>, regr_fn=<function reg_logit>, binr_fn=<function bin_kmeans>, cachable_dimr=True, strict=True), **kwargs)[source]#

runs HiDDEN on a given AnnData object.

Parameters:

  • x (AnnData) – input AnnData object

  • cond_col (str) – string indicating obs column in adata representing condition value

  • control_val (object) – value representing the control condition in the provided condition column

  • algo (Pipeline) – algorithm pipeline (expected to use parameter V as original condition annotation)

  • kwargs – additional variables to pass into pipeline

Returns:

tuple[Series, Series] – 2-tuple consisting of:

  • 1-d series of prediction outputs by model

  • binarized labels from prediction values

found.find.HiDDENg(x, /, cond_col, control_val, group_by, algo=Pipeline(dimr_fn=<function run_pca>, regr_fn=<function reg_logit>, binr_fn=<function bin_kmeans>, cachable_dimr=True, strict=True), which_grouped=None, grp_specific_args=None, **kwargs)[source]#

runs HiDDEN on a given AnnData object, given some set of grouping factors.

Parameters:

  • x (AnnData) – input AnnData object

  • cond_col (str) – string indicating obs column in adata representing condition value

  • control_val (object) – value representing the control condition in the provided condition column

  • group_by (str | tuple[str]) – set of column names in x.obs specifying grouping

  • algo (Pipeline) – algorithm pipeline (expected to use parameter V as original condition annotation)

  • which_grouped (str | tuple[str] | list[str] | dict[str, Callable[[object, ndarray[tuple[int], dtype[integer]]], object]] | None) – set of pipeline arguments which should be indexed by grouping (set to None to group all pipeline arguments which support indexing)

  • grp_specific_args (Mapping[object, dict[str, object]] | None) – any additional arguments that are to be provided on a group-specific basis

  • kwargs – additional variables to pass into pipeline across every group

Returns:

tuple[Series, Series] – 2-tuple consisting of:

  • 1-d series of prediction outputs by model

  • binarized labels from prediction values

found.find.HiDDENt(x, /, cond_col, control_val, tuner, algo=Pipeline(dimr_fn=<function run_pca>, regr_fn=<function reg_logit>, binr_fn=<function bin_kmeans>, cachable_dimr=True, strict=True), **kwargs)[source]#

runs HiDDEN on a given AnnData object, trying to optimize for a specific set of dimensions using provided Tuner.

Parameters:

  • x (AnnData) – input AnnData object

  • cond_col (str) – string indicating obs column in adata representing condition value

  • control_val (object) – value representing the control condition in the provided condition column

  • tuner (Tuner[TypeVar(P), TypeVar(S)]) – provided tuner which attempts to optimize pipeline for a specific hyperparameter

  • algo (Pipeline) – algorithm pipeline (expected to use parameter V as original condition annotation)

  • kwargs – additional variables to pass into pipeline

Returns:

tuple[TypeVar(P), Mapping[TypeVar(P), tuple[Series, Series, TypeVar(S)]]] – 2-tuple consisting of:

  • selected optimal pipeline hyperparameters

  • dictionary with keys being pipeline hyperparameters and values being a 3-tuple of:

    • 1-d series of prediction outputs by model

    • model adjusted labels

    • optional scoring value generated by tuner for specific hyperparameter configuration

found.find.HiDDENgt(x, /, cond_col, control_val, group_by, tuner, algo=Pipeline(dimr_fn=<function run_pca>, regr_fn=<function reg_logit>, binr_fn=<function bin_kmeans>, cachable_dimr=True, strict=True), which_grouped=None, grp_specific_args=None, **kwargs)[source]#

runs HiDDEN on a given AnnData object, given some set of grouping factors, trying to optimize for a specific set of dimensions using provided Tuner.

Parameters:

  • x (AnnData) – input AnnData object

  • cond_col (str) – string indicating obs column in adata representing condition value

  • control_val (object) – value representing the control condition in the provided condition column

  • group_by (str | tuple[str]) – set of column names in x.obs specifying grouping

  • tuner (Tuner[TypeVar(P), TypeVar(S)]) – provided tuner which attempts to optimize pipeline for a specific hyperparameter

  • algo (Pipeline) – algorithm pipeline (expected to use parameter V as original condition annotation)

  • which_grouped (str | tuple[str] | list[str] | dict[str, Callable[[object, ndarray[tuple[int], dtype[integer]]], object]] | None) – set of pipeline arguments which should be indexed by grouping (set to None to group all pipeline arguments which support indexing)

  • grp_specific_args (Mapping[TypeVar(G), dict[str, object]] | None) – any additional arguments that are to be provided on a group-specific basis

  • kwargs – additional variables to pass into pipeline across every group

Returns:

tuple[Mapping[TypeVar(G), TypeVar(P)], ByParamAccessor[TypeVar(G), TypeVar(P), TypeVar(S)], Callable[[TypeVar(G)], Mapping[TypeVar(P), tuple[Series, Series, TypeVar(S)]]]] – 3-tuple consisting of:

  • mapping from each group to selected hyper-parameter for that group

  • accessor function which given a mapping of groups to hyper-parameters (and an optional default hyper-parameter for unspecified groups), returns a 3-tuple of:

    • 1-d series of prediction outputs by model, ordered by their original order within the provided

    • model adjusted labels

    • mapping of group to score value given provided configuration

  • accessor function which given a specific group, returns a HiDDENt-style output dictionary for just that group