pertpy 0.11.3__py3-none-any.whl → 0.11.5__py3-none-any.whl

pertpy/__init__.py

@@ -2,7 +2,7 @@
 
 __author__ = "Lukas Heumos"
 __email__ = "lukas.heumos@posteo.net"
-__version__ = "0.11.3"
+__version__ = "0.11.5"
 
 import warnings
 

pertpy/tools/_coda/_base_coda.py

@@ -16,7 +16,7 @@ from jax import config, random
 from lamin_utils import logger
 from matplotlib import cm, rcParams
 from matplotlib import image as mpimg
-from matplotlib.colors import ListedColormap
+from matplotlib.colors import Colormap
 from mudata import MuData
 from numpyro.infer import HMC, MCMC, NUTS, initialization
 from rich import box, print
@@ -34,7 +34,6 @@ if TYPE_CHECKING:
     from ete4 import Tree
     from jax._src.typing import Array
     from matplotlib.axes import Axes
-    from matplotlib.colors import Colormap
     from matplotlib.figure import Figure
 
 config.update("jax_enable_x64", True)
@@ -1141,7 +1140,7 @@ class CompositionalModel2(ABC):
         level_names: list[str],
         figsize: tuple[float, float] | None = None,
         dpi: int | None = 100,
-        palette: ListedColormap | None = cm.tab20,
+        palette: str | Colormap | None = cm.tab20,
         show_legend: bool | None = True,
     ) -> plt.Axes:
         """Plots a stacked barplot for one (discrete) covariate.
@@ -1156,12 +1155,15 @@ class CompositionalModel2(ABC):
             level_names: Names of the covariate's levels
             figsize: Figure size (matplotlib).
             dpi: Resolution in DPI (matplotlib).
-            palette: The color map for the barplot.
+            palette: The color map (name) for the barplot.
             show_legend: If True, adds a legend.
 
         Returns:
             A :class:`~matplotlib.axes.Axes` object
         """
+        if isinstance(palette, str):
+            palette = getattr(cm, palette)
+
         n_bars, n_types = y.shape
 
         figsize = rcParams["figure.figsize"] if figsize is None else figsize
@@ -1202,7 +1204,7 @@ class CompositionalModel2(ABC):
         feature_name: str,
         *,
         modality_key: str = "coda",
-        palette: ListedColormap | None = cm.tab20,
+        palette: str | Colormap | None = cm.tab20,
         show_legend: bool | None = True,
         level_order: list[str] = None,
         figsize: tuple[float, float] | None = None,
@@ -1217,7 +1219,7 @@ class CompositionalModel2(ABC):
             modality_key: If data is a MuData object, specify which modality to use.
             figsize: Figure size.
             dpi: Dpi setting.
-            palette: The matplotlib color map for the barplot.
+            palette: The matplotlib color map (name) for the barplot.
             show_legend: If True, adds a legend.
             level_order: Custom ordering of bars on the x-axis.
             {common_plot_args}
@@ -1299,7 +1301,7 @@ class CompositionalModel2(ABC):
         plot_facets: bool = True,
         plot_zero_covariate: bool = True,
         plot_zero_cell_type: bool = False,
-        palette: str | ListedColormap | None = cm.tab20,
+        palette: str | Colormap | None = cm.tab20,
         level_order: list[str] = None,
         args_barplot: dict | None = None,
         figsize: tuple[float, float] | None = None,
@@ -1321,7 +1323,7 @@ class CompositionalModel2(ABC):
             plot_zero_cell_type: If True, plot cell type that have zero effect. If False, do not plot.
             figsize: Figure size.
             dpi: Figure size.
-            palette: The seaborn color map for the barplot.
+            palette: The seaborn color map (name) for the barplot.
             level_order: Custom ordering of bars on the x-axis.
             args_barplot: Arguments passed to sns.barplot.
             {common_plot_args}
@@ -1397,7 +1399,7 @@ class CompositionalModel2(ABC):
 
         # If plot as facets, create a FacetGrid and map barplot to it.
         if plot_facets:
-            if isinstance(palette, ListedColormap):
+            if isinstance(palette, Colormap):
                 palette = np.array([palette(i % palette.N) for i in range(len(plot_df["Cell Type"].unique()))]).tolist()
             if figsize is not None:
                 height = figsize[0]
@@ -1437,7 +1439,7 @@ class CompositionalModel2(ABC):
         else:
             _, ax = plt.subplots(figsize=figsize, dpi=dpi)
             if len(covariate_names) == 1:
-                if isinstance(palette, ListedColormap):
+                if isinstance(palette, Colormap):
                     palette = np.array(
                         [palette(i % palette.N) for i in range(len(plot_df["Cell Type"].unique()))]
                     ).tolist()
@@ -1451,7 +1453,7 @@ class CompositionalModel2(ABC):
                 )
                 ax.set_title(covariate_names[0])
             else:
-                if isinstance(palette, ListedColormap):
+                if isinstance(palette, Colormap):
                     palette = np.array([palette(i % palette.N) for i in range(len(covariate_names))]).tolist()
                 sns.barplot(
                     data=plot_df,
@@ -1485,7 +1487,7 @@ class CompositionalModel2(ABC):
         cell_types: list | None = None,
         args_boxplot: dict | None = None,
         args_swarmplot: dict | None = None,
-        palette: str | None = "Blues",
+        palette: str | Colormap | None = "Blues",
         show_legend: bool | None = True,
         level_order: list[str] = None,
         figsize: tuple[float, float] | None = None,
@@ -1510,7 +1512,7 @@ class CompositionalModel2(ABC):
             args_swarmplot: Arguments passed to sns.swarmplot.
             figsize: Figure size.
             dpi: Dpi setting.
-            palette: The seaborn color map for the barplot.
+            palette: The seaborn color map (name) for the barplot.
             show_legend: If True, adds a legend.
             level_order: Custom ordering of bars on the x-axis.
             {common_plot_args}
@@ -1535,6 +1537,8 @@ class CompositionalModel2(ABC):
             args_swarmplot = {}
         if isinstance(data, MuData):
             data = data[modality_key]
+        if isinstance(palette, Colormap):
+            palette = list(palette(range(len(data.obs[feature_name].unique()))))
 
         # y scale transformations
         if y_scale == "relative":
@@ -2104,7 +2108,7 @@ class CompositionalModel2(ABC):
         modality_key_1: str = "rna",
         modality_key_2: str = "coda",
         color_map: Colormap | str | None = None,
-        palette: str | Sequence[str] | None = None,
+        palette: str | Sequence[str] | Colormap | None = None,
         ax: Axes = None,
         return_fig: bool = False,
         **kwargs,
@@ -2122,7 +2126,7 @@ class CompositionalModel2(ABC):
             modality_key_1: Key to the cell-level AnnData in the MuData object.
             modality_key_2: Key to the aggregated sample-level AnnData object in the MuData object.
             color_map: The color map to use for plotting.
-            palette: The color palette to use for plotting.
+            palette: The color palette (name) to use for plotting.
             ax: A matplotlib axes object. Only works if plotting a single component.
             {common_plot_args}
             **kwargs: All other keyword arguments are passed to `scanpy.plot.umap()`
@@ -2154,9 +2158,6 @@ class CompositionalModel2(ABC):
             >>> tasccoda_model.run_nuts(
             ...     tasccoda_data, modality_key="coda", rng_key=1234, num_samples=10000, num_warmup=1000
             ... )
-            >>> tasccoda_model.run_nuts(
-            ...     tasccoda_data, modality_key="coda", rng_key=1234, num_samples=10000, num_warmup=1000
-            ... )
             >>> sc.tl.umap(tasccoda_data["rna"])
             >>> tasccoda_model.plot_effects_umap(tasccoda_data,
             >>>                         effect_name=["effect_df_condition[T.Salmonella]",
@@ -2173,6 +2174,10 @@ class CompositionalModel2(ABC):
         data_coda = mdata[modality_key_2]
         if isinstance(effect_name, str):
             effect_name = [effect_name]
+        if isinstance(palette, Colormap):
+            palette = {
+                cluster: palette(i % palette.N) for i, cluster in enumerate(data_rna.obs[cluster_key].unique().tolist())
+            }
         for _, effect in enumerate(effect_name):
             data_rna.obs[effect] = [data_coda.varm[effect].loc[f"{c}", "Effect"] for c in data_rna.obs[cluster_key]]
         if kwargs.get("vmin"):

pertpy/tools/_dialogue.py

@@ -80,7 +80,7 @@ class Dialogue:
         Returns:
             A Pandas DataFrame of pseudobulk counts
         """
-        # TODO: Replace with decoupler's implementation
+        # TODO: Replace with scanpy get implementation
         pseudobulk = {"Genes": adata.var_names.values}
 
         for category in adata.obs.loc[:, groupby].cat.categories:

pertpy/tools/_differential_gene_expression/_base.py

@@ -572,9 +572,7 @@ class MethodBase(ABC):
         if any(adata.obs[[groupby, pairedby]].value_counts() > 1):
             logger.info("Performing pseudobulk for paired samples")
             ps = PseudobulkSpace()
-            adata = ps.compute(
-                adata, target_col=groupby, groups_col=pairedby, layer_key=layer, mode="sum", min_cells=1, min_counts=1
-            )
+            adata = ps.compute(adata, target_col=groupby, groups_col=pairedby, layer_key=layer, mode="sum")
 
         X = adata.layers[layer] if layer is not None else adata.X
         with contextlib.suppress(AttributeError):

pertpy/tools/_distances/_distances.py

@@ -3,6 +3,7 @@ from __future__ import annotations
 from abc import ABC, abstractmethod
 from typing import TYPE_CHECKING, Literal, NamedTuple
 
+import jax
 import numpy as np
 import pandas as pd
 from numba import jit
@@ -685,6 +686,7 @@ class WassersteinDistance(AbstractDistance):
     def __init__(self) -> None:
         super().__init__()
         self.accepts_precomputed = False
+        self.solver = jax.jit(Sinkhorn())
 
     def __call__(self, X: np.ndarray, Y: np.ndarray, **kwargs) -> float:
         X = np.asarray(X, dtype=np.float64)
@@ -699,8 +701,7 @@ class WassersteinDistance(AbstractDistance):
 
     def solve_ot_problem(self, geom: Geometry, **kwargs):
         ot_prob = LinearProblem(geom)
-        solver = Sinkhorn()
-        ot = solver(ot_prob, **kwargs)
+        ot = self.solver(ot_prob, **kwargs)
         cost = float(ot.reg_ot_cost)
 
         # Check for NaN or invalid cost

pertpy/tools/_milo.py

@@ -364,19 +364,32 @@ class Milo:
             # Set up rpy2 to run edgeR
             edgeR, limma, stats, base = self._setup_rpy2()
 
+            import rpy2.robjects as ro
+            from rpy2.robjects import numpy2ri, pandas2ri
+            from rpy2.robjects.conversion import localconverter
+            from rpy2.robjects.vectors import FloatVector
+
             # Define model matrix
             if not add_intercept or model_contrasts is not None:
                 design = design + " + 0"
-            model = stats.model_matrix(object=stats.formula(design), data=design_df)
+            design_df = design_df.astype(dict.fromkeys(design_df.select_dtypes(exclude=["number"]).columns, "category"))
+            with localconverter(ro.default_converter + pandas2ri.converter):
+                design_r = pandas2ri.py2rpy(design_df)
+            formula_r = stats.formula(design)
+            model = stats.model_matrix(object=formula_r, data=design_r)
 
             # Fit NB-GLM
-            dge = edgeR.DGEList(counts=count_mat[keep_nhoods, :][:, keep_smp], lib_size=lib_size[keep_smp])
+            counts_filtered = count_mat[np.ix_(keep_nhoods, keep_smp)]
+            lib_size_filtered = lib_size[keep_smp]
+            count_mat_r = numpy2ri.py2rpy(counts_filtered)
+            lib_size_r = FloatVector(lib_size_filtered)
+            dge = edgeR.DGEList(counts=count_mat_r, lib_size=lib_size_r)
             dge = edgeR.calcNormFactors(dge, method="TMM")
             dge = edgeR.estimateDisp(dge, model)
             fit = edgeR.glmQLFit(dge, model, robust=True)
-
             # Test
-            n_coef = model.shape[1]
+            model_np = np.array(model)
+            n_coef = model_np.shape[1]
             if model_contrasts is not None:
                 r_str = """
                 get_model_cols <- function(design_df, design){
@@ -387,32 +400,36 @@ class Milo:
                 from rpy2.robjects.packages import STAP
 
                 get_model_cols = STAP(r_str, "get_model_cols")
-                model_mat_cols = get_model_cols.get_model_cols(design_df, design)
-                model_df = pd.DataFrame(model)
+                with localconverter(ro.default_converter + numpy2ri.converter + pandas2ri.converter):
+                    model_mat_cols = get_model_cols.get_model_cols(design_df, design)
+                with localconverter(ro.default_converter + pandas2ri.converter + numpy2ri.converter):
+                    model_df = pandas2ri.rpy2py(model)
+                model_df = pd.DataFrame(model_df)
                 model_df.columns = model_mat_cols
                 try:
-                    mod_contrast = limma.makeContrasts(contrasts=model_contrasts, levels=model_df)
+                    with localconverter(ro.default_converter + pandas2ri.converter):
+                        mod_contrast = limma.makeContrasts(contrasts=model_contrasts, levels=model_df)
                 except ValueError:
                     logger.error("Model contrasts must be in the form 'A-B' or 'A+B'")
                     raise
-                res = base.as_data_frame(
-                    edgeR.topTags(edgeR.glmQLFTest(fit, contrast=mod_contrast), sort_by="none", n=np.inf)
-                )
+                with localconverter(ro.default_converter + pandas2ri.converter + numpy2ri.converter):
+                    res = base.as_data_frame(
+                        edgeR.topTags(edgeR.glmQLFTest(fit, contrast=mod_contrast), sort_by="none", n=np.inf)
+                    )
             else:
-                res = base.as_data_frame(edgeR.topTags(edgeR.glmQLFTest(fit, coef=n_coef), sort_by="none", n=np.inf))
-
-            from rpy2.robjects import conversion
-
-            res = conversion.rpy2py(res)
+                with localconverter(ro.default_converter + numpy2ri.converter + pandas2ri.converter):
+                    res = base.as_data_frame(
+                        edgeR.topTags(edgeR.glmQLFTest(fit, coef=n_coef), sort_by="none", n=np.inf)
+                    )
             if not isinstance(res, pd.DataFrame):
                 res = pd.DataFrame(res)
-
+            # The columns of res looks like e.g. table.A, table.B, so remove the prefix
+            res.columns = [col.replace("table.", "") for col in res.columns]
         # Save outputs
         res.index = sample_adata.var_names[keep_nhoods]  # type: ignore
         if any(col in sample_adata.var.columns for col in res.columns):
             sample_adata.var = sample_adata.var.drop(res.columns, axis=1)
         sample_adata.var = pd.concat([sample_adata.var, res], axis=1)
-
         # Run Graph spatial FDR correction
         self._graph_spatial_fdr(sample_adata, neighbors_key=adata.uns["nhood_neighbors_key"])
 
@@ -657,11 +674,8 @@ class Milo:
         self,
     ):
         """Set up rpy2 to run edgeR."""
-        from rpy2.robjects import numpy2ri, pandas2ri
         from rpy2.robjects.packages import importr
 
-        numpy2ri.activate()
-        pandas2ri.activate()
         edgeR = self._try_import_bioc_library("edgeR")
         limma = self._try_import_bioc_library("limma")
         stats = importr("stats")
@@ -1007,6 +1021,8 @@ class Milo:
         subset_nhoods: list[str] = None,
         log_counts: bool = False,
         return_fig: bool = False,
+        ax=None,
+        show: bool = True,
     ) -> Figure | None:
         """Plot boxplot of cell numbers vs condition of interest.
 
@@ -1036,18 +1052,36 @@ class Milo:
         pl_df = pd.merge(pl_df, nhood_adata.var)
         pl_df["log_n_cells"] = np.log1p(pl_df["n_cells"])
         if not log_counts:
-            sns.boxplot(data=pl_df, x=test_var, y="n_cells", color="lightblue")
-            sns.stripplot(data=pl_df, x=test_var, y="n_cells", color="black", s=3)
-            plt.ylabel("# cells")
+            sns.boxplot(data=pl_df, x=test_var, y="n_cells", color="lightblue", ax=ax)
+            sns.stripplot(data=pl_df, x=test_var, y="n_cells", color="black", s=3, ax=ax)
+            if ax:
+                ax.set_ylabel("# cells")
+            else:
+                plt.ylabel("# cells")
         else:
-            sns.boxplot(data=pl_df, x=test_var, y="log_n_cells", color="lightblue")
-            sns.stripplot(data=pl_df, x=test_var, y="log_n_cells", color="black", s=3)
-            plt.ylabel("log(# cells + 1)")
+            sns.boxplot(data=pl_df, x=test_var, y="log_n_cells", color="lightblue", ax=ax)
+            sns.stripplot(data=pl_df, x=test_var, y="log_n_cells", color="black", s=3, ax=ax)
+            if ax:
+                ax.set_ylabel("log(# cells + 1)")
+            else:
+                plt.ylabel("log(# cells + 1)")
 
-        plt.xticks(rotation=90)
-        plt.xlabel(test_var)
+        if ax:
+            ax.tick_params(axis="x", rotation=90)
+            ax.set_xlabel(test_var)
+        else:
+            plt.xticks(rotation=90)
+            plt.xlabel(test_var)
 
         if return_fig:
             return plt.gcf()
-        plt.show()
+
+        if ax is None:
+            plt.show()
+
+        if return_fig:
+            return plt.gcf()
+        if show:
+            plt.show()
+
         return None

pertpy/tools/_mixscape.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import copy
+import warnings
 from collections import OrderedDict
 from typing import TYPE_CHECKING, Literal
 
@@ -10,11 +11,12 @@ import pandas as pd
 import scanpy as sc
 import seaborn as sns
 from fast_array_utils.stats import mean, mean_var
+from pandas.errors import PerformanceWarning
 from scanpy import get
 from scanpy._utils import _check_use_raw, sanitize_anndata
 from scanpy.plotting import _utils
 from scanpy.tools._utils import _choose_representation
-from scipy.sparse import csr_matrix, spmatrix
+from scipy.sparse import csr_matrix, issparse, spmatrix
 from sklearn.mixture import GaussianMixture
 
 from pertpy._doc import _doc_params, doc_common_plot_args
@@ -103,6 +105,9 @@ class Mixscape:
 
         adata.layers["X_pert"] = adata.X.copy()
 
+        # Work with LIL for efficient indexing but don't store it in AnnData as LIL is not supported anymore
+        X_pert_lil = adata.layers["X_pert"].tolil() if issparse(adata.layers["X_pert"]) else adata.layers["X_pert"]
+
         control_mask = adata.obs[pert_key] == control
 
         if ref_selection_mode == "split_by":
@@ -110,9 +115,8 @@ class Mixscape:
                 split_mask = adata.obs[split_by] == split
                 control_mask_group = control_mask & split_mask
                 control_mean_expr = mean(adata.X[control_mask_group], axis=0)
-                adata.layers["X_pert"][split_mask] = (
-                    np.repeat(control_mean_expr.reshape(1, -1), split_mask.sum(), axis=0)
-                    - adata.layers["X_pert"][split_mask]
+                X_pert_lil[split_mask] = (
+                    np.repeat(control_mean_expr.reshape(1, -1), split_mask.sum(), axis=0) - X_pert_lil[split_mask]
                 )
         else:
             if split_by is None:
@@ -129,49 +133,43 @@ class Mixscape:
 
             for split_mask in split_masks:
                 control_mask_split = control_mask & split_mask
-
                 R_split = representation[split_mask]
                 R_control = representation[np.asarray(control_mask_split)]
-
                 eps = kwargs.pop("epsilon", 0.1)
                 nn_index = NNDescent(R_control, **kwargs)
                 indices, _ = nn_index.query(R_split, k=n_neighbors, epsilon=eps)
-
                 X_control = np.expm1(adata.X[np.asarray(control_mask_split)])
-
                 n_split = split_mask.sum()
                 n_control = X_control.shape[0]
 
                 if batch_size is None:
                     col_indices = np.ravel(indices)
                     row_indices = np.repeat(np.arange(n_split), n_neighbors)
-
                     neigh_matrix = csr_matrix(
                         (np.ones_like(col_indices, dtype=np.float64), (row_indices, col_indices)),
                         shape=(n_split, n_control),
                     )
                     neigh_matrix /= n_neighbors
-                    adata.layers["X_pert"][np.asarray(split_mask)] = (
-                        sc.pp.log1p(neigh_matrix @ X_control) - adata.layers["X_pert"][np.asarray(split_mask)]
+                    X_pert_lil[np.asarray(split_mask)] = (
+                        sc.pp.log1p(neigh_matrix @ X_control) - X_pert_lil[np.asarray(split_mask)]
                     )
                 else:
                     split_indices = np.where(split_mask)[0]
                     for i in range(0, n_split, batch_size):
                         size = min(i + batch_size, n_split)
                         select = slice(i, size)
-
                         batch = np.ravel(indices[select])
                         split_batch = split_indices[select]
-
                         size = size - i
-
                         means_batch = X_control[batch]
                         batch_reshaped = means_batch.reshape(size, n_neighbors, -1)
                         means_batch, _ = mean_var(batch_reshaped, axis=1)
+                        X_pert_lil[split_batch] = np.log1p(means_batch) - X_pert_lil[split_batch]
 
-                        adata.layers["X_pert"][split_batch] = (
-                            np.log1p(means_batch) - adata.layers["X_pert"][split_batch]
-                        )
+        if issparse(X_pert_lil):
+            adata.layers["X_pert"] = X_pert_lil.tocsr()
+        else:
+            adata.layers["X_pert"] = X_pert_lil
 
         if copy:
             return adata
@@ -531,26 +529,29 @@ class Mixscape:
             gene_targets = list(set(adata[split_mask].obs[labels]).difference([control]))
             adata_split = adata[split_mask].copy()
             # find top DE genes between cells with targeting and non-targeting gRNAs
-            sc.tl.rank_genes_groups(
-                adata_split,
-                layer=layer,
-                groupby=labels,
-                groups=gene_targets,
-                reference=control,
-                method=test_method,
-                use_raw=False,
-            )
-            # get DE genes for each target gene
-            for gene in gene_targets:
-                logfc_threshold_mask = (
-                    np.abs(adata_split.uns["rank_genes_groups"]["logfoldchanges"][gene]) >= logfc_threshold
+            with warnings.catch_warnings():
+                warnings.simplefilter("ignore", RuntimeWarning)
+                warnings.simplefilter("ignore", PerformanceWarning)
+                sc.tl.rank_genes_groups(
+                    adata_split,
+                    layer=layer,
+                    groupby=labels,
+                    groups=gene_targets,
+                    reference=control,
+                    method=test_method,
+                    use_raw=False,
                 )
-                de_genes = adata_split.uns["rank_genes_groups"]["names"][gene][logfc_threshold_mask]
-                pvals_adj = adata_split.uns["rank_genes_groups"]["pvals_adj"][gene][logfc_threshold_mask]
-                de_genes = de_genes[pvals_adj < pval_cutoff]
-                if len(de_genes) < min_de_genes:
-                    de_genes = np.array([])
-                perturbation_markers[(category, gene)] = de_genes
+                # get DE genes for each target gene
+                for gene in gene_targets:
+                    logfc_threshold_mask = (
+                        np.abs(adata_split.uns["rank_genes_groups"]["logfoldchanges"][gene]) >= logfc_threshold
+                    )
+                    de_genes = adata_split.uns["rank_genes_groups"]["names"][gene][logfc_threshold_mask]
+                    pvals_adj = adata_split.uns["rank_genes_groups"]["pvals_adj"][gene][logfc_threshold_mask]
+                    de_genes = de_genes[pvals_adj < pval_cutoff]
+                    if len(de_genes) < min_de_genes:
+                        de_genes = np.array([])
+                    perturbation_markers[(category, gene)] = de_genes
 
         return perturbation_markers
 
@@ -711,7 +712,10 @@ class Mixscape:
         if "mixscape_class" not in adata.obs:
             raise ValueError("Please run `pt.tl.mixscape` first.")
         adata_subset = adata[(adata.obs[labels] == target_gene) | (adata.obs[labels] == control)].copy()
-        sc.tl.rank_genes_groups(adata_subset, layer=layer, groupby=labels, method=method)
+        with warnings.catch_warnings():
+            warnings.simplefilter("ignore", RuntimeWarning)
+            warnings.simplefilter("ignore", PerformanceWarning)
+            sc.tl.rank_genes_groups(adata_subset, layer=layer, groupby=labels, method=method)
         sc.pp.scale(adata_subset, max_value=vmax)
         sc.pp.subsample(adata_subset, n_obs=subsample_number)
 
@@ -998,8 +1002,7 @@ class Mixscape:
             ys = keys
 
         if multi_panel and groupby is None and len(ys) == 1:
-            # This is a quick and dirty way for adapting scales across several
-            # keys if groupby is None.
+            # This is a quick and dirty way for adapting scales across several keys if groupby is None.
             y = ys[0]
 
             g = sns.catplot(

pertpy/tools/_perturbation_space/_discriminator_classifiers.py

@@ -226,7 +226,7 @@ class MLPClassifierSpace(PerturbationSpace):
         # Fix class unbalance (likely to happen in perturbation datasets)
         # Usually control cells are overrepresented such that predicting control all time would give good results
         # Cells with rare perturbations are sampled more
-        train_weights = 1 / (1 + torch.sum(torch.tensor(train_dataset.labels), dim=1))
+        train_weights = 1 / (1 + torch.sum(torch.tensor(train_dataset.labels.to_list()), dim=1))
         train_sampler = WeightedRandomSampler(train_weights, len(train_weights))
 
         self.train_dataloader = DataLoader(train_dataset, batch_size=batch_size, sampler=train_sampler, num_workers=4)

pertpy/tools/_perturbation_space/_simple.py

@@ -1,21 +1,15 @@
 from __future__ import annotations
 
-from typing import TYPE_CHECKING
+from typing import Literal
 
-import matplotlib.pyplot as plt
 import numpy as np
+import scanpy as sc
 from anndata import AnnData
-from decoupler import get_pseudobulk as dc_get_pseudobulk
-from decoupler import plot_psbulk_samples as dc_plot_psbulk_samples
 from sklearn.cluster import DBSCAN, KMeans
 
-from pertpy._doc import _doc_params, doc_common_plot_args
 from pertpy.tools._perturbation_space._clustering import ClusteringSpace
 from pertpy.tools._perturbation_space._perturbation_space import PerturbationSpace
 
-if TYPE_CHECKING:
-    from matplotlib.pyplot import Figure
-
 
 class CentroidSpace(PerturbationSpace):
     """Computes the centroids per perturbation of a pre-computed embedding."""
@@ -126,9 +120,9 @@ class PseudobulkSpace(PerturbationSpace):
         groups_col: str = None,
         layer_key: str = None,
         embedding_key: str = None,
-        **kwargs,
+        mode: Literal["count_nonzero", "mean", "sum", "var", "median"] = "sum",
     ) -> AnnData:  # type: ignore
-        """Determines pseudobulks of an AnnData object. It uses Decoupler implementation.
+        """Determines pseudobulks of an AnnData object.
 
         Args:
             adata: Anndata object of size cells x genes
@@ -137,7 +131,7 @@ class PseudobulkSpace(PerturbationSpace):
                 The summarized expression per perturbation (target_col) and group (groups_col) is computed.
             layer_key: If specified pseudobulk computation is done by using the specified layer. Otherwise, computation is done with .X
             embedding_key: `obsm` key of the AnnData embedding to use for computation. Defaults to the 'X' matrix otherwise.
-            **kwargs: Are passed to decoupler's get_pseuobulk.
+            mode: Pseudobulk aggregation function
 
         Returns:
              AnnData object with one observation per perturbation.
@@ -167,53 +161,16 @@ class PseudobulkSpace(PerturbationSpace):
                 adata = adata_emb
 
         adata.obs[target_col] = adata.obs[target_col].astype("category")
-        ps_adata = dc_get_pseudobulk(adata, sample_col=target_col, layer=layer_key, groups_col=groups_col, **kwargs)  # type: ignore
+        ps_adata = sc.get.aggregate(
+            adata, by=[target_col] if groups_col is None else [target_col, groups_col], func=mode, layer=layer_key
+        )
+        if mode in ps_adata.layers:
+            ps_adata.X = ps_adata.layers[mode]
 
         ps_adata.obs[target_col] = ps_adata.obs[target_col].astype("category")
 
         return ps_adata
 
-    @_doc_params(common_plot_args=doc_common_plot_args)
-    def plot_psbulk_samples(  # pragma: no cover # noqa: D417
-        self,
-        adata: AnnData,
-        groupby: str,
-        *,
-        return_fig: bool = False,
-        **kwargs,
-    ) -> Figure | None:
-        """Plot the pseudobulk samples of an AnnData object.
-
-        Plot the count number vs. the number of cells per pseudobulk sample.
-
-        Args:
-            adata: Anndata containing pseudobulk samples.
-            groupby: `.obs` column to color the samples by.
-            {common_plot_args}
-            **kwargs: Are passed to decoupler's plot_psbulk_samples.
-
-        Returns:
-            If `return_fig` is `True`, returns the figure, otherwise `None`.
-
-        Examples:
-            >>> import pertpy as pt
-            >>> adata = pt.dt.zhang_2021()
-            >>> ps = pt.tl.PseudobulkSpace()
-            >>> pdata = ps.compute(
-            ...     adata, target_col="Patient", groups_col="Cluster", mode="sum", min_cells=10, min_counts=1000
-            ... )
-            >>> ps.plot_psbulk_samples(pdata, groupby=["Patient", "Major celltype"], figsize=(12, 4))
-
-        Preview:
-            .. image:: /_static/docstring_previews/pseudobulk_samples.png
-        """
-        fig = dc_plot_psbulk_samples(adata, groupby, return_fig=True, **kwargs)
-
-        if return_fig:
-            return fig
-        plt.show()
-        return None
-
 
 class KMeansSpace(ClusteringSpace):
     """Computes K-Means clustering of the expression values."""

pertpy/tools/_scgen/_scgen.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 from typing import TYPE_CHECKING, Any
 
+import anndata as ad
 import jax.numpy as jnp
 import matplotlib.pyplot as plt
 import numpy as np
@@ -248,7 +249,7 @@ class Scgen(JaxTrainingMixin, BaseModelClass):
                 temp_cell[batch_ind[study]].X = batch_list[study].X
             shared_ct.append(temp_cell)
 
-        all_shared_ann = AnnData.concatenate(*shared_ct, batch_key="concat_batch", index_unique=None)
+        all_shared_ann = ad.concat(shared_ct, label="concat_batch", index_unique=None)
         if "concat_batch" in all_shared_ann.obs.columns:
             del all_shared_ann.obs["concat_batch"]
         if len(not_shared_ct) < 1:

{pertpy-0.11.3.dist-info → pertpy-0.11.5.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pertpy
-Version: 0.11.3
+Version: 0.11.5
 Summary: Perturbation Analysis in the scverse ecosystem.
 Project-URL: Documentation, https://pertpy.readthedocs.io
 Project-URL: Source, https://github.com/scverse/pertpy
@@ -49,7 +49,6 @@ Requires-Python: <3.14,>=3.11
 Requires-Dist: adjusttext
 Requires-Dist: arviz
 Requires-Dist: blitzgsea
-Requires-Dist: decoupler
 Requires-Dist: fast-array-utils
 Requires-Dist: lamin-utils
 Requires-Dist: mudata
@@ -132,6 +131,12 @@ You can install _pertpy_ in less than a minute via [pip] from [PyPI]:
 pip install pertpy
 ```
 
+or [conda-forge]:
+
+```console
+conda install -c conda-forge pertpy
+```
+
 ### Differential gene expression
 
 If you want to use the differential gene expression interface, please install pertpy by running:
@@ -180,6 +185,7 @@ pip install rpy2
 [pip]: https://pip.pypa.io/
 [pypi]: https://pypi.org/
 [api]: https://pertpy.readthedocs.io/en/latest/api.html
+[conda-forge]: https://anaconda.org/conda-forge/pertpy
 [//]: # "numfocus-fiscal-sponsor-attribution"
 
 pertpy is part of the scverse® project ([website](https://scverse.org), [governance](https://scverse.org/about/roles)) and is fiscally sponsored by [NumFOCUS](https://numfocus.org/).

{pertpy-0.11.3.dist-info → pertpy-0.11.5.dist-info}/RECORD

@@ -1,4 +1,4 @@
-pertpy/__init__.py,sha256=nuIzUydNMtNnhu1E4ffxU9gCgIdLi3liUv7dylR-2_I,716
+pertpy/__init__.py,sha256=KIxMlqyHlppcGM5Uc2HpTwCEtGFavXRPW50dM5dFB7U,716
 pertpy/_doc.py,sha256=j5TMNC-DA9yIMqIIUNpjpcVgWfRqyBBfvbRjnCM_OLs,427
 pertpy/_types.py,sha256=IcHCojCUqx8CapibNkcYf2TUqjBFP2ujeELvn_IBSBQ,154
 pertpy/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -19,18 +19,18 @@ pertpy/preprocessing/_guide_rna_mixture.py,sha256=pT_YkjmN4iEJ-THBROu_dpbr8E6u8G
 pertpy/tools/__init__.py,sha256=xjfw3Dd_KGytjjCID0uEs6Fz7DalH46fCjVL2Zf2kOo,2629
 pertpy/tools/_augur.py,sha256=tc1YKyc0BwzrEGgctsfyy7DsTNKxyvy7ZvWraTWCc1A,55262
 pertpy/tools/_cinemaot.py,sha256=54-rS0AEj31dMe7iU4kEmLoAunq3jNuhsBE3IEp9hrI,38071
-pertpy/tools/_dialogue.py,sha256=cCSwo9ge1pOLoA7QHTPb3b865juCFWUaKX5aD7UoSjo,52355
+pertpy/tools/_dialogue.py,sha256=mygIZm5i_bnEE37TTQtr1efl_KJq-ejzeL3V1Bmr7Pg,52354
 pertpy/tools/_enrichment.py,sha256=55mwotLH9DXQOhl85MCkxXu-MX0RysLyrPheJysAnF0,21369
-pertpy/tools/_milo.py,sha256=r-kZcpAcoQuhi41AnVuzh-cMIcV3HB3-RGzynHyDc1A,43712
-pertpy/tools/_mixscape.py,sha256=qjXGyH-oeBFte0efuHJfhVEbivnzUVWREwC40ef6Se8,57203
+pertpy/tools/_milo.py,sha256=zIYG0aP8B39_eiNgpZONhTKmDvcRwCzOLo5FMOTMUms,45530
+pertpy/tools/_mixscape.py,sha256=HfrpBeRlxHXaOpZkF2FmX7dg35kUB1rL0_-n2aSi2_0,57905
 pertpy/tools/decoupler_LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
 pertpy/tools/transferlearning_MMD_LICENSE,sha256=MUvDA-o_j9htRpI8fStVdCRuyLdPkQUuIH0a_EIc57w,1069
 pertpy/tools/_coda/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-pertpy/tools/_coda/_base_coda.py,sha256=-rpdipPLFd4cFXyLMN7uFgv-pFJseKaqDmyWRBrGfws,111519
+pertpy/tools/_coda/_base_coda.py,sha256=NjKIQBtTIUENnRmeIC2O8cMdU_9DKaJ5_AHPvFnc8XQ,111744
 pertpy/tools/_coda/_sccoda.py,sha256=0Ret6O56kAfCNOdBvtxqiyuj2rUPp18SV1GVK1AvYGU,22607
 pertpy/tools/_coda/_tasccoda.py,sha256=BTaOAmL458zQ_og3x4ENlDnJHD6_F4YkdCoXWsF4i1U,30465
 pertpy/tools/_differential_gene_expression/__init__.py,sha256=SEydWg0iT3Y1pApjnCAOuHxFeI6xVUfgyBHv2s3LADU,487
-pertpy/tools/_differential_gene_expression/_base.py,sha256=PpfH_RZXsN79Bu0yBFPE9TXEqNsZ4bSzSbhM0wZou2I,38322
+pertpy/tools/_differential_gene_expression/_base.py,sha256=ELx0e9DChJGO3yRpCLUOykt3oNOyDAPOQZZGSwzBSR0,38265
 pertpy/tools/_differential_gene_expression/_checks.py,sha256=hH_GP0lWGO-5zrCFX4YiIVCZBCuK0ZJ0jFmdlx2Qm4k,1639
 pertpy/tools/_differential_gene_expression/_dge_comparison.py,sha256=LXhp5djKKCAk9VI7OqxOuja849G5lnd8Ehcs9Epk8rg,4159
 pertpy/tools/_differential_gene_expression/_edger.py,sha256=nSHMDA4drGq_sJwUXs5I2EbMHwqjiS08GqOU_1_dXPc,4798
@@ -39,20 +39,20 @@ pertpy/tools/_differential_gene_expression/_simple_tests.py,sha256=SfU8s_P2JzEA1
 pertpy/tools/_differential_gene_expression/_statsmodels.py,sha256=90h9EPuoCtNxAbJ1Xq4j_E4yYJJpk64zTP7GyTdmrxY,2220
 pertpy/tools/_distances/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pertpy/tools/_distances/_distance_tests.py,sha256=6_nqfHUfKxkI2Yhkzspq3ujMpq56zV_Ddn7bgPzgjyo,13513
-pertpy/tools/_distances/_distances.py,sha256=89d1zShW_9dhphup2oWx5hMOFC7RdogOY56doMuBFts,50473
+pertpy/tools/_distances/_distances.py,sha256=_XbVU8dlYt_Jl2thYPUWg7HT6OXVe-Ki6qthF566sqQ,50503
 pertpy/tools/_perturbation_space/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 pertpy/tools/_perturbation_space/_clustering.py,sha256=pNx_SpPkZfCbgF7vzHWqAaiiHdbxPaA-L-hTWTbzFhI,3528
 pertpy/tools/_perturbation_space/_comparison.py,sha256=-NzCPRT-IlhJ9hOz7NQLSk0riIzr2C0yZvX6zm3kon4,4291
-pertpy/tools/_perturbation_space/_discriminator_classifiers.py,sha256=gDid9Z1_AAPHPWuNgAkbP7yrgcC0qjjqTuWjTzTAAZo,23373
+pertpy/tools/_perturbation_space/_discriminator_classifiers.py,sha256=a53-YmUwDHQBCT7ZWe_RH7PZsGXvoSHmJaQyL0CBJng,23383
 pertpy/tools/_perturbation_space/_metrics.py,sha256=y8-baP8WRdB1iDgvP3uuQxSCDxA2lcxvEHHM2C_vWHY,3248
 pertpy/tools/_perturbation_space/_perturbation_space.py,sha256=8RxVUkVEPZj5YZ-C-NP5zO4aYYVD04PzlsYuaIG-wjY,19447
-pertpy/tools/_perturbation_space/_simple.py,sha256=nnagHJ_aPv4ZCqfnVLdVUT_JShtIXg7iEP_sCMD3JLY,14271
+pertpy/tools/_perturbation_space/_simple.py,sha256=AJlHRaEP-vViBeMDvvMtUnXMuIKqZVc7wggnjsHMfMw,12721
 pertpy/tools/_scgen/__init__.py,sha256=uERFlFyF88TH0uLiwmsUGEfHfLVCiZMFuk8gO5f7164,45
 pertpy/tools/_scgen/_base_components.py,sha256=Qq8myRUm43q9XBrZ9gBggfa2cSV2wbz_KYoLgH7iF1A,3009
-pertpy/tools/_scgen/_scgen.py,sha256=31T8ez0FxABIbunJHCk8xvGulHFb8RHXSsyM_z1WsPY,30850
+pertpy/tools/_scgen/_scgen.py,sha256=AQNGsDe-9HEqli3oq7UBDg68ofLCoXm-R_jnLFQ-rlc,30856
 pertpy/tools/_scgen/_scgenvae.py,sha256=bPk4v7EdJc7ROdLuDitHiX_Pvwa7Flw2qHRUwBvjLJY,3889
 pertpy/tools/_scgen/_utils.py,sha256=qz5QUn_Bvk2NGyYVzp3jgjWTFOMt1YyHwUo6HWtoThY,2871
-pertpy-0.11.3.dist-info/METADATA,sha256=PM1yN_AADeouMbFT9X2m4Qv4VfPJ2PFgAbC2FNScIXs,8726
-pertpy-0.11.3.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-pertpy-0.11.3.dist-info/licenses/LICENSE,sha256=XuiT2hxeRInhquEIBKMZ5M21n5syhDQ4XbABoposIAg,1100
-pertpy-0.11.3.dist-info/RECORD,,
+pertpy-0.11.5.dist-info/METADATA,sha256=YEYgYTHkjmyWyboRL3RhBaSxOw86O5vr0wpXdvaLTGk,8827
+pertpy-0.11.5.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+pertpy-0.11.5.dist-info/licenses/LICENSE,sha256=XuiT2hxeRInhquEIBKMZ5M21n5syhDQ4XbABoposIAg,1100
+pertpy-0.11.5.dist-info/RECORD,,