Merge pull request #157 from saezlab/circle

add Circle plot, address issues with AnnData, and scipy sparse

.bumpversion.cfg

@@ -1,5 +1,5 @@
 [bumpversion]
-current_version = 1.4.0
+current_version = 1.5.0
 commit = True
 tag = True
 files = pyproject.toml liana/__init__.py

.github/workflows/main.yml

@@ -12,7 +12,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.8", "3.10"]
+        python-version: ["3.9", "3.10"]
 
     steps:
     - uses: actions/checkout@v3
@@ -43,14 +43,14 @@ jobs:
 
   build-macOS:
 
-    runs-on: macOS-12
+    runs-on: macOS-13
 
     steps:
     - uses: actions/checkout@v2
-    - name: Set up Python 3.8
+    - name: Set up Python 3.9
       uses: actions/setup-python@v2
       with:
-        python-version: 3.8
+        python-version: 3.9
     - name: Install Poetry
       run: |
         python -m pip install --upgrade pip

docs/source/release_notes.rst

@@ -1,6 +1,13 @@
 Changelog
 =============
 
+1.5.0 (17.01.2025)
+- New ``circle_plot`` is now available (Merged #139). Thanks to @WeipengMO.
+
+- Update bivariate metrics to no longer save in place but rather return the AnnData
+
+- Issue related to .A for a csr_matrix after a certain scipy version #155, #135
+
 1.4.0 (02.09.2024)
 -------------------------------------------------
 

liana/__init__.py

@@ -1,4 +1,4 @@
-__version__ = '1.4.0'
+__version__ = '1.5.0'
 
 from liana import method as mt, plotting as pl, resource as rs, multi as mu, utils as ut, testing
 

liana/method/sc/_liana_pipe.py

@@ -460,8 +460,8 @@ def _assign_min_or_max(x, x_ascending):
 
 
 def _trimean(a, axis=0):
-    quantiles = np.quantile(a.A, q=[0.25, 0.75], axis=axis)
-    median = np.median(a.A, axis=axis)
+    quantiles = np.quantile(a.toarray(), q=[0.25, 0.75], axis=axis)
+    median = np.median(a.toarray(), axis=axis)
     return (quantiles[0] + 2 * median + quantiles[1]) / 4
 
 def _cluster_stats(adata):
@@ -471,7 +471,7 @@ def _cluster_stats(adata):
         temp = adata[adata.obs['@label'].isin([label])]
 
         cluster_stats.loc[label, 'mean'] = temp.X.mean()
-        cluster_stats.loc[label, 'std'] = np.std(temp.X.A)
+        cluster_stats.loc[label, 'std'] = np.std(temp.X.toarray())
 
     return cluster_stats
 

liana/method/sp/_bivariate/_local_functions.py

@@ -36,12 +36,13 @@ def __call__(self,
             weight = _spatialdm_weight_norm(weight)
         else:
             if issparse(x_mat):
-               x_mat = x_mat.A
+               x_mat = x_mat.toarray()
             if issparse(y_mat):
-                y_mat = y_mat.A
+                y_mat = y_mat.toarray()
 
         if self.name.__contains__("masked") or weight.shape[0] < 10000:
-            weight = weight.todense().A
+            if issparse(weight):
+                weight = weight.todense().A
 
         local_scores = self.fun(x_mat, y_mat, weight)
 
@@ -180,7 +181,7 @@ def _get_local_var(self, x_sigma, y_sigma, weight, spot_n):
         return std
 
     def _norm_max(self, X, axis=0):
-        X = X / X.max(axis=axis).A
+        X = X / X.max(axis=axis).toarray()
         X = _zscore(X, axis=axis)
         X = np.where(np.isnan(X), 0, X)
 

liana/method/sp/_bivariate/_spatial_bivariate.py

@@ -44,8 +44,6 @@ def __call__(self,
                  remove_self_interactions: bool = True,
                  complex_sep: (None | str) = "_",
                  xy_sep: str = V.lr_sep,
-                 inplace: bool = V.inplace,
-                 key_added: str = "local_scores",
                  verbose: bool = V.verbose,
                  **kwargs
                  ) -> Union[AnnData, pd.DataFrame] | None:
@@ -76,9 +74,6 @@ def __call__(self,
             Separator to use for interaction names.
         remove_self_interactions: bool
             Whether to remove self-interactions. `True` by default.
-        %(inplace)s
-        key_added: str
-            Key in `mdata.mod` (if MuData) or `adata.obsm` (if AnnData) where the local scores will be stored.
         %(verbose)s
 
         **kwargs : dict, optional
@@ -107,14 +102,8 @@ def __call__(self,
 
         Returns
         -------
-        If `inplace` is `True`, the results are added to `mdata` and `None` is returned.
-        Note that `var`, `obs`, `obsm`, `uns` and `obsp` attributes are copied to the output object.
-        When `mdata` is an `AnnData` object, the results are stored in `mdata.obsm[key_added]`
-        When `mdata` is an `MuData` object, the results are stored in `mdata.mod[key_added]`
-
-        `AnnData` objects in `obsm` will not be copied to the output object.
-
-        If `inplace` is `False`, the results are returned.
+        An AnnData object, (optionally) with multiple layers which correspond categories/p-values, and the
+        actual scores are stored in `.X`. Moreover, global stats are stored in ``.var``.
 
         """
 
@@ -251,7 +240,7 @@ def __call__(self,
         if local_pvals is not None:
             local_scores.layers['pvals'] = csr_matrix(local_pvals)
 
-        return self._handle_return(mdata, xy_stats, local_scores, key_added, inplace)
+        return local_scores
 
     def _process_anndata(self,
                          adata,
@@ -331,15 +320,6 @@ def _rename_means(self, lr_stats, entity):
         df.columns = df.columns.map(lambda x: entity + '_' + str(x) if x != 'gene' else 'gene')
         return df.rename(columns={'gene': entity})
 
-    def _handle_return(self, data, stats, local_scores, key_added, inplace=False):
-        if not inplace:
-            return stats, local_scores
-
-        if isinstance(data, MuData):
-            data.mod[key_added] = local_scores
-        else:
-            data.obsm[key_added] = local_scores
-
     def _handle_connectivity(self, adata, connectivity_key):
         if connectivity_key not in adata.obsp.keys():
             raise ValueError(f'No connectivity matrix founds in mdata.obsp[{connectivity_key}]')
@@ -358,8 +338,8 @@ def _encode_cats(self, a, weight):
         return a
 
     def _categorize(self, x_mat, y_mat, weight):
-        x_cats = self._encode_cats(x_mat.A, weight)
-        y_cats = self._encode_cats(y_mat.A, weight)
+        x_cats = self._encode_cats(x_mat.toarray(), weight)
+        y_cats = self._encode_cats(y_mat.toarray(), weight)
         cats = x_cats + y_cats
         cats = np.where(cats == 2, 1, np.where(cats == 0, -1, 0))
 

liana/plotting/__init__.py

@@ -2,3 +2,4 @@
 from ._dotplot import dotplot, dotplot_by_sample
 from ._misty_plots import target_metrics, contributions, interactions
 from ._tileplot import tileplot
+from ._circle_plot import circle_plot