future proofing scFates

requirements.txt

@@ -7,7 +7,7 @@ scanpy>=1.7.0
 leidenalg>=0.8.1
 numpy<2.0
 joblib>=0.17.0
-pandas<2.0
+pandas<=2.2.3
 statsmodels>=0.11.1
 matplotlib<3.9
 plotly>=4.8.1
@@ -18,7 +18,7 @@ typing_extensions
 scikit-learn
 scikit-misc
 simpleppt>=1.1.3
-elpigraph-python==0.3.1
+elpigraph-python==0.3.2
 networkx<3.0
 igraph>=0.9.6
 adjustText>=0.7.3

scFates/plot/features.py

@@ -182,7 +182,7 @@ def trends(
     fitted = pd.DataFrame(
         adata[:, features].layers["fitted"], index=adata.obs_names, columns=features
     ).T.copy(deep=True)
-    g = adata.obs.groupby("seg")
+    g = adata.obs.groupby("seg",observed=False )
 
     dct = graph["milestones"]
     keys = np.array(list(dct.keys()))
@@ -476,7 +476,7 @@ def add_frames(axis, vert):
             gs_emb = gridspec.GridSpecFromSubplotSpec(1, 1, subplot_spec=gsubs[0])
             axemb = fig.add_subplot(gs_emb[0])
             axs = axs + [axemb]
-        adata_temp.obs["mean_trajectory"] = 0
+        adata_temp.obs["mean_trajectory"] = 0.0
         adata_temp.obs.loc[
             fitted_sorted.columns, "mean_trajectory"
         ] = fitted_sorted.mean(axis=0).values

scFates/plot/matrix.py

@@ -238,7 +238,7 @@ def matrix(
 
         if return_data:
             data = pd.DataFrame(
-                0, index=features, columns=adata_sub.obs.split.cat.categories
+                0.0, index=features, columns=adata_sub.obs.split.cat.categories
             )
             allidx = data.stack().index
             for group in data.columns:
@@ -338,7 +338,7 @@ def matrix(
             borderpad=2,
         )
 
-        mappable = cm.ScalarMappable(cmap=cm.get_cmap(cmap))
+        mappable = cm.ScalarMappable(cmap=plt.get_cmap(cmap))
         cbar = fig.colorbar(mappable, cax=position, orientation="horizontal", aspect=50)
         cbar.set_ticks([0, 1])
         cbar.set_ticklabels(

scFates/plot/slide_cors.py

@@ -198,7 +198,7 @@ def slide_cors(
             emb[np.argsort(freq), 0],
             emb[np.argsort(freq), 1],
             s=10,
-            c=freq[np.argsort(freq)],
+            c=freq.iloc[np.argsort(freq)],
             cmap=gr,
             rasterized=True,
         )
@@ -327,11 +327,11 @@ def slide_cors(
                 pairwise_distances(np.array([0, 0]).reshape(1, -1), fB).argsort()
             )[0][:top_focus]
             texts_A = [
-                ax_scat.text(fA.loc[t][0], fA.loc[t][1], t, **kwargs_text)
+                ax_scat.text(fA.loc[t].iloc[0], fA.loc[t].iloc[1], t, **kwargs_text)
                 for t in fA.index[topA]
             ]
             texts_B = [
-                ax_scat.text(fB.loc[t][0], fB.loc[t][1], t, **kwargs_text)
+                ax_scat.text(fB.loc[t].iloc[0], fB.loc[t].iloc[1], t, **kwargs_text)
                 for t in fB.index[topB]
             ]
             texts = texts_A + texts_B

scFates/plot/trajectory.py

@@ -748,7 +748,7 @@ def _get_color_values(
     # when plotting, the color of the dots is determined for each plot
     # the data is either categorical or continuous and the data could be in
     # 'obs' or in 'var'
-    if not is_categorical_dtype(values):
+    if not isinstance(values, pd.CategoricalDtype):
         return values, False
     else:  # is_categorical_dtype(values)
         color_key = f"{value_to_plot}_colors"

scFates/plot/utils.py

@@ -129,13 +129,11 @@ def make_projection_available(projection):
 
 
 def is_categorical(data, c=None):
-    from pandas.api.types import is_categorical_dtype as cat
-
     if c is None:
-        return cat(data)  # if data is categorical/array
+        return isinstance(data, pd.CategoricalDtype)  # if data is categorical/array
     if not is_view(data):  # if data is anndata view
         strings_to_categoricals(data)
-    return isinstance(c, str) and c in data.obs.keys() and cat(data.obs[c])
+    return isinstance(c, str) and c in data.obs.keys() and isinstance(data.obs[c], pd.CategoricalDtype)
 
 
 def is_view(adata):

scFates/tests/test_w_plots.py

@@ -2,6 +2,7 @@
 import scanpy as sc
 import numpy as np
 import pandas as pd
+import matplotlib.pyplot as plt
 
 
 def test_pipeline():
@@ -80,6 +81,8 @@ def test_pipeline():
     scf.pl.trajectory(adata_2, color_seg="milestones", arrows=True)
     scf.pl.trajectory(adata_2, color_seg="seg", arrows=True)
 
+    plt.close()
+
     df = scf.tl.getpath(adata, root_milestone="43", milestones=["18"])
 
     adata.obsm["X_umap3d"] = np.concatenate(
@@ -89,6 +92,8 @@ def test_pipeline():
     scf.pl.trajectory_3d(adata)
     scf.pl.trajectory_3d(adata, color="seg")
 
+    plt.close()
+
     adata_s1 = scf.tl.subset_tree(
         adata, root_milestone="88", milestones=["18"], mode="substract", copy=True
     )
@@ -100,7 +105,7 @@ def test_pipeline():
 
     adata_lim = scf.tl.subset_tree(adata, t_max=adata.obs.t.max() / 4 * 3, copy=True)
 
-    scf.tl.test_association(adata, n_jobs=2)
+    scf.tl.test_association(adata, n_jobs=1)
     scf.tl.test_association(adata, A_cut=0.3, reapply_filters=True)
     scf.tl.test_association(adata_2, layer="scaled", A_cut=0.3)
     nsigni = adata.var.signi.sum()
@@ -148,6 +153,7 @@ def test_pipeline():
         plot_emb=False,
         ordering="quantile",
     )
+    plt.close()
 
     scf.pl.trends(
         adata,
@@ -158,6 +164,7 @@ def test_pipeline():
         ordering="pearson",
         show=False,
     )
+    plt.close()
 
     scf.pl.single_trend(adata, feature=adata.var_names[0], color_exp="k")
 
@@ -173,15 +180,15 @@ def test_pipeline():
     )
 
     scf.tl.test_fork(
-        adata, layer="scaled", root_milestone="43", milestones=["24", "18"], n_jobs=2
+        adata, layer="scaled", root_milestone="43", milestones=["24", "18"], n_jobs=1
     )
-    scf.tl.test_fork(adata, root_milestone="43", milestones=["24", "18"], n_jobs=2)
+    scf.tl.test_fork(adata, root_milestone="43", milestones=["24", "18"], n_jobs=1)
     signi_fdr_nonscaled = adata.uns["43->24<>18"]["fork"].signi_fdr.sum()
     adata.uns["43->24<>18"]["fork"]["fdr"] = 0.02
     scf.pl.test_fork(adata, root_milestone="43", milestones=["24", "18"])
 
     scf.tl.test_fork(
-        adata, root_milestone="43", milestones=["24", "18"], n_jobs=2, rescale=True
+        adata, root_milestone="43", milestones=["24", "18"], n_jobs=1, rescale=True
     )
     signi_fdr_rescaled = adata.uns["43->24<>18"]["fork"].signi_fdr.sum()
 

scFates/tools/bifurcation_tools.py

@@ -42,7 +42,7 @@ def test_fork(
     copy: bool = False,
 ):
 
-    """\
+    """
     Test for branch differential gene expression and differential upregulation from progenitor to terminal state.
 
     First, differential gene expression between two branches is performed. The following model is used:
@@ -154,7 +154,7 @@ def get_branches(m):
             return ddf
 
         brcells = pd.concat([get_branches(m) for m in milestones])
-        brcells = brcells.loc[brcells.index.drop_duplicates(False)]
+        brcells = brcells.loc[brcells.index.drop_duplicates(keep=False)]
 
         matw = None
         if matw is None:
@@ -324,7 +324,7 @@ def gt_fun(data):
 
     tmin = np.min([sdf.loc[sdf.i == i, "t"].max() for i in sdf.i.unique()])
     sdf = sdf.loc[sdf.t <= tmin]
-    Amps = sdf.groupby("i").apply(lambda x: np.mean(x.exp))
+    Amps = sdf.groupby("i")['exp'].mean()
     res = Amps - Amps.max()
     res["de_p"] = rmgcv.summary_gam(m)[3][1]
 
@@ -351,7 +351,7 @@ def branch_specific(
     copy: bool = False,
 ):
 
-    """\
+    """
     Assign genes differentially expressed between two post-bifurcation branches.
 
     Parameters
@@ -439,7 +439,7 @@ def activation(
     layer=None,
 ):
 
-    """\
+    """
     Identify pseudotime of activation of branch-specififc features.
 
     This aims in classifying the genes according to their their activation timing
@@ -522,7 +522,7 @@ def activation_map(m):
             df = adata.obs.loc[:, ["t", "seg"]]
         else:
             df = adata.uns["pseudotime_list"][str(m)]
-        acti = pd.Series(0, index=stats.index)
+        acti = pd.Series(0.0, index=stats.index)
 
         for leave in leaves:
             subtree = getpath(img, root, graph["tips"], leave, graph, df).sort_values(
@@ -617,9 +617,9 @@ def get_activation(data):
     subtree = data[0]
     subtree["exp"] = data[1]
     # load parameters
-    deriv_cut = subtree["deriv_cut"][0]
-    nwin = subtree["nwin"][0]
-    steps = subtree["steps"][0]
+    deriv_cut = subtree["deriv_cut"].iloc[0]
+    nwin = subtree["nwin"].iloc[0]
+    steps = subtree["steps"].iloc[0]
 
     wf = warnings.filters.copy()
     warnings.filterwarnings("ignore")
@@ -676,7 +676,7 @@ def activation_lm(
     layer=None,
 ):
 
-    """\
+    """
     A more robust version of `tl.activation`.
 
     This is considered to be a more robust version of :func:`scFates.tl.activation`.

scFates/tools/dendrogram.py

@@ -96,7 +96,7 @@ def dendrogram(adata: anndata.AnnData, crowdedness: float = 1, n_jobs: int = 1):
     adata.obsm["X_dendro"] = dend.loc[adata.obs_names].values
 
     # generate segments
-    newseg = newseg[newseg.argsort()]
+    newseg = newseg.iloc[newseg.argsort()]
 
     segments = []
     for i, n in enumerate(newseg.index):
@@ -264,7 +264,7 @@ def _group_longform(self, vals, grouper, order):
         """Group a long-form variable by another with correct order."""
 
         # Group the val data
-        grouped_vals = vals.groupby(grouper)
+        grouped_vals = vals.groupby(grouper,observed=False)
         out_data = []
         for g in order:
             try:

scFates/tools/fit.py

@@ -198,7 +198,7 @@ def fit(
 def gt_fun(data):
     sdf = data[0]
     sdf["exp"] = data[1]
-    gamma = sdf["gamma"][0]
+    gamma = sdf["gamma"].iloc[0]
 
     global rmgcv
     global rstats

scFates/tools/graph_operations.py

@@ -1,6 +1,7 @@
 from typing import Optional, Union, Iterable
 from typing_extensions import Literal
 from anndata import AnnData
+import anndata as ad
 import numpy as np
 import pandas as pd
 from pandas import DataFrame
@@ -433,9 +434,7 @@ def attach_tree(
 
     newcells = np.concatenate([adata.obs_names, adata_branch.obs_names])
 
-    adata = adata.concatenate(
-        adata_branch, batch_key=None, index_unique=None, uns_merge="first", join="outer"
-    )
+    adata = ad.concat([adata,adata_branch],uns_merge="first", join="outer")
 
     logg.info("    tree refitting")
     use_rep = graph["use_rep"]

scFates/tools/linearity_deviation.py

@@ -93,7 +93,7 @@ def lindev_map(m):
 
         t_perc = [cells.t.quantile(p / 100) for p in percentiles]
 
-        df[name + "_lindev_sel"] = np.nan
+        df[name + "_lindev_sel"] = 'none'
         df.loc[
             adata.obs_names.isin(cells.index), name + "_lindev_sel"
         ] = "putative bridge"

scFates/tools/pseudotime.py

@@ -290,7 +290,7 @@ def map_on_edges(v):
 
         if vcells.shape[0] > 0:
             nv = np.array(g.neighborhood(v, order=1))[1:]
-            nvd = g.shortest_paths(v, nv, weights=g.es["weight"])
+            nvd = g.distances(v, nv, weights=g.es["weight"])
 
             ndf = pd.DataFrame(
                 {
@@ -314,7 +314,7 @@ def map_on_edges(v):
                 for i in range(ndf.shape[0])
             ]
 
-            ndf["seg"] = 0
+            ndf["seg"] = '0'
             isinfork = (graph["pp_info"].loc[ndf.v0, "PP"].isin(graph["forks"])).values
             ndf.loc[isinfork, "seg"] = (
                 graph["pp_info"].loc[ndf.loc[isinfork, "v1"], "seg"].values
@@ -332,7 +332,7 @@ def map_on_edges(v):
     df.sort_values("cell", inplace=True)
     # df.index = graph["cells_fitted"]
 
-    df["edge"] = df.apply(lambda x: str(int(x[1])) + "|" + str(int(x[2])), axis=1)
+    df["edge"] = df.apply(lambda x: str(int(x.iloc[1])) + "|" + str(int(x.iloc[2])), axis=1)
 
     df.drop(["cell", "v0", "v1", "d"], axis=1, inplace=True)
 
@@ -417,11 +417,11 @@ def unroll_circle(adata: AnnData, copy: bool = False):
 
     adata = adata.copy() if copy else adata
     pp_seg = adata.uns["graph"]["pp_seg"]
-    adata.obs.t[adata.obs.seg == "0"] = (
+    adata.obs.loc[adata.obs.seg == "0",'t'] = (
         -adata.obs.t[adata.obs.seg == "0"] + adata.obs.t[adata.obs.seg == "0"].max()
     )
 
-    adata.obs.t[adata.obs.seg == "0"] = (
+    adata.obs.loc[adata.obs.seg == "0",'t'] = (
         adata.obs.t[adata.obs.seg == "0"] + adata.obs.t[adata.obs.seg == "0"].max()
     )
 
@@ -452,7 +452,7 @@ def unroll_circle(adata: AnnData, copy: bool = False):
             "n": 0,
             "from": a,
             "to": b,
-            "d": g.shortest_paths(a, b, weights="weight")[0][0],
+            "d": g.distances(a, b, weights="weight")[0][0],
         },
         index=[0],
     )

scFates/tools/root.py

@@ -124,9 +124,9 @@ def root(
 
         sub_g = g.vs.select(group=0).subgraph()
         a, b = np.argwhere(np.array(sub_g.degree()) == 1).ravel()
-        dst_0 = sub_g.shortest_paths(a, b, weights="weight")[0][0]
+        dst_0 = sub_g.distances(a, b, weights="weight")[0][0]
         a, b = sub_g.vs["label"][a], sub_g.vs["label"][b]
-        dst_1 = g.shortest_paths(root, furthest, weights="weight")[0][0]
+        dst_1 = g.distances(root, furthest, weights="weight")[0][0]
         pp_seg = pd.concat(
             [
                 pd.Series([0, a, b, dst_0], index=["n", "from", "to", "d"]),
@@ -290,7 +290,7 @@ def roots(adata: AnnData, roots, meeting, copy: bool = False):
     nodes = np.argwhere(
         np.apply_along_axis(arr=(csr > 0).todense(), axis=0, func1d=np.sum) != 2
     ).flatten()
-    pp_seg = pd.DataFrame(columns=["n", "from", "to", "d"])
+    pp_seg = []
     for node1, node2 in itertools.combinations(nodes, 2):
         paths12 = g.get_shortest_paths(node1, node2)
         paths12 = np.array([val for sublist in paths12 for val in sublist])
@@ -300,19 +300,19 @@ def roots(adata: AnnData, roots, meeting, copy: bool = False):
             path_root = root_dist_matrix[[node1, node2]]
             fro = fromto[np.argmin(path_root)]
             to = fromto[np.argmax(path_root)]
-            pp_info.loc[paths12, "seg"] = pp_seg.shape[0] + 1
-            pp_seg = pp_seg.append(
+            pp_info.loc[paths12, "seg"] = len(pp_seg) + 1
+            pp_seg.append(
                 pd.DataFrame(
                     {
-                        "n": pp_seg.shape[0] + 1,
+                        "n": len(pp_seg) + 1,
                         "from": fro,
                         "to": to,
                         "d": shortest_path(csr, directed=False, indices=fro)[to],
                     },
-                    index=[pp_seg.shape[0] + 1],
+                    index=[len(pp_seg) + 1],
                 )
             )
-
+    pp_seg = pd.concat(pp_seg, axis=0)
     pp_seg["n"] = pp_seg["n"].astype(int).astype(str)
     pp_seg["n"] = pp_seg["n"].astype(int).astype(str)