decoupler 2.0.1__tar.gz → 2.0.2__tar.gz

{decoupler-2.0.1 → decoupler-2.0.2}/.github/workflows/test.yaml

@@ -78,6 +78,7 @@ jobs:
           MPLBACKEND: agg
           PLATFORM: ${{ matrix.os }}
           DISPLAY: :42
+          NUMBA_JIT_COVERAGE: "1"
         run: uvx hatch run ${{ matrix.env.name }}:run-cov
       - name: generate coverage report
         run: uvx hatch run ${{ matrix.env.name }}:coverage xml

{decoupler-2.0.1 → decoupler-2.0.2}/LICENSE

@@ -1,6 +1,6 @@
 BSD 3-Clause License
 
-Copyright (c) 2025, Pau Badia i Mompel
+Copyright (c) 2025, Pau Badia i Mompel, Saez lab
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without

{decoupler-2.0.1 → decoupler-2.0.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: decoupler
-Version: 2.0.1
+Version: 2.0.2
 Summary: Python package to perform enrichment analysis from omics data.
 Project-URL: Documentation, https://decoupler.readthedocs.io/
 Project-URL: Homepage, https://github.com/PauBadiaM/decoupler
@@ -9,7 +9,7 @@ Author: Pau Badia i Mompel
 Maintainer-email: Pau Badia i Mompel <pau.badia@uni-heidelberg.de>
 License: BSD 3-Clause License
         
-        Copyright (c) 2025, Pau Badia i Mompel
+        Copyright (c) 2025, Pau Badia i Mompel, Saez lab
         All rights reserved.
         
         Redistribution and use in source and binary forms, with or without
@@ -47,6 +47,7 @@ Requires-Dist: adjusttext
 Requires-Dist: anndata
 Requires-Dist: docrep
 Requires-Dist: igraph
+Requires-Dist: ipywidgets
 Requires-Dist: marsilea
 Requires-Dist: numba
 Requires-Dist: requests
@@ -134,7 +135,7 @@ There are several alternative options to install decoupler:
 pip install decoupler
 ```
 
-2. Install the latest stable version from [conda-forge][conda] using mamba or conda:
+2. Install the latest stable version from [conda-forge][conda] using mamba or conda (pay attention to the `-py` suffix at the end):
 
 ```bash
 mamba create -n=dcp conda-forge::decoupler-py

{decoupler-2.0.1 → decoupler-2.0.2}/README.md

@@ -53,7 +53,7 @@ There are several alternative options to install decoupler:
 pip install decoupler
 ```
 
-2. Install the latest stable version from [conda-forge][conda] using mamba or conda:
+2. Install the latest stable version from [conda-forge][conda] using mamba or conda (pay attention to the `-py` suffix at the end):
 
 ```bash
 mamba create -n=dcp conda-forge::decoupler-py

{decoupler-2.0.1 → decoupler-2.0.2}/pyproject.toml

@@ -4,7 +4,7 @@ requires = [ "hatchling" ]
 
 [project]
 name = "decoupler"
-version = "2.0.1"
+version = "2.0.2"
 description = "Python package to perform enrichment analysis from omics data."
 readme = "README.md"
 license = { file = "LICENSE" }
@@ -27,6 +27,7 @@ dependencies = [
   "numba",
   "docrep",
   "tqdm",
+  "ipywidgets",
   "xgboost",
   "seaborn",
   "requests",

{decoupler-2.0.1 → decoupler-2.0.2}/src/decoupler/bm/metric/_auc.py

@@ -63,16 +63,13 @@ def auprc(
     # Compute prc
     ps = tps + fps
     msk = ps != 0
-    if pi0 > 0 and pi0 < 1:
-        # Siblini W., Fréry J., He-Guelton L., Oblé F., Wang YQ. (2020) Master
-        # Your Metrics with Calibration. In: Berthold M., Feelders A., Krempl G.
-        # (eds) Advances in Intelligent Data Analysis XVIII. IDA 2020. Lecture
-        # Notes in Computer Science, vol 12080. Springer, Cham
-        pi = np.sum(y_true) / y_true.size
-        ratio = pi * (1 - pi0) / (pi0 * (1 - pi))
-        prc = tps[msk] / (tps[msk] + ratio * fps[msk])
-    else:
-        prc = np.divide(tps[msk], ps[msk])
+    # Siblini W., Fréry J., He-Guelton L., Oblé F., Wang YQ. (2020) Master
+    # Your Metrics with Calibration. In: Berthold M., Feelders A., Krempl G.
+    # (eds) Advances in Intelligent Data Analysis XVIII. IDA 2020. Lecture
+    # Notes in Computer Science, vol 12080. Springer, Cham
+    pi = np.sum(y_true) / y_true.size
+    ratio = pi * (1 - pi0) / (pi0 * (1 - pi))
+    prc = tps[msk] / (tps[msk] + ratio * fps[msk])
     # Compute rcl
     rcl = tps / tps[-1]
     # Flip and add limits
@@ -94,6 +91,8 @@ def auc(
     """
     # Normalize to make comparable
     norm = np.nanmax(np.abs(y_score), axis=1)
+    msk = norm == 0.
+    norm[msk] = 1.
     y_score = y_score / norm.reshape(-1, 1)
     assert ((-1. <= y_score) & (y_score <= 1.)).all()
     # Flatten and remove nans

{decoupler-2.0.1 → decoupler-2.0.2}/src/decoupler/bm/metric/_hmean.py

@@ -40,6 +40,12 @@ def hmean(
     -------
     Dataframe containing the harmonic mean per metric.
     """
+    # Validate
+    assert isinstance(df, pd.DataFrame), 'df must be pandas.DataFrame'
+    assert isinstance(metrics, (str, list)), 'metrics must be str or list'
+    if isinstance(metrics, str):
+        metrics = [metrics]
+    # Run
     d_metrics = {
         'auc': {
             'name': 'H(auroc, auprc)',
@@ -63,7 +69,6 @@ def hmean(
         # Compute harmonic mean
         name = d_metrics[metric]['name']
         tmp[name] = _hmean(tmp[cols[0]], tmp[cols[1]], beta=beta)
-        #tmp = tmp.drop(columns=cols)
         if i == 0:
             hdf.append(tmp)
         else:

{decoupler-2.0.1 → decoupler-2.0.2}/src/decoupler/mt/_mdt.py

@@ -1,9 +1,12 @@
+import warnings
 from typing import Tuple
 
 import numpy as np
 import scipy.sparse as sps
 from tqdm.auto import tqdm
-from xgboost import XGBRegressor
+with warnings.catch_warnings():
+    warnings.filterwarnings("ignore", category=FutureWarning, module="xgboost")
+    from xgboost import XGBRegressor
 
 from decoupler._docs import docs
 from decoupler._log import _log

{decoupler-2.0.1 → decoupler-2.0.2}/src/decoupler/mt/_udt.py

@@ -1,9 +1,12 @@
+import warnings
 from typing import Tuple
 
 import numpy as np
 import scipy.sparse as sps
 from tqdm.auto import tqdm
-from xgboost import XGBRegressor
+with warnings.catch_warnings():
+    warnings.filterwarnings("ignore", category=FutureWarning, module="xgboost")
+    from xgboost import XGBRegressor
 
 from decoupler._docs import docs
 from decoupler._log import _log

{decoupler-2.0.1 → decoupler-2.0.2}/src/decoupler/pl/_order.py

@@ -53,7 +53,7 @@ def order(
             bp.ax.imshow(
                 colors,
                 aspect='auto',
-                extent=[0, xmax, 1.05 * ymax, 1.2 * ymax],
+                extent=[xmin, xmax, 1.05 * ymax, 1.2 * ymax],
                 transform=bp.ax.transData,
                 zorder=2
             )
@@ -75,9 +75,9 @@ def order(
             .groupby(['name', 'order'], as_index=False)['value'].mean()
             .pivot(index='name', columns='order', values='value')
         )
-        img = bp.ax.imshow(mat, extent=[0, xmax, 0, n_names], aspect='auto', **kw_order)
+        img = bp.ax.imshow(mat, extent=[xmin, xmax, 0, n_names], aspect='auto', **kw_order)
         if has_cbar:
-            bp.ax.imshow(colors, aspect='auto', extent=[0, xmax, n_names, 1.1 * n_names], zorder=2)
+            bp.ax.imshow(colors, aspect='auto', extent=[xmin, xmax, n_names, 1.1 * n_names], zorder=2)
             bp.ax.axhline(y=n_names, c='black', lw=1)
             bp.ax.set_ylim(0, 1.1 * n_names)
         bp.fig.colorbar(img, ax=bp.ax, shrink=0.5, label='Mean value', location='top')
@@ -85,5 +85,5 @@ def order(
         bp.ax.set_yticklabels(np.flip(mat.index))
         bp.ax.grid(axis='y', visible=False)
         bp.ax.set_xlabel('order')
-    bp.ax.set_xlim(0, 1)
+    bp.ax.set_xlim(xmin, xmax)
     return bp._return()

{decoupler-2.0.1 → decoupler-2.0.2}/src/decoupler/pl/_order_targets.py

@@ -140,9 +140,11 @@ def order_targets(
     ax.grid(axis='y', visible=False)
     ax.set_ylabel('Targets')
     yticklabels = []
+    # Find minmax
+    omin, omax = df_ftr['order'].min(), df_ftr['order'].max()
     # Add neg targets
     if neg_names.size > 0:
-        img = ax.imshow(mat.loc[neg_names], extent=[0, 1, 0, neg_names.size], aspect='auto', cmap=neg_cmap, vmin=vmin, vmax=vmax)
+        img = ax.imshow(mat.loc[neg_names], extent=[omin, omax, 0, neg_names.size], aspect='auto', cmap=neg_cmap, vmin=vmin, vmax=vmax)
         yticklabels.extend(list(neg_names)[::-1])
         cbar_mappable = ScalarMappable(cmap=neg_cmap, norm=Normalize(vmin=vmin, vmax=vmax))
         pos = ax.get_position().bounds
@@ -152,22 +154,22 @@ def order_targets(
     ax.axhline(y=neg_names.size, c='black', lw=1)
     # Add pos targets
     if pos_names.size > 0:
-        img = ax.imshow(mat.loc[pos_names], extent=[0, 1, neg_names.size, neg_names.size + pos_names.size], aspect='auto', cmap='Reds', vmin=vmin, vmax=vmax)
+        img = ax.imshow(mat.loc[pos_names], extent=[omin, omax, neg_names.size, neg_names.size + pos_names.size], aspect='auto', cmap='Reds', vmin=vmin, vmax=vmax)
         yticklabels.extend(list(pos_names)[::-1])
         cbar_mappable = ScalarMappable(cmap=pos_cmap, norm=Normalize(vmin=vmin, vmax=vmax))
         pos = ax.get_position().bounds
-        #  (left, bottom, width, height)
         cax = bp.fig.add_axes([0.97, pos[1] + (pos[3] / 2) + .025, 0.05, (pos[3] / 2) - .025])
         cax.grid(axis='y', visible=False)
         bp.fig.colorbar(cbar_mappable, cax=cax, aspect=5, shrink=0.5, label='+ target\nvalues', location='right')
     # Plot labels
     ax.set_ylim(0, n_names)
     if has_cbar:
-        ax.imshow(colors, aspect='auto', extent=[0, 1, n_names, 1.1 * n_names], zorder=2)
+        ax.imshow(colors, aspect='auto', extent=[omin, omax, n_names, 1.1 * n_names], zorder=2)
         ax.axhline(y=n_names, c='black', lw=1)
         ax.set_ylim(0, 1.1 * n_names)
     # Format plot
     ax.set_yticks(np.arange(n_names) + 0.5)
     ax.set_yticklabels(yticklabels)
+    ax.set_xlim(omin, omax)
     bp.fig.subplots_adjust(hspace=0)
     return bp._return()

{decoupler-2.0.1 → decoupler-2.0.2}/src/decoupler/pp/anndata.py

@@ -151,6 +151,10 @@ def _sample_group_by(
     # Use one column if the same
     if sample == group:
         group = None
+    # Handle list columns
+    ocols = obs.select_dtypes(include='object').columns
+    for ocol in ocols:
+        obs[ocol] = obs[ocol].str.join('_')
     if group is None:
         # Filter extra columns in obs
         cols = obs.groupby(sample, observed=True).nunique(dropna=False).eq(other=1).all(axis=0)
@@ -703,10 +707,9 @@ def bin_order(
     if sps.issparse(X):
         X = X.toarray()
     y = adata.obs[order].values
-    # Normalize to 0 and 1
-    y = np.abs(y) / np.abs(y).max()
     # Make windows
-    bin_edges = np.linspace(0, 1, nbins + 1)
+    ymin, ymax = y.min(), y.max()
+    bin_edges = np.linspace(start=ymin, stop=ymax, num=nbins + 1)
     bin_midpoints = (bin_edges[:-1] + bin_edges[1:]) / 2
     # Prepare label colors
     cols = ['name', 'midpoint', 'value']
@@ -735,5 +738,4 @@ def bin_order(
     df = pd.concat(dfs)
     df = df[cols]
     df = df.rename(columns={'midpoint': 'order'}).reset_index(drop=True)
-    df['order'] = df['order'] / df['order'].max()
     return df

{decoupler-2.0.1 → decoupler-2.0.2}/src/decoupler/tl/_rankby_order.py

@@ -1,8 +1,12 @@
+import warnings
+
 import pandas as pd
 import scipy.stats as sts
 import scipy.sparse as sps
+with warnings.catch_warnings():
+    warnings.filterwarnings("ignore", category=FutureWarning, module="xgboost")
+    from xgboost import XGBRegressor
 from anndata import AnnData
-from xgboost import XGBRegressor
 
 from decoupler._docs import docs
 from decoupler.pp.data import extract

decoupler-2.0.2/tests/bm/test_benchmark.py

@@ -0,0 +1,59 @@
+import pandas as pd
+import scipy.sparse as sps
+import pytest
+
+import decoupler as dc
+
+
+@pytest.mark.parametrize(
+    'metrics,groupby,runby,sfilt,thr,emin,mnet',
+    [
+        ['auc', None, 'expr', False, 0.05, 5, False],
+        ['auc', None, 'expr', True, 0.05, 5, False],
+        [['auc'], None, 'expr', False, 0.05, 5, False],
+        [['auc', 'fscore'], 'group', 'expr', False, 0.05, 5, False],
+        [['auc', 'fscore', 'qrank'], None, 'source', False, 0.05, 2, False],
+        [['auc', 'fscore', 'qrank'], 'group', 'source', False, 0.05, 1, False],
+        [['auc', 'fscore', 'qrank'], 'bm_group', 'expr', True, 0.05, 5, False],
+        [['auc', 'fscore', 'qrank'], 'source', 'expr', True, 0.05, 5, False],
+    ]
+)
+def test_benchmark(
+    bdata,
+    net,
+    metrics,
+    groupby,
+    runby,
+    sfilt,
+    thr,
+    emin,
+    mnet,
+    rng,
+):
+    dc.mt.ulm(data=bdata, net=net, tmin=0)
+    if mnet:
+        net = {'w_net': net, 'unw_net': net.drop(columns=['weight'])}
+        bdata = bdata.copy()
+        bdata.obs['source'] = rng.choice(['x', 'y', 'z'], size=bdata.n_obs, replace=True)
+        bdata.X = sps.csr_matrix(bdata.X)
+    df = dc.bm.benchmark(
+        adata=bdata,
+        net=net,
+        metrics=metrics,
+        groupby=groupby,
+        runby=runby,
+        sfilt=sfilt,
+        thr=thr,
+        emin=emin,
+        kws_decouple={
+            'cons': True,
+            'tmin': 3,
+            'methods': ['ulm', 'zscore', 'aucell']
+        },
+        verbose=True
+    )
+    assert isinstance(df, pd.DataFrame)
+    cols = {'method', 'metric', 'score'}
+    assert cols.issubset(df.columns)
+    hdf = dc.bm.metric.hmean(df, metrics=metrics)
+    assert isinstance(hdf, pd.DataFrame)

decoupler-2.0.2/tests/bm/test_pl.py

@@ -0,0 +1,89 @@
+import pandas as pd
+import matplotlib
+matplotlib.use("Agg")
+import matplotlib.pyplot as plt
+from matplotlib.figure import Figure
+import pytest
+
+import decoupler as dc
+
+@pytest.fixture
+def df():
+    df = pd.DataFrame(
+        data=[
+            ['aucell', 'auroc', 0.45],
+            ['aucell', 'auprc', 0.55],
+            ['ulm', 'auroc', 0.9],
+            ['ulm', 'auprc', 0.8],
+            ['aucell', 'recall', 0.45],
+            ['aucell', 'precision', 0.55],
+            ['ulm', 'recall', 0.9],
+            ['ulm', 'precision', 0.8],
+            ['aucell', '1-qrank', 0.45],
+            ['aucell', '-log10(pval)', 0.9],
+            ['ulm', '1-qrank', 0.9],
+            ['ulm', '-log10(pval)', 5.6],
+        ],
+        columns = ['method', 'metric', 'score']
+    )
+    return df
+
+
+@pytest.fixture
+def hdf(
+    df,
+):
+    hdf = dc.bm.metric.hmean(df)
+    return hdf
+
+
+def test_auc(
+    df,
+):
+    fig = dc.bm.pl.auc(df=df, hue=None, return_fig=True)
+    assert isinstance(fig, Figure)
+    plt.close(fig)
+    fig = dc.bm.pl.auc(df=df, hue='method', return_fig=True)
+    assert isinstance(fig, Figure)
+    plt.close(fig)
+
+
+def test_fscore(
+    df,
+):
+    fig = dc.bm.pl.fscore(df=df, hue=None, return_fig=True)
+    assert isinstance(fig, Figure)
+    plt.close(fig)
+    fig = dc.bm.pl.fscore(df=df, hue='method', return_fig=True)
+    assert isinstance(fig, Figure)
+    plt.close(fig)
+
+
+def test_qrank(
+    df,
+):
+    fig = dc.bm.pl.qrank(df=df, hue=None, return_fig=True)
+    assert isinstance(fig, Figure)
+    plt.close(fig)
+    fig = dc.bm.pl.qrank(df=df, hue='method', return_fig=True)
+    assert isinstance(fig, Figure)
+    plt.close(fig)
+
+
+def test_bar(
+    hdf,
+):
+    fig = dc.bm.pl.bar(df=hdf, x='H(auroc, auprc)', y='method', hue=None, return_fig=True)
+    assert isinstance(fig, Figure)
+    plt.close(fig)
+    fig = dc.bm.pl.bar(df=hdf, x='H(auroc, auprc)', y='method', hue='method', return_fig=True)
+    assert isinstance(fig, Figure)
+    plt.close(fig)
+
+
+def test_summary(
+    hdf,
+):
+    fig = dc.bm.pl.summary(df=hdf, y='method', figsize=(6, 3), return_fig=True)
+    assert isinstance(fig, Figure)
+    plt.close(fig)

{decoupler-2.0.1 → decoupler-2.0.2}/tests/conftest.py

@@ -49,6 +49,7 @@ def pdata(
 @pytest.fixture
 def bdata():
     adata, _ = dc.ds.toy_bench(nobs=100, nvar=20, bval=2, seed=42, verbose=False)
+    adata.obs['bm_group'] = adata.obs.apply(lambda x: [x['sample'], x['group']], axis=1)
     return adata
 
 

{decoupler-2.0.1 → decoupler-2.0.2}/tests/pp/test_anndata.py

@@ -250,7 +250,6 @@ def test_bin_order(
     cols = {'name', 'order', 'value'}
     assert cols.issubset(df.columns)
     assert ((0. <= df['order']) & (df['order'] <= 1.)).all()
-    assert df['order'].unique().size == np.min([tdata.n_obs, nbins])
     if label is not None:
         lcols = {'label', 'color'}
         assert lcols.issubset(df.columns)

decoupler-2.0.1/tests/bm/test_benchmark.py

@@ -1,46 +0,0 @@
-import pandas as pd
-import pytest
-
-import decoupler as dc
-
-
-@pytest.mark.parametrize(
-    'metrics,groupby,runby,sfilt,thr,emin',
-    [
-        ['auc', None, 'expr', False, 0.05, 5],
-        ['auc', None, 'expr', True, 0.05, 5],
-        [['auc'], None, 'expr', False, 0.05, 5],
-        [['auc', 'fscore'], 'group', 'expr', False, 0.05, 5],
-        [['auc', 'fscore', 'qrank'], None, 'source', False, 0.05, 2],
-        [['auc', 'fscore', 'qrank'], 'group', 'source', False, 0.05, 1],
-    ]
-)
-def test_benchmark(
-    bdata,
-    net,
-    metrics,
-    groupby,
-    runby,
-    sfilt,
-    thr,
-    emin,
-):
-    df = dc.bm.benchmark(
-        adata=bdata,
-        net=net,
-        metrics=metrics,
-        groupby=groupby,
-        runby=runby,
-        sfilt=sfilt,
-        thr=thr,
-        emin=emin,
-        kws_decouple={
-            'cons': True,
-            'tmin': 3,
-            'methods': ['ulm', 'zscore']
-        },
-    )
-    assert isinstance(df, pd.DataFrame)
-    cols = {'method', 'metric', 'score'}
-    print(df.columns)
-    assert cols.issubset(df.columns)