Removed .A for .toarray() due to scipy update

PauBadiaM · PauBadiaM · commit 0fda9127c990 · 2024-07-05T13:33:28.000+02:00
diff --git a/decoupler/method_aucell.py b/decoupler/method_aucell.py
@@ -67,7 +67,7 @@ def aucell(mat, net, n_up, verbose):
     for i in tqdm(range(mat.shape[0]), disable=not verbose):
 
         if isinstance(mat, csr_matrix):
-            row = mat[i].A[0]
+            row = mat[i].toarray()[0]
         else:
             row = mat[i]
 
diff --git a/decoupler/method_gsea.py b/decoupler/method_gsea.py
@@ -195,7 +195,7 @@ def gsea(mat, net, times=1000, seed=42, verbose=False):
     for i in tqdm(range(n_samples), disable=not verbose):
 
         if isinstance(mat, csr_matrix):
-            row = mat[i].A[0]
+            row = mat[i].toarray()[0]
         else:
             row = mat[i]
 
diff --git a/decoupler/method_gsva.py b/decoupler/method_gsva.py
@@ -224,7 +224,7 @@ def run_gsva(mat, net, source='source', target='target', kcdf=False, mx_diff=Tru
 
     # Run GSVA
     if isinstance(m, csr_matrix):
-        m = m.A
+        m = m.toarray()
     estimate = gsva(m, net, kcdf=kcdf, verbose=verbose)
 
     # Transform to df
diff --git a/decoupler/method_mdt.py b/decoupler/method_mdt.py
@@ -42,7 +42,7 @@ def mdt(mat, net, trees=100, min_leaf=5, n_jobs=4, seed=42, verbose=False):
     # For each sample
     for i in tqdm(range(mat.shape[0]), disable=not verbose):
         if isinstance(mat, csr_matrix):
-            sample = mat[i].A[0]
+            sample = mat[i].toarray()[0]
         else:
             sample = mat[i]
         acts[i] = fit_rf(sr, net, sample, trees=trees, min_leaf=min_leaf, n_jobs=n_jobs, seed=seed)
diff --git a/decoupler/method_mlm.py b/decoupler/method_mlm.py
@@ -55,7 +55,7 @@ def mlm(mat, net, batch_size=10000, verbose=False):
 
             # Subset batch
             srt, end = i * batch_size, i * batch_size + batch_size
-            y = mat[srt:end].A.T
+            y = mat[srt:end].toarray().T
 
             # Compute MLM for batch
             es[srt:end] = fit_mlm(net, y, inv, df)[:, 1:]
diff --git a/decoupler/method_ora.py b/decoupler/method_ora.py
@@ -114,7 +114,7 @@ def ora(mat, net, n_up_msk, n_bt_msk, n_background=20000, verbose=False):
     for i in tqdm(range(n_samples), disable=not verbose):
 
         if isinstance(mat, csr_matrix):
-            row = mat[i].A[0]
+            row = mat[i].toarray()[0]
         else:
             row = mat[i]
 
diff --git a/decoupler/method_udt.py b/decoupler/method_udt.py
@@ -42,7 +42,7 @@ def udt(mat, net, min_leaf=5, seed=42, verbose=False):
     # For each sample and regulator fit dt
     for i in tqdm(range(mat.shape[0]), disable=not verbose):
         if isinstance(mat, csr_matrix):
-            sample = mat[i].A[0]
+            sample = mat[i].toarray()[0]
         else:
             sample = mat[i]
         for j in range(net.shape[1]):
diff --git a/decoupler/method_ulm.py b/decoupler/method_ulm.py
@@ -42,7 +42,7 @@ def ulm(mat, net, batch_size=10000, verbose=False):
 
             # Subset batch
             srt, end = i * batch_size, i * batch_size + batch_size
-            batch = mat[srt:end].A.T
+            batch = mat[srt:end].toarray().T
 
             # Compute R for batch
             r = mat_cor(net, batch)
diff --git a/decoupler/method_viper.py b/decoupler/method_viper.py
@@ -195,7 +195,7 @@ def viper(mat, net, pleiotropy=True, reg_sign=0.05, n_targets=10, penalty=20, ba
 
             # Subset batch
             srt, end = i*batch_size, i*batch_size+batch_size
-            tmp = mat[srt:end].A
+            tmp = mat[srt:end].toarray()
 
             # Compute VIPER for batch
             nes[srt:end] = aREA(tmp, net)
@@ -211,7 +211,7 @@ def viper(mat, net, pleiotropy=True, reg_sign=0.05, n_targets=10, penalty=20, ba
 
             # Extract per sample
             if isinstance(mat, csr_matrix):
-                ss_i = mat[i].A[0]
+                ss_i = mat[i].toarray()[0]
             else:
                 ss_i = mat[i]
             nes_i = nes[i]
diff --git a/decoupler/method_wmean.py b/decoupler/method_wmean.py
@@ -84,7 +84,7 @@ def wmean(mat, net, times, batch_size, seed, verbose):
 
             # Subset batch
             srt, end = i * batch_size, i * batch_size + batch_size
-            tmp = mat[srt:end].A
+            tmp = mat[srt:end].toarray()
 
             # Run WMEAN
             estimate[srt:end] = tmp.dot(net) / div
diff --git a/decoupler/method_wsum.py b/decoupler/method_wsum.py
@@ -82,7 +82,7 @@ def wsum(mat, net, times, batch_size, seed, verbose):
 
             # Subset batch
             srt, end = i * batch_size, i * batch_size + batch_size
-            tmp = mat[srt:end].A
+            tmp = mat[srt:end].toarray()
 
             # Run WSUM
             estimate[srt:end] = tmp.dot(net)
diff --git a/decoupler/plotting.py b/decoupler/plotting.py
@@ -444,7 +444,7 @@ def plot_violins(mat, thr=None, log=False, use_raw=False, figsize=(7, 5), dpi=10
     # Format
     x = np.repeat(r, m.shape[1])
     if isinstance(m, csr_matrix):
-        y = m.A.flatten()
+        y = m.toarray().flatten()
     else:
         y = m.flatten()
 
diff --git a/decoupler/tests/test_aucell.py b/decoupler/tests/test_aucell.py
@@ -18,7 +18,7 @@ def test_aucell():
     assert act[2, 0] < 0.7
     assert act[3, 0] < 0.7
     assert np.all((0. <= act) * (act <= 1.))
-    act = aucell(m.A, net, n_up, False)
+    act = aucell(m.toarray(), net, n_up, False)
     assert act[0, 0] > 0.7
     assert act[1, 0] > 0.7
     assert act[2, 0] < 0.7
diff --git a/decoupler/tests/test_gsea.py b/decoupler/tests/test_gsea.py
@@ -28,10 +28,10 @@ def test_gsea():
     es, nes, pval = gsea(m, net, times=10)
     assert nes[0][0] > 0.
     assert nes[0][1] < 0.
-    es, nes, pval = gsea(m.A, net, times=10)
+    es, nes, pval = gsea(m.toarray(), net, times=10)
     assert nes[0][0] > 0.
     assert nes[0][1] < 0.
-    es, nes, pval = gsea(m.A, net, times=0)
+    es, nes, pval = gsea(m.toarray(), net, times=0)
     assert (nes is None) & (pval is None)
     assert es[0][0] > 0.
     assert es[0][1] < 0.
diff --git a/decoupler/tests/test_gsva.py b/decoupler/tests/test_gsva.py
@@ -10,13 +10,13 @@ def test_init_cdfs():
 
 
 def test_density():
-    m = csr_matrix(np.array([[1, 0, 2], [1., 0, 3], [0, 0, 0]], dtype=np.float32)).A
+    m = csr_matrix(np.array([[1, 0, 2], [1., 0, 3], [0, 0, 0]], dtype=np.float32)).toarray()
     density(m, kcdf=True)
     density(m, kcdf=False)
 
 
 def test_gsva():
-    m = csr_matrix(np.array([[1, 0, 2], [1., 0, 3], [0, 0, 0]], dtype=np.float32)).A
+    m = csr_matrix(np.array([[1, 0, 2], [1., 0, 3], [0, 0, 0]], dtype=np.float32)).toarray()
     net = pd.Series([np.array([1, 3], dtype=np.int64), np.array([1, 3], dtype=np.int64)], index=['T1', 'T2'])
     gsva(m, net)
 
diff --git a/decoupler/tests/test_mdt.py b/decoupler/tests/test_mdt.py
@@ -37,7 +37,7 @@ def test_mdt():
     ])
     a, b = mdt(m, net, seed=42, trees=100, min_leaf=2)[0]
     assert a > b
-    a, b = mdt(m.A, net, seed=42, trees=100, min_leaf=2)[0]
+    a, b = mdt(m.toarray(), net, seed=42, trees=100, min_leaf=2)[0]
     assert a > b
 
 
diff --git a/decoupler/tests/test_mlm.py b/decoupler/tests/test_mlm.py
@@ -14,7 +14,7 @@ def test_mlm():
     assert act[2, 0] < 0
     assert act[3, 0] < 0
     assert np.all((0. <= pvl) * (pvl <= 1.))
-    act, pvl = mlm(m.A, net)
+    act, pvl = mlm(m.toarray(), net)
     assert act[0, 0] > 0
     assert act[1, 0] > 0
     assert act[2, 0] < 0
diff --git a/decoupler/tests/test_pre.py b/decoupler/tests/test_pre.py
@@ -19,7 +19,7 @@ def test_check_mat():
     assert nr.size == 2
     assert nc.size == 2
     assert type(nm) is csr_matrix
-    nm, nr, nc = check_mat(m.A, r, c, verbose=True)
+    nm, nr, nc = check_mat(m.toarray(), r, c, verbose=True)
     assert nm.shape[0] == 2
     assert nm.shape[1] == 2
     assert nr.size == 2
diff --git a/decoupler/tests/test_udt.py b/decoupler/tests/test_udt.py
@@ -38,7 +38,7 @@ def test_udt():
     ])
     a, b = udt(m, net, seed=42, min_leaf=1)[0]
     assert a > b
-    a, b = udt(m.A, net, seed=42, min_leaf=1)[0]
+    a, b = udt(m.toarray(), net, seed=42, min_leaf=1)[0]
     assert a > b
 
 
diff --git a/decoupler/tests/test_ulm.py b/decoupler/tests/test_ulm.py
@@ -65,7 +65,7 @@ def test_ulm():
     assert act[2, 0] < 0
     assert act[3, 0] < 0
     assert np.all((0. <= pvl) * (pvl <= 1.))
-    act, pvl = ulm(m.A, net)
+    act, pvl = ulm(m.toarray(), net)
     assert act[0, 0] > 0
     assert act[1, 0] > 0
     assert act[2, 0] < 0
diff --git a/decoupler/tests/test_utilsanndata.py b/decoupler/tests/test_utilsanndata.py
@@ -79,7 +79,7 @@ def test_check_X():
     X_neg = csr_matrix(np.array([[1, 0, -2], [1, 0, -3], [0, 0, 0]]))
     X_inf = csr_matrix(np.array([[1, 0, np.nan], [1, 0, 3], [0, 0, 0]]))
     check_X(X, mode='sum', skip_checks=False)
-    check_X(X.A, mode='sum', skip_checks=False)
+    check_X(X.toarray(), mode='sum', skip_checks=False)
     with pytest.raises(ValueError):
         check_X(X_inf, mode='sum', skip_checks=False)
     with pytest.raises(ValueError):
@@ -89,13 +89,13 @@ def test_check_X():
     with pytest.raises(ValueError):
         check_X(X_float, mode='sum', skip_checks=False)
     check_X(X_neg, mode='sum', skip_checks=True)
-    check_X(X_neg.A, mode='sum', skip_checks=True)
+    check_X(X_neg.toarray(), mode='sum', skip_checks=True)
     check_X(X_float, mode='sum', skip_checks=True)
-    check_X(X_float.A, mode='sum', skip_checks=True)
+    check_X(X_float.toarray(), mode='sum', skip_checks=True)
     check_X(X_neg, mode=sum, skip_checks=False)
-    check_X(X_neg.A, mode=sum, skip_checks=False)
+    check_X(X_neg.toarray(), mode=sum, skip_checks=False)
     check_X(X_neg, mode={'sum': sum}, skip_checks=False)
-    check_X(X_neg.A, mode={'sum': sum}, skip_checks=False)
+    check_X(X_neg.toarray(), mode={'sum': sum}, skip_checks=False)
 
 
 def test_format_psbulk_inputs():
@@ -195,15 +195,15 @@ def test_compute_psbulk():
                                                    new_obs, min_cells, min_counts, mode, dtype)
     assert np.all(np.sum(psbulk, axis=1) == counts)
     assert np.all(np.array(psbulk.shape) == np.array(props.shape))
-    psbulk, ncells, counts, props = compute_psbulk(n_rows, n_cols, X.A, sample_col, groups_col, smples, groups, obs,
+    psbulk, ncells, counts, props = compute_psbulk(n_rows, n_cols, X.toarray(), sample_col, groups_col, smples, groups, obs,
                                                    new_obs, min_cells, min_counts, mode, dtype)
     assert np.all(np.sum(psbulk, axis=1) == counts)
     assert np.all(np.array(psbulk.shape) == np.array(props.shape))
     psbulk, ncells, counts, props = compute_psbulk(n_rows, n_cols, X, sample_col, groups_col, smples, groups, obs,
                                                    new_obs, min_cells, 9, mode, dtype)
     assert np.sum(psbulk, axis=1)[2] == 0.
     assert np.all(psbulk[2] == props[2])
-    psbulk, ncells, counts, props = compute_psbulk(n_rows, n_cols, X.A, sample_col, groups_col, smples, groups, obs,
+    psbulk, ncells, counts, props = compute_psbulk(n_rows, n_cols, X.toarray(), sample_col, groups_col, smples, groups, obs,
                                                    new_obs, min_cells, 9, mode, dtype)
     assert np.sum(psbulk, axis=1)[2] == 0.
     assert np.all(psbulk[2] == props[2])
@@ -217,7 +217,7 @@ def test_compute_psbulk():
     assert np.all(np.sum(psbulk, axis=1) == counts)
     assert np.all(np.array(psbulk.shape) == np.array(props.shape))
     assert psbulk.shape[0] == 2
-    psbulk, ncells, counts, props = compute_psbulk(n_rows, n_cols, X.A, sample_col, None, smples, groups, obs,
+    psbulk, ncells, counts, props = compute_psbulk(n_rows, n_cols, X.toarray(), sample_col, None, smples, groups, obs,
                                                    new_obs, min_cells, min_counts, mode, dtype)
     assert np.all(np.sum(psbulk, axis=1) == counts)
     assert np.all(np.array(psbulk.shape) == np.array(props.shape))
@@ -226,7 +226,7 @@ def test_compute_psbulk():
                                                    new_obs, min_cells, 21, mode, dtype)
     assert np.sum(psbulk, axis=1)[0] == 0.
     assert np.all(psbulk[0] == props[0])
-    psbulk, ncells, counts, props = compute_psbulk(n_rows, n_cols, X.A, sample_col, None, smples, groups, obs,
+    psbulk, ncells, counts, props = compute_psbulk(n_rows, n_cols, X.toarray(), sample_col, None, smples, groups, obs,
                                                    new_obs, min_cells, 21, mode, dtype)
     assert np.sum(psbulk, axis=1)[0] == 0.
     assert np.all(psbulk[0] == props[0])
diff --git a/decoupler/tests/test_wmean.py b/decoupler/tests/test_wmean.py
@@ -14,7 +14,7 @@ def test_wmean():
     assert norm[2, 0] < 0
     assert norm[3, 0] < 0
     assert np.all((0. <= pvl) * (pvl <= 1.))
-    est, norm, corr, pvl = wmean(m.A, net, 1000, 10000, 42, True)
+    est, norm, corr, pvl = wmean(m.toarray(), net, 1000, 10000, 42, True)
     assert norm[0, 0] > 0
     assert norm[1, 0] > 0
     assert norm[2, 0] < 0
diff --git a/decoupler/tests/test_wsum.py b/decoupler/tests/test_wsum.py
@@ -14,7 +14,7 @@ def test_wsum():
     assert norm[2, 0] < 0
     assert norm[3, 0] < 0
     assert np.all((0. <= pvl) * (pvl <= 1.))
-    est, norm, corr, pvl = wsum(m.A, net, 1000, 10000, 42, True)
+    est, norm, corr, pvl = wsum(m.toarray(), net, 1000, 10000, 42, True)
     assert norm[0, 0] > 0
     assert norm[1, 0] > 0
     assert norm[2, 0] < 0
diff --git a/decoupler/utils.py b/decoupler/utils.py
@@ -405,7 +405,7 @@ def dense_run(func, mat, net, source='source', target='target', weight='weight',
 
         # Extract single sample
         if isinstance(m, csr_matrix):
-            sample = m[i].A[0]
+            sample = m[i].toarray()[0]
         else:
             sample = m[i]
         idx = sample != 0.
diff --git a/decoupler/utils_anndata.py b/decoupler/utils_anndata.py
@@ -226,7 +226,7 @@ def compute_psbulk(n_rows, n_cols, X, sample_col, groups_col, smples, groups, ob
             # Get cells from specific sample
             profile = X[obs[sample_col] == smp]
             if isinstance(X, csr_matrix):
-                profile = profile.A
+                profile = profile.toarray()
 
             # Skip if few cells or not enough counts
             ncell = profile.shape[0]
@@ -261,7 +261,7 @@ def compute_psbulk(n_rows, n_cols, X, sample_col, groups_col, smples, groups, ob
                 # Get cells from specific sample and group
                 profile = X[(obs[sample_col] == smp) & (obs[groups_col] == grp)]
                 if isinstance(X, csr_matrix):
-                    profile = profile.A
+                    profile = profile.toarray()
 
                 # Skip if few cells or not enough counts
                 ncell = profile.shape[0]
