Update concatenation function during batch removal

.github/workflows/test.yml

@@ -15,7 +15,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: [3.8, 3.9, 3.10]
+        python-version: ['3.8', '3.9', '3.10']
 
     steps:
     - uses: actions/checkout@v2

docs/tutorials/scgen_perturbation_prediction.ipynb

@@ -14,7 +14,8 @@
    "outputs": [],
    "source": [
     "import sys\n",
-    "#if branch is stable, will install via pypi, else will install from source\n",
+    "\n",
+    "# if branch is stable, will install via pypi, else will install from source\n",
     "branch = \"stable\"\n",
     "IN_COLAB = \"google.colab\" in sys.modules\n",
     "\n",
@@ -78,8 +79,10 @@
     }
    ],
    "source": [
-    "train = sc.read(\"./tests/data/train_kang.h5ad\",\n",
-    "                backup_url='https://drive.google.com/uc?id=1r87vhoLLq6PXAYdmyyd89zG90eJOFYLk')"
+    "train = sc.read(\n",
+    "    \"./tests/data/train_kang.h5ad\",\n",
+    "    backup_url=\"https://drive.google.com/uc?id=1r87vhoLLq6PXAYdmyyd89zG90eJOFYLk\",\n",
+    ")"
    ]
   },
   {
@@ -95,8 +98,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "train_new = train[~((train.obs[\"cell_type\"] == \"CD4T\") &\n",
-    "                    (train.obs[\"condition\"] == \"stimulated\"))]"
+    "train_new = train[\n",
+    "    ~((train.obs[\"cell_type\"] == \"CD4T\") & (train.obs[\"condition\"] == \"stimulated\"))\n",
+    "]"
    ]
   },
   {
@@ -198,10 +202,7 @@
    ],
    "source": [
     "model.train(\n",
-    "    max_epochs=100,\n",
-    "    batch_size=32,\n",
-    "    early_stopping=True,\n",
-    "    early_stopping_patience=25\n",
+    "    max_epochs=100, batch_size=32, early_stopping=True, early_stopping_patience=25\n",
     ")"
    ]
   },
@@ -258,8 +259,13 @@
    "source": [
     "sc.pp.neighbors(latent_adata)\n",
     "sc.tl.umap(latent_adata)\n",
-    "sc.pl.umap(latent_adata, color=['condition', 'cell_type'], wspace=0.4, frameon=False,\n",
-    "           save='latentspace_batch32_klw000005_z100__100e.pdf')"
+    "sc.pl.umap(\n",
+    "    latent_adata,\n",
+    "    color=[\"condition\", \"cell_type\"],\n",
+    "    wspace=0.4,\n",
+    "    frameon=False,\n",
+    "    save=\"latentspace_batch32_klw000005_z100__100e.pdf\",\n",
+    ")"
    ]
   },
   {
@@ -339,11 +345,9 @@
    ],
    "source": [
     "pred, delta = model.predict(\n",
-    "    ctrl_key='control',\n",
-    "    stim_key='stimulated',\n",
-    "    celltype_to_predict='CD4T'\n",
+    "    ctrl_key=\"control\", stim_key=\"stimulated\", celltype_to_predict=\"CD4T\"\n",
     ")\n",
-    "pred.obs['condition'] = 'pred'"
+    "pred.obs[\"condition\"] = \"pred\""
    ]
   },
   {
@@ -380,8 +384,12 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "ctrl_adata = train[((train.obs['cell_type'] == 'CD4T') & (train.obs['condition'] == 'control'))]\n",
-    "stim_adata = train[((train.obs['cell_type'] == 'CD4T') & (train.obs['condition'] == 'stimulated'))]"
+    "ctrl_adata = train[\n",
+    "    ((train.obs[\"cell_type\"] == \"CD4T\") & (train.obs[\"condition\"] == \"control\"))\n",
+    "]\n",
+    "stim_adata = train[\n",
+    "    ((train.obs[\"cell_type\"] == \"CD4T\") & (train.obs[\"condition\"] == \"stimulated\"))\n",
+    "]"
    ]
   },
   {
@@ -433,8 +441,12 @@
    ],
    "source": [
     "sc.tl.pca(eval_adata)\n",
-    "sc.pl.pca(eval_adata, color=\"condition\", frameon=False,\n",
-    "           save='pred_stim_b32_klw000005_z100__100e.pdf')"
+    "sc.pl.pca(\n",
+    "    eval_adata,\n",
+    "    color=\"condition\",\n",
+    "    frameon=False,\n",
+    "    save=\"pred_stim_b32_klw000005_z100__100e.pdf\",\n",
+    ")"
    ]
   },
   {
@@ -466,7 +478,7 @@
     }
    ],
    "source": [
-    "CD4T = train[train.obs[\"cell_type\"] ==\"CD4T\"]"
+    "CD4T = train[train.obs[\"cell_type\"] == \"CD4T\"]"
    ]
   },
   {
@@ -527,7 +539,7 @@
     "    labels={\"x\": \"predicted\", \"y\": \"ground truth\"},\n",
     "    path_to_save=\"./reg_mean1.pdf\",\n",
     "    show=True,\n",
-    "    legend=False\n",
+    "    legend=False,\n",
     ")"
    ]
   },
@@ -567,11 +579,11 @@
     "    eval_adata,\n",
     "    axis_keys={\"x\": \"pred\", \"y\": \"stimulated\"},\n",
     "    gene_list=diff_genes[:10],\n",
-    "    top_100_genes= diff_genes,\n",
-    "    labels={\"x\": \"predicted\",\"y\": \"ground truth\"},\n",
+    "    top_100_genes=diff_genes,\n",
+    "    labels={\"x\": \"predicted\", \"y\": \"ground truth\"},\n",
     "    path_to_save=\"./reg_mean1.pdf\",\n",
     "    show=True,\n",
-    "    legend=False\n",
+    "    legend=False,\n",
     ")"
    ]
   },

pyproject.toml

@@ -53,7 +53,7 @@ python-igraph = {version = "*", optional = true}
 scanpy = ">=1.6"
 scanpydoc = {version = ">=0.5", optional = true}
 scikit-misc = {version = ">=0.1.3", optional = true}
-scvi-tools = ">=0.20.0"
+scvi-tools = "<0.20.0"
 seaborn = ">=0.11"
 sphinx = {version = ">=4.1,<4.4", optional = true}
 sphinx-autodoc-typehints = {version = "*", optional = true}

readthedocs.yml

@@ -4,7 +4,7 @@ build:
 sphinx:
   configuration: docs/conf.py
 python:
-  version: 3.10
+  version: 3.8
   install:
   - method: pip
     path: .

scgen/_scgen.py

@@ -5,7 +5,7 @@
 import scanpy as sc
 import torch
 from adjustText import adjust_text
-from anndata import AnnData
+from anndata import AnnData, concat
 from matplotlib import pyplot
 from scipy import stats
 from scvi import REGISTRY_KEYS
@@ -265,9 +265,7 @@ def batch_removal(self, adata: Optional[AnnData] = None) -> AnnData:
                 batch_list[study].X = delta + batch_list[study].X
                 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 = 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:

scgen/_scgenvae.py

@@ -116,7 +116,6 @@ def loss(
         inference_outputs,
         generative_outputs,
     ):
-
         x = tensors[REGISTRY_KEYS.X_KEY]
         qz_m = inference_outputs["qz_m"]
         qz_v = inference_outputs["qz_v"]
@@ -156,7 +155,10 @@ def sample(
             tensor with shape (n_cells, n_genes, n_samples)
         """
         inference_kwargs = dict(n_samples=n_samples)
-        inference_outputs, generative_outputs, = self.forward(
+        (
+            inference_outputs,
+            generative_outputs,
+        ) = self.forward(
             tensors,
             inference_kwargs=inference_kwargs,
             compute_loss=False,

scgen/_utils.py

@@ -60,7 +60,6 @@ def balancer(
     adata,
     cell_type_key,
 ):
-
     """
     Makes cell type population equal.
 

tests/test_scgen.py

@@ -1,10 +1,10 @@
-import scvi
 import scanpy as sc
+import scvi
+
 from scgen import SCGEN
 
 
 def test_scgen():
-
     adata = scvi.data.synthetic_iid()
     SCGEN.setup_anndata(
         adata,