update all notebooks with tvn

recursionpharma · Dec 21, 2023 · 807f85a · 807f85a
1 parent af3260b
commit 807f85a
Show file tree

Hide file tree

Showing 4 changed files with 64 additions and 250 deletions.
diff --git a/efaar_benchmarking/data_loading.py b/efaar_benchmarking/data_loading.py
@@ -12,7 +12,7 @@
 from efaar_benchmarking.constants import PERISCOPE_PLATE_COL
 
 
-def load_periscope(cell_type="HeLa", normalized=False) -> tuple[pd.DataFrame, pd.DataFrame]:
+def load_periscope(cell_type="HeLa", normalized=True) -> tuple[pd.DataFrame, pd.DataFrame]:
     """
     Load PERISCOPE (cpg0021) data for a specific cell type.
     Find more information about the dataset here: https://www.biorxiv.org/content/10.1101/2023.08.06.552164v1
@@ -21,7 +21,7 @@ def load_periscope(cell_type="HeLa", normalized=False) -> tuple[pd.DataFrame, pd
 
     Parameters:
     cell_type (str, optional): The cell type to load data for. Defaults to "HeLa".
-    normalized (bool, optional): Whether to load normalized data. Defaults to False.
+    normalized (bool, optional): Whether to load normalized data. Defaults to True.
 
     Returns:
     tuple[pd.DataFrame, pd.DataFrame]: A tuple containing two DataFrames named `features` and `metadata`,

diff --git a/notebooks/jump_map_building.ipynb b/notebooks/jump_map_building.ipynb
diff --git a/notebooks/periscope_map_building.ipynb b/notebooks/periscope_map_building.ipynb
diff --git a/notebooks/replogle_map_building.ipynb b/notebooks/replogle_map_building.ipynb
@@ -35,20 +35,29 @@
    },
    "outputs": [],
    "source": [
-    "adata = load_replogle(\"genome_wide\", \"normalized\") # use for PCA Embeddings\n",
-    "embeddings = embed_by_pca_anndata(adata) # PCA Embeddings\n",
-    "# adata = load_replogle(\"genome_wide\", \"raw\") # use for scVI Embeddings\n",
-    "# embeddings = embed_by_scvi_anndata(adata) # scVI Embeddings\n",
+    "# adata = load_replogle(\"genome_wide\", \"normalized\") # use for PCA Embeddings\n",
+    "# embeddings = embed_by_pca_anndata(adata) # PCA Embeddings\n",
+    "adata = load_replogle(\"genome_wide\", \"raw\") # use for scVI Embeddings\n",
+    "embeddings = embed_by_scvi_anndata(adata) # scVI Embeddings\n",
     "metadata = adata.obs\n",
     "del adata\n",
-    "embeddings = align_on_controls(embeddings, metadata, pert_col=REPLOGLE_PERT_LABEL_COL, control_key=REPLOGLE_CONTROL_PERT_LABEL, scale=True)\n",
-    "phenoprint_pvals = univariate_consistency_benchmark(embeddings, metadata, pert_col=REPLOGLE_PERT_LABEL_COL, keys_to_drop=[REPLOGLE_CONTROL_PERT_LABEL])\n",
-    "map_data = aggregate(embeddings, metadata, pert_col=REPLOGLE_PERT_LABEL_COL, control_key=REPLOGLE_CONTROL_PERT_LABEL)\n",
-    "del embeddings, metadata\n",
-    "metrics = multivariate_benchmark(map_data, recall_thr_pairs=recall_threshold_pairs, pert_col=REPLOGLE_PERT_LABEL_COL)\n",
-    "plot_recall(metrics)\n",
-    "metrics.groupby('source')['recall_0.05_0.95'].mean()"
+    "for fn in [centerscale_on_controls, tvn_on_controls]:\n",
+    "    embeddings = fn(embeddings, metadata, pert_col=REPLOGLE_PERT_LABEL_COL, control_key=REPLOGLE_CONTROL_PERT_LABEL)\n",
+    "    consistency_pvals = univariate_consistency_benchmark(embeddings, metadata, pert_col=REPLOGLE_PERT_LABEL_COL, keys_to_drop=[REPLOGLE_CONTROL_PERT_LABEL])\n",
+    "    map_data = aggregate(embeddings, metadata, pert_col=REPLOGLE_PERT_LABEL_COL, control_key=REPLOGLE_CONTROL_PERT_LABEL)\n",
+    "    for right_sided in [True, False]:\n",
+    "        metrics = multivariate_benchmark(map_data, recall_thr_pairs=recall_threshold_pairs, pert_col=REPLOGLE_PERT_LABEL_COL, n_null_samples = 10000, n_iterations = 1, right_sided=right_sided)\n",
+    "        plot_recall({'Replogle map': metrics}, right_sided=right_sided, title=\"Right tail only\" if right_sided else \"Both tails\")\n",
+    "        print(metrics.groupby('source')['recall_0.05_0.95'].mean())"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "fc927454",
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {