label pbmc10k clusters (pinellolab#338)

* add script to label clusters * set pbmc10k cell_state to celltype * set pbmc10k embedding to umap * update pbmc10k config * include instructions to download pyensembl reference * fix case matching in ribosomal gene filtration * update pbmc10k metrics * update pipeline lock file * compare standard to coarse-grained clusterings * add pbmc10k coarse clustering * update lock file
lijc0804 · Jun 5, 2023 · e3d50cf · e3d50cf
1 parent 32c8156
commit e3d50cf
Show file tree

Hide file tree

Showing 9 changed files with 243 additions and 124 deletions.
diff --git a/pyrovelocity/config.py b/pyrovelocity/config.py
@@ -374,8 +374,8 @@ def create_reports_config(model_name: str, model_number: int):
                 "pyrovelocity",
                 "pbmc10k",
                 2,
-                "tsne",
-                cell_state="leiden",
+                "umap",
+                cell_state="celltype",
                 offset=True,
                 max_epochs=2000,
             ),
@@ -726,9 +726,9 @@ def create_reports_config(model_name: str, model_number: int):
             "pyrovelocity",
             "pbmc10k",
             2,
-            "tsne",
+            "umap",
             gpu_id=0,
-            cell_state="leiden",
+            cell_state="celltype",
             offset=True,
             max_epochs=2000,
         ),

diff --git a/pyrovelocity/plots/rainbow.py b/pyrovelocity/plots/rainbow.py
@@ -119,7 +119,9 @@ def adjust_subfigure(subfig):
 
 
 def pareto_frontier_genes(volcano_data, num_genes):
-    volcano_data = volcano_data.loc[~volcano_data.index.str.startswith(("Rpl", "Rps"))]
+    volcano_data = volcano_data.loc[
+        ~volcano_data.index.str.contains(("^Rpl|^Rps"), case=False)
+    ]
     pareto_frontier = pd.DataFrame()
 
     if len(volcano_data) < num_genes:

diff --git a/reproducibility/figures/config.yaml b/reproducibility/figures/config.yaml
@@ -559,7 +559,7 @@ model_training:
     metrics_path: models/pbmc10k_model2/metrics.json
     run_info_path: models/pbmc10k_model2/run_info.json
     vector_field_parameters:
-      basis: tsne
+      basis: umap
     gpu_id: 0
     training_parameters:
       _target_: pyrovelocity.api.train_model
@@ -573,7 +573,7 @@ model_training:
       likelihood: Poisson
       num_samples: 30
       log_every: 100
-      cell_state: leiden
+      cell_state: celltype
       patient_improve: 0.0001
       patient_init: 45
       seed: 99