Merge pull request #72 from Lotfollahi-lab/revision

Revision
Lotfollahi-lab · Aug 22, 2024 · 46a63d0 · 46a63d0
2 parents d22ef4e + 58ffcc7
commit 46a63d0
Show file tree

Hide file tree

Showing 9 changed files with 1,508 additions and 1,921 deletions.
diff --git a/docs/tutorials/notebooks/mouse_brain_multimodal.ipynb b/docs/tutorials/notebooks/mouse_brain_multimodal.ipynb
diff --git a/docs/tutorials/notebooks/mouse_cns_sample_integration.ipynb b/docs/tutorials/notebooks/mouse_cns_sample_integration.ipynb
diff --git a/docs/tutorials/notebooks/mouse_cns_single_sample.ipynb b/docs/tutorials/notebooks/mouse_cns_single_sample.ipynb
diff --git a/docs/tutorials/notebooks/mouse_cns_spatial_reference_mapping.ipynb b/docs/tutorials/notebooks/mouse_cns_spatial_reference_mapping.ipynb
diff --git a/src/nichecompass/models/nichecompass.py b/src/nichecompass/models/nichecompass.py
@@ -179,7 +179,10 @@ class NicheCompass(BaseModelMixin):
         in masks but can be learned de novo).
     cat_covariates_embeds_nums:
         List of number of embedding nodes for all categorical covariates.
-
+    use_cuda_if_available:
+        If `True`, use cuda if available.
+    seed:
+        Random seed to get reproducible results.
     kwargs:
         NicheCompass kwargs (to support legacy versions).
     """
@@ -240,6 +243,8 @@ def __init__(self,
                  n_addon_gp: int=100,
                  cat_covariates_embeds_nums: Optional[List[int]]=None,
                  include_edge_kl_loss: bool=True,
+                 use_cuda_if_available: bool=True,
+                 seed: int=0,
                  **kwargs):
         self.adata = adata
         self.adata_atac = adata_atac
@@ -282,6 +287,15 @@ def __init__(self,
         self.active_gp_thresh_ratio_ = active_gp_thresh_ratio
         self.active_gp_type_ = active_gp_type
         self.include_edge_kl_loss_ = include_edge_kl_loss
+        self.seed_ = seed
+
+        # Set seed for reproducibility
+        np.random.seed(self.seed_)
+        if use_cuda_if_available & torch.cuda.is_available():
+            torch.cuda.manual_seed(self.seed_)
+            torch.manual_seed(self.seed_)
+        else:
+            torch.manual_seed(self.seed_)
 
         # Retrieve gene program masks
         if gp_targets_mask_key in adata.varm:
@@ -590,7 +604,7 @@ def train(self,
               lambda_l1_masked: float=0.,
               l1_targets_categories: Optional[list]=["target_gene"],
               l1_sources_categories: Optional[list]=None,
-              lambda_l1_addon: float=0.,
+              lambda_l1_addon: float=30.,
               edge_val_ratio: float=0.1,
               node_val_ratio: float=0.1,
               edge_batch_size: int=256,

diff --git a/src/nichecompass/train/trainer.py b/src/nichecompass/train/trainer.py
@@ -155,8 +155,10 @@ def __init__(self,
         print("\n--- INITIALIZING TRAINER ---")
 
         # Set seed and use GPU if available
+        np.random.seed(self.seed_)
         if use_cuda_if_available & torch.cuda.is_available():
             torch.cuda.manual_seed(self.seed_)
+            torch.manual_seed(self.seed_)
             self.device = torch.device("cuda")
         else:
             torch.manual_seed(self.seed_)

diff --git a/src/nichecompass/utils/__init__.py b/src/nichecompass/utils/__init__.py
@@ -11,9 +11,10 @@
 from .gene_programs import (add_gps_from_gp_dict_to_adata,
                             extract_gp_dict_from_collectri_tf_network,
                             extract_gp_dict_from_nichenet_lrt_interactions,
-                            extract_gp_dict_from_mebocost_es_interactions,
+                            extract_gp_dict_from_mebocost_ms_interactions,
                             extract_gp_dict_from_omnipath_lr_interactions,
                             filter_and_combine_gp_dict_gps,
+                            filter_and_combine_gp_dict_gps_v2,
                             get_unique_genes_from_gp_dict)
 
 __all__ = ["add_gps_from_gp_dict_to_adata",
@@ -25,9 +26,10 @@
            "visualize_communication_gp_network",
            "extract_gp_dict_from_collectri_tf_network",
            "extract_gp_dict_from_nichenet_lrt_interactions",
-           "extract_gp_dict_from_mebocost_es_interactions",
+           "extract_gp_dict_from_mebocost_ms_interactions",
            "extract_gp_dict_from_omnipath_lr_interactions",
            "filter_and_combine_gp_dict_gps",
+           "filter_and_combine_gp_dict_gps_v2",
            "get_gene_annotations",
            "generate_enriched_gp_info_plots",
            "plot_non_zero_gene_count_means_dist",

diff --git a/src/nichecompass/utils/gene_programs.py b/src/nichecompass/utils/gene_programs.py
diff --git a/src/nichecompass/utils/utils.py b/src/nichecompass/utils/utils.py
@@ -18,8 +18,8 @@ def load_R_file_as_df(R_file_path: str,
                       save_df_to_disk: bool=False,
                       df_save_path: Optional[str]=None) -> pd.DataFrame:
     """
-    Helper to load an R file either from ´url´ if specified or from ´file_path´ 
-    on disk and convert it to a pandas DataFrame.
+    Helper to load an R file either from ´url´ if specified or from
+    ´R_file_path´ on disk and convert it to a pandas DataFrame.
 
     Parameters
     ----------
@@ -39,7 +39,7 @@ def load_R_file_as_df(R_file_path: str,
     """
     if url is None:
         if not os.path.exists(R_file_path):
-            raise ValueError("Please specify a valid ´file_path´ or ´url´.")
+            raise ValueError("Please specify a valid ´R_file_path´ or ´url´.")
         result_odict = pyreadr.read_r(R_file_path)
     else:
         result_odict = pyreadr.read_r(pyreadr.download_file(url, R_file_path))