flexynesis 0.2.0__tar.gz → 0.2.1__tar.gz

{flexynesis-0.2.0 → flexynesis-0.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: flexynesis
-Version: 0.2.0
+Version: 0.2.1
 Summary: A deep-learning based multi-omics bulk sequencing data integration suite with a focus on (pre-)clinical endpoint prediction.
 Author-email: Bora Uyar <bora.uyar@mdc-berlin.de>, Taras Savchyn <Taras.Savchyn@mdc-berlin.de>, Ricardo Wurmus <Ricardo.Wurmus@mdc-berlin.de>, Ahmet Sarigun <Ahmet.Sariguen@mdc-berlin.de>
 Project-URL: homepage, https://github.com/BIMSBbioinfo/flexynesis

{flexynesis-0.2.0 → flexynesis-0.2.1}/flexynesis/data.py

@@ -680,7 +680,15 @@ class MultiOmicDatasetNW(Dataset):
             if indices:  # Ensure there are common features in this data type
                 all_features[:, :, i] = data_matrix[:, indices]
         return all_features
+    
+    def subset(self, indices):
+        # Create a subset of the main multiomic dataset
+        dataset_subset = self.multiomic_dataset.subset(indices)
         
+        # Create a new instance of MultiOmicDatasetNW with the subsetted multiomic dataset
+        return MultiOmicDatasetNW(dataset_subset, self.interaction_df.copy())
+
+
     def __getitem__(self, idx):
         node_features_tensor = self.node_features_tensor[idx]
         y_dict = {target_name: self.labels[target_name][idx] for target_name in self.labels}

{flexynesis-0.2.0 → flexynesis-0.2.1}/flexynesis/models/__init__.py

@@ -3,5 +3,4 @@ from .supervised_vae import supervised_vae
 from .triplet_encoder import MultiTripletNetwork
 from .crossmodal_pred import CrossModalPred
 from .gnn_early import GNN
-
 __all__ = ["DirectPred", "supervised_vae", "MultiTripletNetwork", "CrossModalPred", "GNN"]

{flexynesis-0.2.0 → flexynesis-0.2.1}/flexynesis/models/gnn_early.py

@@ -95,15 +95,16 @@ class GNN(pl.LightningModule):
             for var in self.variables:
                 self.log_vars[var] = nn.Parameter(torch.zeros(1))
         
-        self.encoders = flexGCN(
-                        node_count = dataset[0][0].shape[0], #number of nodes
-                        node_feature_count= dataset[0][0].shape[1], # number of node features
-                        node_embedding_dim=int(self.config["node_embedding_dim"]),  
-                        num_convs = int(self.config['num_convs']), # Number of convolutional layers 
-                        output_dim=self.config["latent_dim"],
-                        act = self.config['activation'],
-                        conv = self.gnn_conv_type
-        )
+        self.encoders = nn.ModuleList([ 
+            flexGCN(
+                node_count = dataset[0][0].shape[0], #number of nodes
+                node_feature_count= dataset[0][0].shape[1], # number of node features
+                node_embedding_dim=int(self.config["node_embedding_dim"]),  
+                num_convs = int(self.config['num_convs']), # Number of convolutional layers 
+                output_dim=self.config["latent_dim"],
+                act = self.config['activation'],
+                conv = self.gnn_conv_type
+            )])
 
         # Init output layers
         self.MLPs = nn.ModuleDict()
@@ -129,14 +130,15 @@ class GNN(pl.LightningModule):
         Returns:
             dict: Outputs from the MLPs, one for each target variable.
         """
-        embeddings = self.encoders(x, edge_index)
+        # notice we are using the first encoder (it is currently a early fusion method)
+        embeddings = self.encoders[0](x, edge_index)
         outputs = {}
         for var, mlp in self.MLPs.items():
             outputs[var] = mlp(embeddings)
         return outputs
 
             
-    def training_step(self, batch):
+    def training_step(self, batch, batch_idx, log = True):
         """
         Performs a training step including loss calculation and logging.
 
@@ -164,10 +166,11 @@ class GNN(pl.LightningModule):
 
         total_loss = self.compute_total_loss(losses)
         losses["train_loss"] = total_loss
-        self.log_dict(losses, on_step=False, on_epoch=True, prog_bar=True, batch_size=len(batch))
+        if log:
+            self.log_dict(losses, on_step=False, on_epoch=True, prog_bar=True, batch_size=len(batch))
         return total_loss
 
-    def validation_step(self, batch):
+    def validation_step(self, batch, batch_idx, log = True):
         """
         Performs a validation step, computing losses for a batch of data.
 
@@ -194,7 +197,8 @@ class GNN(pl.LightningModule):
 
         total_loss = sum(losses.values())
         losses["val_loss"] = total_loss
-        self.log_dict(losses, on_step=False, on_epoch=True, prog_bar=True, batch_size=len(batch))
+        if log:
+            self.log_dict(losses, on_step=False, on_epoch=True, prog_bar=True, batch_size=len(batch))
         return total_loss
             
     def configure_optimizers(self):
@@ -341,7 +345,8 @@ class GNN(pl.LightningModule):
         for x, y_dict, samples in dataloader:
             x = x.to(device)  # Move data to GPU
 
-            embeddings = self.encoders(x, edge_index).detach().cpu().numpy()  # Compute embeddings and move to CPU
+            # notice we are using the first encoder (it is currently a early fusion method)
+            embeddings = self.encoders[0](x, edge_index).detach().cpu().numpy()  # Compute embeddings and move to CPU
             all_embeddings.append(embeddings)
             sample_ids.extend(samples)  
             
@@ -455,8 +460,12 @@ class GNN(pl.LightningModule):
             features = dataset.common_features
             target_class_label = dataset.label_mappings[target_var].get(i) if target_var in dataset.label_mappings else ''
             for l in range(len(layers)): 
-                # extracting node feature attributes coming from different omic layers 
-                importances = imp[i].squeeze().detach().numpy()[:,l]
+                # Extracting node feature attributes coming from different omic layers
+                importances_array = imp[i].squeeze().detach().numpy()
+                if importances_array.ndim == 1:
+                    importances = importances_array  # Use the array as is if it is 1-dimensional
+                else:
+                    importances = importances_array[:, l]  # Use the original indexing for 2D arrays
                 df_list.append(pd.DataFrame({'target_variable': target_var, 
                                              'target_class': i, 
                                              'target_class_label': target_class_label,

{flexynesis-0.2.0 → flexynesis-0.2.1}/flexynesis.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: flexynesis
-Version: 0.2.0
+Version: 0.2.1
 Summary: A deep-learning based multi-omics bulk sequencing data integration suite with a focus on (pre-)clinical endpoint prediction.
 Author-email: Bora Uyar <bora.uyar@mdc-berlin.de>, Taras Savchyn <Taras.Savchyn@mdc-berlin.de>, Ricardo Wurmus <Ricardo.Wurmus@mdc-berlin.de>, Ahmet Sarigun <Ahmet.Sariguen@mdc-berlin.de>
 Project-URL: homepage, https://github.com/BIMSBbioinfo/flexynesis

{flexynesis-0.2.0 → flexynesis-0.2.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "flexynesis"
-version = "0.2.0"
+version = "0.2.1"
 authors = [
     {name = "Bora Uyar", email = "bora.uyar@mdc-berlin.de"},
     {name = "Taras Savchyn", email = "Taras.Savchyn@mdc-berlin.de"},