Update app.py

app.py

@@ -84,25 +84,21 @@ def sequence_to_kmer_vector(sequence: str, k: int = 4) -> np.ndarray:
 ###############################################################################
 # 3. SHAP-VALUE (ABLATION) CALCULATION
 ###############################################################################
-
 def calculate_shap_values(model, x_tensor):
     model.eval()
     device = next(model.parameters()).device
     
     try:
-        # Create background as a torch tensor instead of a numpy array
+        # Create background as a torch tensor
         background = torch.zeros((300, x_tensor.shape[1]), device=device)
-        
-        # Use DeepExplainer with the torch tensor background
         explainer = shap.DeepExplainer(model, background)
         shap_values_all = explainer.shap_values(x_tensor)
-        
         # Get SHAP values for human class (index 1)
         shap_values = shap_values_all[1][0]
     except Exception as e:
         print(f"DeepExplainer failed, falling back to KernelExplainer: {str(e)}")
         
-        # Define a wrapper that converts inputs to proper shape and tensor
+        # Define a wrapper that ensures input is a 2D tensor
         def model_predict(x):
             if not isinstance(x, np.ndarray):
                 x = np.array(x)
@@ -114,14 +110,14 @@ def calculate_shap_values(model, x_tensor):
                 probs = torch.softmax(output, dim=1)[:, 1]  # Human probability
                 return probs.cpu().numpy()
         
-        # Use a numpy background for KernelExplainer
+        # Use a numpy background with 300 samples
         background = np.zeros((300, x_tensor.shape[1]))
-        
         explainer = shap.KernelExplainer(model_predict, background)
         x_numpy = x_tensor.cpu().numpy()
-        shap_values = explainer.shap_values(x_numpy, nsamples=100)
+        # Increase nsamples to 1000 to provide enough data for regression
+        shap_values = explainer.shap_values(x_numpy, nsamples=1000)
     
-    # Get human probability from model prediction
+    # Get human probability from the model prediction
     with torch.no_grad():
         output = model(x_tensor)
         probs = torch.softmax(output, dim=1)