Update app.py

Added raw frequencies and scaled frequencies and importance for each.

app.py

@@ -126,6 +126,9 @@ def predict(file_obj):
             kmer_vector = scaler.transform(kmer_vector.reshape(1, -1))
             X_tensor = torch.FloatTensor(kmer_vector).to(device)
             
+            # Get raw frequency vector before scaling
+            raw_freq_vector = sequence_to_kmer_vector(seq)
+            
             # Get predictions and feature importance
             with torch.no_grad():
                 output = model(X_tensor)
@@ -135,8 +138,8 @@ def predict(file_obj):
             importance = model.get_feature_importance(X_tensor)
             kmer_importance = importance[0].cpu().numpy()
             
-            # Weight importance by actual k-mer frequency
-            kmer_importance *= kmer_vector[0]
+            # Normalize importance scores to match original scale
+            kmer_importance = kmer_importance / np.max(np.abs(kmer_importance)) * 0.002
             
             # Get top 10 k-mers
             top_k = 10
@@ -145,7 +148,7 @@ def predict(file_obj):
                 {
                     'kmer': list(kmer_dict.keys())[list(kmer_dict.values()).index(i)],
                     'importance': float(kmer_importance[i]),
-                    'frequency': float(kmer_vector[0][i])
+                    'frequency': float(raw_freq_vector[i])
                 }
                 for i in top_indices
             ]
@@ -163,7 +166,7 @@ Non-human probability: {float(probs[0][0]):0.4f}
 Most influential k-mers:"""
             
             for kmer in important_kmers:
-                result += f"\n  {kmer['kmer']}: importance={kmer['importance']:.4f}, frequency={kmer['frequency']:.4f}"
+                result += f"\n  {kmer['kmer']}: importance={kmer['importance']:.4f}, raw_freq={kmer['raw_frequency']:.4f} ({kmer['raw_frequency']*100:.2f}%), scaled_freq={kmer['scaled_frequency']:.4f}"
             
             results.append(result)
     except Exception as e: