fix AF2 models multimer prediction fetch

folding_studio_demo/app.py

@@ -341,6 +341,16 @@ def create_antibody_discovery_tab():
         inputs=[prediction_dataframe],
         outputs=[prediction_dataframe],
     )
+    with gr.Row(visible=False) as explanation_row:
+        gr.Markdown(
+            """
+            We now have the predicted structures along with the models confidence scores of all complexes. Let's see if we can find a correlation 
+            between the confidence scores and the binding affinity.
+            Spearman and Pearson are statistical methods commonly used to measure the correlation between 
+            two variables. Higher values indicate a stronger correlation.
+            Here **Boltz Complex ipLDDT** is the best predictor of binding affinity.
+            """,
+        )
     with gr.Row(visible=False) as correlation_row:
         with gr.Column(scale=0):
             with gr.Row():
@@ -366,7 +376,7 @@ def create_antibody_discovery_tab():
                 label="Score data to display",
                 choices=SCORE_COLUMNS,
                 multiselect=False,
-                value=SCORE_COLUMNS[0],
+                value="Boltz Complex ipLDDT",
             )
             score_description = gr.Markdown(
                 get_score_description(correlation_column.value)
@@ -386,12 +396,14 @@ def create_antibody_discovery_tab():
             ),
             gr.Row(visible=True),
             gr.Row(visible=True),
+            gr.Row(visible=True)
         ),
         inputs=[correlation_type],
         outputs=[
             prediction_dataframe,
             correlation_ranking_plot,
             regression_plot,
+            explanation_row,
             correlation_row,
             regression_row,
         ],