feat: Add proprietary data disclosure checkbox to submission form

Adds a checkbox for participants to declare whether proprietary (non-public)
data was used in training. The flag is stored in the submission metadata,
propagated through the leaderboard pipeline, and displayed as a
"Proprietary Data" (Yes/No) column on both leaderboards.

Backend handles missing field gracefully via COALESCE in DuckDB queries
and explicit column defaulting in validation metadata builders, ensuring
backwards compatibility with existing submissions.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

app.py

@@ -83,6 +83,7 @@ _ACTIVITY_EMPTY = pd.DataFrame(
         "username",
         "Submitted",
         "model_report_link",
+        "Proprietary Data",
         "MAE",
         "RAE",
         "R2",
@@ -96,6 +97,7 @@ _STRUCTURE_EMPTY = pd.DataFrame(
         "username",
         "Submitted",
         "model_report_link",
+        "Proprietary Data",
         "LDDT-PLI",
         "BiSyRMSD",
         "LDDT-LP",
@@ -118,6 +120,12 @@ def _prepare_activity_df(df: pd.DataFrame) -> pd.DataFrame:
     df["Submitted"] = pd.to_datetime(df["Submitted"], utc=True).dt.strftime(
         "%Y-%m-%d %H:%M UTC"
     )
+    df["Proprietary Data"] = (
+        df["used_proprietary_data"].fillna(False).map({True: "Yes", False: "No"})
+        if "used_proprietary_data" in df.columns
+        else "No"
+    )
+    df = df.drop(columns=["used_proprietary_data"], errors="ignore")
     return df
 
 
@@ -134,6 +142,12 @@ def _prepare_structure_df(df: pd.DataFrame) -> pd.DataFrame:
         "%Y-%m-%d %H:%M UTC"
     )
     df["model_report_link"] = df["model_report_link"].fillna("")
+    df["Proprietary Data"] = (
+        df["used_proprietary_data"].fillna(False).map({True: "Yes", False: "No"})
+        if "used_proprietary_data" in df.columns
+        else "No"
+    )
+    df = df.drop(columns=["used_proprietary_data"], errors="ignore")
     return df
 
 
@@ -247,6 +261,7 @@ def submit_predictions(
     affiliation,
     model_tag,
     paper_checkbox,
+    proprietary_data_checkbox,
     track_select,
     file_input,
 ):
@@ -264,6 +279,7 @@ def submit_predictions(
         affiliation: Institutional affiliation (optional).
         model_tag: Link to method report (optional).
         paper_checkbox: Opt-in for future publication inclusion.
+        proprietary_data_checkbox: Whether proprietary data was used in training.
         track_select: "Activity Prediction" or "Structure Prediction".
         file_input: Path to the uploaded submission file.
 
@@ -387,6 +403,7 @@ def submit_predictions(
         affiliation=affiliation.strip(),
         model_report_link=model_tag.strip(),
         include_in_publication=paper_checkbox,
+        used_proprietary_data=proprietary_data_checkbox,
         track=track_select,
         filename=file_path.name,
     )
@@ -588,6 +605,7 @@ We thank the experimentalists at **Octant** and **UCSF** (Fraser Lab) for all th
                             "username",
                             "Submitted",
                             "model_report_link",
+                            "Proprietary Data",
                             "MAE",
                             "RAE",
                             "R2",
@@ -605,6 +623,7 @@ We thank the experimentalists at **Octant** and **UCSF** (Fraser Lab) for all th
                             "str",
                             "str",
                             "str",
+                            "str",
                         ],
                     )
                 with gr.TabItem("Structure Prediction (Pose)"):
@@ -620,6 +639,7 @@ We thank the experimentalists at **Octant** and **UCSF** (Fraser Lab) for all th
                             "username",
                             "Submitted",
                             "model_report_link",
+                            "Proprietary Data",
                             "LDDT-PLI",
                             "BiSyRMSD",
                             "LDDT-LP",
@@ -635,6 +655,7 @@ We thank the experimentalists at **Octant** and **UCSF** (Fraser Lab) for all th
                             "str",
                             "str",
                             "str",
+                            "str",
                         ],
                     )
             leaderboard_timer.tick(fn=load_activity_leaderboard, outputs=[activity_lb])
@@ -700,6 +721,11 @@ We thank the experimentalists at **Octant** and **UCSF** (Fraser Lab) for all th
                         label="Include me in a future Challenge publication",
                         value=False,
                     )
+                    proprietary_data_checkbox = gr.Checkbox(
+                        label="I used proprietary data (not publicly available) in training my model",
+                        value=False,
+                        info="Displayed publicly on the leaderboard.",
+                    )
 
                 # --- Column 3: Track & file ---
                 with gr.Column(scale=1):
@@ -739,6 +765,7 @@ We thank the experimentalists at **Octant** and **UCSF** (Fraser Lab) for all th
                     affiliation,
                     model_tag,
                     paper_checkbox,
+                    proprietary_data_checkbox,
                     track_select,
                     file_input,
                 ],

models.py

@@ -37,6 +37,7 @@ class Submission(BaseModel):
         affiliation: Institutional affiliation — stored privately.
         model_report_link: URL to method report (required before deadline).
         include_in_publication: Opt-in for Challenge publication authorship.
+        used_proprietary_data: Whether proprietary data was used in training.
         track: Competition track.
         filename: Original uploaded filename.
         s3_key: Full S3 object key for the uploaded prediction file.
@@ -60,6 +61,7 @@ class Submission(BaseModel):
     affiliation: str = ""
     model_report_link: str = ""
     include_in_publication: bool = False
+    used_proprietary_data: bool = False
 
     # --- submission ---
     track: Literal["Activity Prediction", "Structure Prediction"]