Upload folder using huggingface_hub

app.py

@@ -562,6 +562,27 @@ def proteins_tab():
                 ], width=6),
             ], className="mb-2"),
             dbc.Row([
+                dbc.Col([
+                    html.Label("Cross-filter: field", className="fw-semibold small mb-1"),
+                    dcc.Dropdown(
+                        id="table-cross-field",
+                        options=[{"label": f.replace("_"," ").title(), "value": f}
+                                 for f in EXTRACTION_FIELDS[1:]],
+                        placeholder="Select a field…",
+                        clearable=True,
+                        style={"fontSize": "13px"},
+                    ),
+                ], width=3),
+                dbc.Col([
+                    html.Label("Cross-filter: value", className="fw-semibold small mb-1"),
+                    dcc.Dropdown(
+                        id="table-cross-value",
+                        options=[],
+                        placeholder="Select a value…",
+                        clearable=True,
+                        style={"fontSize": "13px"},
+                    ),
+                ], width=5),
                 dbc.Col([
                     html.Label("Show columns", className="fw-semibold small mb-1"),
                     dcc.Dropdown(
@@ -572,7 +593,7 @@ def proteins_tab():
                         placeholder="Select columns…",
                         style={"fontSize": "13px"},
                     ),
-                ], width=12),
+                ], width=4),
             ]),
         ], className="filter-panel"),
         dbc.Row([
@@ -645,19 +666,41 @@ def clustering_tab():
                     placeholder="All groups",
                     style={"fontSize": "13px"},
                 ),
-            ], width=4),
+            ], width=3),
+            dbc.Col([
+                html.Label("Cross-filter: field", className="fw-semibold small mb-1"),
+                dcc.Dropdown(
+                    id="dd-cross-field",
+                    options=[{"label": f.replace("_"," ").title(), "value": f}
+                             for f in CLUSTER_FIELDS],
+                    placeholder="No cross-filter",
+                    clearable=True,
+                    style={"fontSize": "13px"},
+                ),
+            ], width=3),
+            dbc.Col([
+                html.Label("Cross-filter: value", className="fw-semibold small mb-1"),
+                dcc.Dropdown(
+                    id="dd-cross-value",
+                    options=[],
+                    placeholder="Select a value…",
+                    clearable=True,
+                    disabled=True,
+                    style={"fontSize": "13px"},
+                ),
+            ], width=3),
             dbc.Col([
                 dbc.RadioItems(
                     id="plot-type",
                     options=[
-                        {"label": " UMAP cluster plot", "value": "cluster"},
-                        {"label": " Cluster distribution", "value": "distribution"},
+                        {"label": " UMAP", "value": "cluster"},
+                        {"label": " Distribution", "value": "distribution"},
                     ],
                     value="cluster",
                     inline=True,
                     className="mb-2 mt-4",
                 ),
-            ], width=4),
+            ], width=2),
             dbc.Col([
                 html.Label("Show top N clusters", className="fw-semibold small mb-1"),
                 dcc.Dropdown(
@@ -1183,6 +1226,14 @@ def readme_tab():
                         html.Span("✕ Clear  ", className="fw-semibold"),
                         "Resets the search value and group filter, returning to the full dataset.",
                     ]),
+                    dbc.ListGroupItem([
+                        html.Span("Cross-filter: field / value  ", className="fw-semibold"),
+                        "Select a purification field and one of its most common values to restrict "
+                        "the table to proteins where that field matches. For example, set field = "
+                        "Organism Source and value = 'Escherichia coli', then add Elution Buffer "
+                        "to the visible columns to see all elution buffers used with E. coli proteins. "
+                        "Combines with group and search filters.",
+                    ]),
                 ], flush=True, className="mb-3"),
                 html.H6("Table", className="fw-bold mt-2"),
                 dbc.ListGroup([
@@ -1254,6 +1305,16 @@ def readme_tab():
                         "are merged into a light-gray \"Other\" trace. Choose \"All\" to colour every "
                         "cluster individually (may be slow for large fields).",
                     ]),
+                    dbc.ListGroupItem([
+                        html.Span("Cross-filter: field / value  ", className="fw-semibold"),
+                        "Restrict the plot to proteins that belong to a specific cluster in a "
+                        "different field. For example, set cross-filter field = Organism Source "
+                        "and value = 'Escherichia coli', then set Field = Elution Buffer — the "
+                        "UMAP and distribution chart will show elution buffer clusters only for "
+                        "E. coli proteins. The value dropdown is populated with the top 200 most "
+                        "common cluster labels for the selected cross-filter field and current "
+                        "model / threshold settings.",
+                    ]),
                 ], flush=True, className="mb-3"),
                 html.H6("UMAP cluster plot", className="fw-bold mt-2"),
                 dbc.ListGroup([
@@ -1487,7 +1548,7 @@ def render_tab(tab):
     return html.Div()
 
 
-def _apply_filters(groups, search_field, search_value):
+def _apply_filters(groups, search_field, search_value, cross_field=None, cross_value=None):
     filtered = df.copy()
     if groups:
         filtered = filtered[filtered["groups"].apply(
@@ -1497,9 +1558,32 @@ def _apply_filters(groups, search_field, search_value):
         filtered = filtered[
             filtered[search_field].astype(str).str.contains(search_value, case=False, na=False)
         ]
+    if cross_field and cross_value and cross_field in filtered.columns:
+        filtered = filtered[
+            filtered[cross_field].astype(str).str.contains(cross_value, case=False, na=False)
+        ]
     return filtered
 
 
+@app.callback(
+    Output("table-cross-value", "options"),
+    Output("table-cross-value", "disabled"),
+    Output("table-cross-value", "value"),
+    Input("table-cross-field", "value"),
+)
+def populate_table_cross_values(cross_field):
+    if not cross_field or cross_field not in df.columns:
+        return [], True, None
+    vals = (
+        df[cross_field].dropna().astype(str)
+        .loc[lambda s: s.str.strip() != ""]
+        .value_counts()
+        .head(200)
+        .index.tolist()
+    )
+    return [{"label": v, "value": v} for v in vals], False, None
+
+
 @app.callback(
     Output("protein-table", "data"),
     Output("protein-table", "hidden_columns"),
@@ -1509,9 +1593,11 @@ def _apply_filters(groups, search_field, search_value):
     Input("search-field", "value"),
     Input("search-value", "value"),
     Input("col-selector", "value"),
+    Input("table-cross-field", "value"),
+    Input("table-cross-value", "value"),
 )
-def filter_proteins(groups, search_field, search_value, selected_cols):
-    filtered = _apply_filters(groups, search_field, search_value)
+def filter_proteins(groups, search_field, search_value, selected_cols, cross_field, cross_value):
+    filtered = _apply_filters(groups, search_field, search_value, cross_field, cross_value)
     cols = selected_cols if selected_cols else DEFAULT_COLS
     hidden = [c for c in TABLE_FIELDS if c not in cols]
     records = filtered[TABLE_FIELDS].to_dict("records")
@@ -1529,12 +1615,14 @@ def filter_proteins(groups, search_field, search_value, selected_cols):
     Input("filter-group", "value"),
     Input("search-field", "value"),
     Input("search-value", "value"),
+    Input("table-cross-field", "value"),
+    Input("table-cross-value", "value"),
     prevent_initial_call=True,
 )
-def show_detail(selected_rows, groups, search_field, search_value):
+def show_detail(selected_rows, groups, search_field, search_value, cross_field, cross_value):
     if not selected_rows:
         return html.Div()
-    filtered = _apply_filters(groups, search_field, search_value)
+    filtered = _apply_filters(groups, search_field, search_value, cross_field, cross_value)
     clicked = selected_rows[0]
     if clicked >= len(filtered):
         return dash.no_update
@@ -1999,6 +2087,31 @@ def _distribution_figure(cdf, field_name, top_n=20):
     return fig
 
 
+@app.callback(
+    Output("dd-cross-value", "options"),
+    Output("dd-cross-value", "disabled"),
+    Output("dd-cross-value", "value"),
+    Input("dd-cross-field", "value"),
+    Input("dd-model", "value"),
+    Input("dd-min", "value"),
+    Input("dd-threshold", "value"),
+)
+def populate_cross_filter_values(cross_field, model, min_val, threshold):
+    if not cross_field or not all([model, min_val, threshold]):
+        return [], True, None
+    cdf = _load_cluster_csv(model, min_val, threshold)
+    if cdf is None:
+        return [], True, None
+    labels = (
+        cdf[(cdf["field"] == cross_field) & (cdf["cluster_id"] != -1)]["cluster_label"]
+        .value_counts()
+        .head(200)
+        .index.tolist()
+    )
+    options = [{"label": l, "value": l} for l in labels]
+    return options, False, None
+
+
 @app.callback(
     Output("cluster-graph", "figure"),
     Output("plot-status", "children"),
@@ -2009,8 +2122,10 @@ def _distribution_figure(cdf, field_name, top_n=20):
     Input("plot-type", "value"),
     Input("top-n-clusters", "value"),
     Input("cluster-group-filter", "value"),
+    Input("dd-cross-field", "value"),
+    Input("dd-cross-value", "value"),
 )
-def update_cluster_plot(model, min_val, threshold, field, plot_type, top_n, groups):
+def update_cluster_plot(model, min_val, threshold, field, plot_type, top_n, groups, cross_field, cross_value):
     empty_fig = go.Figure()
     empty_fig.update_layout(paper_bgcolor="#ffffff", plot_bgcolor="#f9fafc")
     if not all([model, min_val, threshold, field, plot_type]):
@@ -2020,6 +2135,13 @@ def update_cluster_plot(model, min_val, threshold, field, plot_type, top_n, grou
     if cdf is None:
         return empty_fig, f"Data file not found for model={model} min={min_val} t={threshold}"
 
+    # Apply cross-field filter: keep only (key, protein_index) in the selected cross-field cluster
+    if cross_field and cross_value:
+        keep = cdf[(cdf["field"] == cross_field) & (cdf["cluster_label"] == cross_value)][["key", "protein_index"]]
+        if keep.empty:
+            return empty_fig, f"No proteins found for {cross_field} = '{cross_value}'."
+        cdf = cdf.merge(keep, on=["key", "protein_index"], how="inner")
+
     field_df = cdf[cdf["field"] == field].copy()
     if field_df.empty:
         return empty_fig, f"No data for field '{field}' in this parameter combination."
@@ -2036,7 +2158,12 @@ def update_cluster_plot(model, min_val, threshold, field, plot_type, top_n, grou
             return empty_fig, f"No data for the selected group(s) in this field."
 
     n = top_n or 0
-    suffix = f" — {', '.join(groups)}" if groups else ""
+    suffix_parts = []
+    if groups:
+        suffix_parts.append(", ".join(groups))
+    if cross_field and cross_value:
+        suffix_parts.append(f"{cross_field.replace('_',' ')}={cross_value}")
+    suffix = f" — {' | '.join(suffix_parts)}" if suffix_parts else ""
     if plot_type == "cluster":
         return _umap_figure(field_df, field + suffix, top_n=n), ""
     else: