Merge branch 'main' of https://github.com/BorgwardtLab/rnaglib-experiments

Author: wisskarrou

plotting_scripts/make_table_splitting_publication.py

@@ -2,7 +2,7 @@
 import pandas as pd
 import seaborn as sns
 import matplotlib.pyplot as plt
-import matplotlib.lines as mlines # Import mlines
+import matplotlib.lines as mlines  # Import mlines
 import matplotlib.patches as mpatches
 
 plt.rcParams["text.usetex"] = True
@@ -114,7 +114,7 @@
 # Add a vertical dotted line between rna_site and rna_site_redundant
 ax = g.ax
 line_pos = list(task_names.keys()).index("rna_site") + 0.5
-ax.axvline(x=line_pos, linestyle="--", color="dimgray", linewidth=2)
+ax.axvline(x=line_pos, ymax=0.75, linestyle="--", color="dimgray", linewidth=2)
 
 # Create handles and labels manually
 handles = []
@@ -132,7 +132,7 @@
     labels.append(distance)
 plt.legend(handles, labels, loc="upper center", ncol=3, title=r"Splitting strategy:", handletextpad=-0.3)
 
-plt.savefig("plotting_scripts/splitting_publication.pdf", format="pdf")
 plt.subplots_adjust(bottom=0.15)  # Adjust the values as needed
+plt.savefig("plotting_scripts/splitting_publication.pdf", format="pdf")
 plt.show()
 plt.clf()

plotting_scripts/make_table_thresholds.py

@@ -1,8 +1,17 @@
 import json
 import pandas as pd
-import matplotlib.pyplot as plt
 import seaborn as sns
 
+import matplotlib.pyplot as plt
+import matplotlib.lines as mlines  # Import mlines
+import matplotlib.patches as mpatches
+
+plt.rcParams["text.usetex"] = True
+plt.rc("font", size=16)  # fontsize of the tick labels
+plt.rc("ytick", labelsize=13)  # fontsize of the tick labels
+plt.rc("xtick", labelsize=13)  # fontsize of the tick labels
+plt.rc("grid", color="grey", alpha=0.2)
+
 # Config
 thresholds = [0.4, 0.6, 0.8, 1]
 SEEDS = [0, 1, 2]
@@ -22,67 +31,97 @@
                 rows.append(
                     {
                         "score": score,
-                        "metric": METRIC,
                         "threshold": f"{threshold:.1f}",
                         "seed": seed,
                         "distance": distance,
                     }
                 )
 
 # Add random performance over three seeds
-for seed in SEEDS:
-    path = f"results/thresholds_rna_site_redundant_rand_0.4_{seed}.json"
-    with open(path) as result:
-        result = json.load(result)
-        score = result[METRIC]
-        rows.append(
-            {
-                "score": score,
-                "metric": METRIC,
-                "threshold": "random",
-                "seed": seed,  # simulate different seeds
-                "distance": "rand",  # used for filtering
-            }
-        )
+for threshold in thresholds:
+    for seed in SEEDS:
+        path = f"results/thresholds_rna_site_redundant_rand_0.4_{seed}.json"
+        with open(path) as result:
+            result = json.load(result)
+            score = result[METRIC]
+            rows.append(
+                {
+                    "score": score,
+                    "threshold": f"{threshold:.1f}",
+                    "seed": seed,  # simulate different seeds
+                    "distance": "rand",  # used for filtering
+                }
+            )
 
 # Build DataFrame
 df = pd.DataFrame(rows)
 df.to_csv("thresholds_with_random.csv")
 
-# Ensure x-axis order: random first, then thresholds
-task_order = ["random"] + [f"{t:.1f}" for t in thresholds]
-
 # Create a column to use for legend hue, exclude 'rand'
-df["distance"] = df["distance"].apply(lambda d: d if d in ["struc", "seq"] else "random")
+# df["distance"] = df["distance"].apply(lambda d: d if d in ["struc", "seq"] else "random")
 
 # Custom color palette (you can tweak these)
 palette = {
     "struc": sns.color_palette()[0],
     "seq": sns.color_palette()[1],
     "random": sns.color_palette()[2],
 }
+
+dist_names = {
+    "struc": r"Structure",
+    "seq": r"Sequence",
+    "rand": r"Random"
+}
+df["distance"] = df["distance"].replace(dist_names)
+
+palette_dict = sns.color_palette("muted")
+palette_dict = {
+    r"Structure": palette_dict[0],
+    r"Sequence": palette_dict[9],
+    r"Random": palette_dict[3],
+}
 print(df)
 
-# Plot
-g = sns.catplot(
+# df_no_rand = df[df["distance"] != r"Random"]
+ax = sns.lineplot(
     data=df,
-    x="threshold",
-    y="score",
-    hue="distance",
-    kind="bar",
-    height=4,
-    aspect=1,
-    order=task_order,
-    palette=palette,
+    x="threshold",  # X-axis is the threshold
+    y="score",  # Y-axis is the score
+    hue="distance",  # Differentiate lines by distance
+    errorbar='sd',  # Use standard deviation for the error band
     legend=False,
+    palette=palette_dict,
+    linewidth=2,
 )
 
-# Format axes
-g.set_axis_labels("Threshold", "Test Score")
-g.set(ylim=(0.5, 0.85))
-g.despine(left=True)
+# Set axis labels and title
+plt.xlabel("Threshold")
+plt.ylabel("Test Score")
+plt.tight_layout()
+
+
+# Create handles and labels manually
+handles = []
+labels = []
+for i, distance in enumerate(dist_names.values()):
+    # Create a dummy rectangle for each distance, using the color from the plot
+    color = palette_dict[distance]  # Get color for this distance
+    # color = sns.color_palette()[i]  # Get color for this distance
+    # handle = mpatches.Circle((0, 0), radius=0.5, color=color)
+    # Using mlines.Line2D with marker 'o'
+    handle = mlines.Line2D([], [], color=color, marker='o', linestyle='None', markersize=10, )
+    # label=distance_name) # Use Line2D with marker 'o' and set markersize
+    # handle = plt.Rectangle((0, 0), 1, 1, color=color)  # Create a rectangle with that color
+    handles.append(handle)
+    labels.append(distance)
+plt.legend(handles, labels, loc="upper center", ncol=3, title=r"Splitting strategy:", handletextpad=-0.3)
 
-# Save + show
-plt.savefig("thresholds.pdf", format="pdf")
+# # Format axes
+# g.set_axis_labels("Threshold", "Test Score")
+ax.set(ylim=(0.5, 1))
+sns.despine()
+#
+# # Save + show
+plt.savefig("plotting_scripts/thresholds.pdf", format="pdf")
 plt.show()
-plt.clf()
+plt.clf()