Standardize docking success criterion for failure mode analysis

BioinfoMachineLearning · Jan 20, 2025 · 1a6b003 · 1a6b003
1 parent 788f6bf
commit 1a6b003
Showing 1 changed file with 8 additions and 6 deletions.
diff --git a/notebooks/failure_modes_analysis_plotting.ipynb b/notebooks/failure_modes_analysis_plotting.ipynb
@@ -763,6 +763,7 @@
     "# sns.histplot(posebusters_complex_pdb_deposition_dates_df[\"Deposition Date\"].values, bins=25)\n",
     "# plt.xlabel(\"Deposition Date\")\n",
     "# plt.xticks(rotation=45)\n",
+    "# plt.tight_layout()\n",
     "# plt.savefig(\"posebusters_benchmark_complex_pdb_deposition_dates.png\")\n",
     "# plt.show()\n",
     "# plt.close(\"all\")"
@@ -783,8 +784,7 @@
    "source": [
     "# find PDB IDs of complexes for which the correct (e.g., RMSD ≤ 2 Å & PB-Valid) binding conformation was not found by any method\n",
     "for dataset in datasets:\n",
-    "    # NOTE: for DockGen, we consider centroid RMSD (cRMSD) ≤ 1 Å as a surrogate docking success criterion\n",
-    "    docking_success_column = \"cRMSD ≤ 1 Å\" if dataset == \"dockgen\" else \"RMSD ≤ 2 Å & PB-Valid\"\n",
+    "    docking_success_column = \"RMSD ≤ 2 Å & PB-Valid\"\n",
     "\n",
     "    for repeat_index in range(1, max_num_repeats_per_method + 1):\n",
     "        dataset_results_table = globals()[f\"results_table_{repeat_index}\"].loc[\n",
@@ -916,8 +916,7 @@
    "source": [
     "# find complexes that AlphaFold 3 failed to correctly predict\n",
     "for dataset in datasets:\n",
-    "    # NOTE: for DockGen, we consider centroid RMSD (cRMSD) ≤ 1 Å as a surrogate docking success criterion\n",
-    "    docking_success_column = \"cRMSD ≤ 1 Å\" if dataset == \"dockgen\" else \"RMSD ≤ 2 Å & PB-Valid\"\n",
+    "    docking_success_column = \"RMSD ≤ 2 Å & PB-Valid\"\n",
     "\n",
     "    for repeat_index in range(1, max_num_repeats_per_method + 1):\n",
     "        dataset_results_table = globals()[f\"results_table_{repeat_index}\"].loc[\n",
@@ -1076,13 +1075,15 @@
     "    sns.histplot(pdb_report_df[\"Refinement Resolution (Å)\"].values)\n",
     "    plt.xlim(0, 10)\n",
     "    plt.xlabel(\"Refinement Resolution (Å)\")\n",
+    "    plt.tight_layout()\n",
     "    plt.savefig(f\"{report_type}_pdb_report_resolution.png\")\n",
     "    plt.show()\n",
     "\n",
     "    plt.close(\"all\")\n",
     "\n",
     "    sns.histplot(pdb_report_df[\"Deposition Date\"].values)\n",
     "    plt.xlabel(\"Deposition Date\")\n",
+    "    plt.tight_layout()\n",
     "    plt.savefig(f\"{report_type}_pdb_report_deposition_date.png\")\n",
     "    plt.show()\n",
     "\n",
@@ -1104,8 +1105,7 @@
    "source": [
     "# find PoseBusters Benchmark set and CASP15 complexes that AlphaFold 3 failed to correctly predict\n",
     "for dataset in [\"posebusters_benchmark\", \"casp15\"]:\n",
-    "    # NOTE: for DockGen, we consider centroid RMSD (cRMSD) ≤ 1 Å as a surrogate docking success criterion\n",
-    "    docking_success_column = \"cRMSD ≤ 1 Å\" if dataset == \"dockgen\" else \"RMSD ≤ 2 Å & PB-Valid\"\n",
+    "    docking_success_column = \"RMSD ≤ 2 Å & PB-Valid\"\n",
     "\n",
     "    for repeat_index in range(1, max_num_repeats_per_method + 1):\n",
     "        dataset_results_table = globals()[f\"results_table_{repeat_index}\"].loc[\n",
@@ -1295,6 +1295,7 @@
     "sns.histplot(failed_complex_pdb_deposition_dates_df[\"Deposition Date\"].values, bins=25)\n",
     "plt.xlabel(\"Deposition Date\")\n",
     "plt.xticks(rotation=45)\n",
+    "plt.tight_layout()\n",
     "plt.savefig(\"af3_failed_complex_pdb_deposition_dates.png\")\n",
     "plt.show()\n",
     "plt.close(\"all\")\n",
@@ -1317,6 +1318,7 @@
     "sns.histplot(top_match_pdb_deposition_dates_df[\"Deposition Date\"].values, bins=30)\n",
     "plt.xlabel(\"Deposition Date\")\n",
     "plt.xticks(rotation=45)\n",
+    "plt.tight_layout()\n",
     "plt.savefig(\"af3_failed_complex_top_match_pdb_deposition_dates.png\")\n",
     "plt.show()\n",
     "plt.close(\"all\")\n",