comparison/masif_site/masif_vs_sppider/masif_sppider_Intpred_comp.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pymesh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import sys\n",
    "sys.path.append(\"../../../source/\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "ground_truth_dir = 'ground_truth/'\n",
    "masif_pred_dir = 'masif_pred/'\n",
    "transient_interactions_file = 'testing_transient.txt'\n",
    "\n",
    "# List of sppider_predictions\n",
    "sppider_pred_dir = 'sppider_pred/'\n",
    "\n",
    "# List of intpred_predictions\n",
    "intpred_pred_dir = 'intpred_pred/intpred_pdbs/'\n",
    "\n",
    "# List of psiver_predictions\n",
    "psiver_pred_dir = 'psiver_pred/psiver_pdbs/'\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The list that we will use has 53 proteins which are the ones that worked on both SPPIDER and IntPred.\n",
    "transient_pdbs = set([''.join(x[0:4]) for x in open(transient_interactions_file).readlines()])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# To follow SPPIDER's implementation  we only label as interface those residues \n",
    "# with points with more than 4% RSA change and at least 5A RSA change.\n",
    "def is_res_iface(res_labels, vertex_areas):\n",
    "    sum_iface_area = 0.0\n",
    "    sum_total_res_area = 0.0\n",
    "\n",
    "    for ix, label in enumerate(res_labels): \n",
    "        sum_total_res_area += vertex_areas[ix]\n",
    "        if label > 0:\n",
    "            sum_iface_area += vertex_areas[ix]\n",
    "\n",
    "    if sum_iface_area > 5.0 and sum_iface_area/sum_total_res_area > 0.04:\n",
    "        return True\n",
    "    else:\n",
    "        return False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get the ground truth per residue (as we have it per surface point.)\n",
    "# Atoms: list of PDB atoms in Biopython format. \n",
    "# Residues: list of PDB residues in Biopython format. \n",
    "# ground_truth_mesh: a pymesh mesh object with labeled interface (the ground truth.)\n",
    "# Returns a dict where the key is the residue id and the value is the score. \n",
    "def assign_labels_per_residue(atoms, residues, ground_truth_mesh):\n",
    "    # For every residue, assign a label on whether it is interface or not, based on the buried surface.\n",
    "    flann = pyflann.FLANN()\n",
    "    at_coord = np.array([at.get_coord() for at in atoms]).astype(float)\n",
    "    verts = ground_truth_mesh.vertices.astype(float)\n",
    "\n",
    "    labels = ground_truth_mesh.get_attribute('vertex_iface')\n",
    "    ground_truth_mesh.add_attribute(\"vertex_area\")\n",
    "    vertex_areas = ground_truth_mesh.get_attribute(\"vertex_area\")\n",
    "\n",
    "    # Compute the closest atom to each vertex. \n",
    "    r,d = flann.nn(at_coord, verts)\n",
    "    d = np.sqrt(d)\n",
    "    n= len(r)\n",
    "    \n",
    "    # For each residue, assign the ground truth based on the closest atom. \n",
    "    res_labels = {}\n",
    "    res_areas = {}\n",
    "    for ix, v in enumerate(verts): \n",
    "        pred = labels[ix]\n",
    "        res_id = atoms[r[ix]].get_parent().get_id()\n",
    "        if res_id not in res_labels:\n",
    "            res_labels[res_id] = []\n",
    "            res_areas[res_id] = []\n",
    "        res_labels[res_id].append(pred)\n",
    "        res_areas[res_id].append(vertex_areas[ix])\n",
    "    res_labels_tmp = []\n",
    "    for key in res_labels:\n",
    "        # To follow SPPIDER's implementation (as an approximation) we only label as interface those residues \n",
    "        # with points with more than 4% RSA change and at least 5A RSA change.\n",
    "        if is_res_iface(res_labels[key], res_areas[key]):\n",
    "            res_labels[key] = 1.0\n",
    "        else:\n",
    "            res_labels[key] = 0.0\n",
    "    return res_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline\n",
    "# Compute the score of each predicted surface. \n",
    "from IPython.core.debugger import set_trace\n",
    "from Bio.PDB import *\n",
    "import pyflann\n",
    "import sklearn.metrics\n",
    "import os\n",
    "\n",
    "# Compute the ROC AUC for SPPIDER/INTPRED for points and residues. \n",
    "# SPPIDER/INTPRED work on residues while we work on points. \n",
    "# This method computes ROC AUCs according to both. \n",
    "def compute_roc_auc_sppider_intpred_psiver(pred_dir, pdbid):\n",
    "    parser = PDBParser()\n",
    "    struct = parser.get_structure(pred_dir+pdbid+'.pdb', pred_dir+pdbid+'.pdb')\n",
    "    atoms = Selection.unfold_entities(struct, 'A')\n",
    "    residues = Selection.unfold_entities(struct, 'R')\n",
    "    at_coord = np.array([at.get_coord() for at in atoms]).astype(float)\n",
    "    \n",
    "    ground_truth_mesh = pymesh.load_mesh(ground_truth_dir+pdbid+'.ply')\n",
    "\n",
    "\n",
    "    flann = pyflann.FLANN()\n",
    "    \n",
    "    verts = ground_truth_mesh.vertices.astype(float)\n",
    "    \n",
    "    # Compute the closest atom to each vertex in vertices\n",
    "    r,d = flann.nn(at_coord, verts)\n",
    "    d = np.sqrt(d)\n",
    "    n= len(r)\n",
    "    \n",
    "    # For each vertex, assign the predictions according to the closest atoms.\n",
    "    pred = np.zeros((n))\n",
    "    labels = ground_truth_mesh.get_attribute('vertex_iface')\n",
    "    \n",
    "    labels_per_residue = assign_labels_per_residue(atoms, residues, ground_truth_mesh)\n",
    "    \n",
    "    res_pred = {}\n",
    "    for res in residues:\n",
    "        if 'CA' in res:\n",
    "            bfactor = res['CA'].bfactor\n",
    "            # If the bfactor is less than zero it means that it was ignored by the program (by intpred in this case.)\n",
    "            if bfactor >= 0:\n",
    "                res_pred[res.get_id()] = bfactor/100\n",
    "\n",
    "    res_pred_y = []\n",
    "    res_label_y = []\n",
    "    for key in labels_per_residue:\n",
    "        if key in res_pred and key in labels_per_residue:\n",
    "            res_pred_y.append(res_pred[key])\n",
    "            res_label_y.append(labels_per_residue[key])\n",
    "\n",
    "    try:\n",
    "        res_roc_auc = sklearn.metrics.roc_auc_score(res_label_y, res_pred_y)\n",
    "    except:\n",
    "        res_roc_auc = 0.5\n",
    "    print('Per residue ROC AUC: {:.2f}'.format(res_roc_auc))\n",
    "    \n",
    "    for vi in range(n):\n",
    "        pred[vi] = atoms[r[vi]].bfactor/100\n",
    "        \n",
    "    roc_auc = sklearn.metrics.roc_auc_score(labels, pred)\n",
    "    print('{} Per point ROC AUC {} : {:.2f}'.format(pred_dir, pdbid, roc_auc))\n",
    "    return labels, pred, res_label_y, res_pred_y, roc_auc, res_roc_auc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Per residue ROC AUC: 0.45\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 1XQS_A : 0.43\n",
      "Per residue ROC AUC: 0.41\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4XL5_C : 0.53\n",
      "Per residue ROC AUC: 0.54\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4HDO_A : 0.57\n",
      "Per residue ROC AUC: 0.49\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3BH6_B : 0.45\n",
      "Per residue ROC AUC: 0.42\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4Y61_B : 0.40\n",
      "Per residue ROC AUC: 0.72\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 5TIH_A : 0.57\n",
      "Per residue ROC AUC: 0.57\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3RT0_A : 0.57\n",
      "Per residue ROC AUC: 0.48\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3MCA_B : 0.57\n",
      "Per residue ROC AUC: 0.43\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4X33_B : 0.33\n",
      "Per residue ROC AUC: 0.68\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3O5T_A : 0.58\n",
      "Per residue ROC AUC: 0.84\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4ZGY_A : 0.80\n",
      "Per residue ROC AUC: 0.64\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4ZRJ_A : 0.54\n",
      "Per residue ROC AUC: 0.35\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 1XG2_A : 0.40\n",
      "Per residue ROC AUC: 0.41\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4DVG_B : 0.41\n",
      "Per residue ROC AUC: 0.43\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3VYR_B : 0.46\n",
      "Per residue ROC AUC: 0.46\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2WUS_A : 0.49\n",
      "Per residue ROC AUC: 0.60\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3QML_D : 0.56\n",
      "Per residue ROC AUC: 0.64\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2V3B_A : 0.55\n",
      "Per residue ROC AUC: 0.68\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4FZV_A : 0.67\n",
      "Per residue ROC AUC: 0.39\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3VV2_A : 0.33\n",
      "Per residue ROC AUC: 0.60\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 1F6M_A : 0.51\n",
      "Per residue ROC AUC: 0.52\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 1ZHH_A : 0.48\n",
      "Per residue ROC AUC: 0.45\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4YEB_A : 0.47\n",
      "Per residue ROC AUC: 0.32\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2V9T_B : 0.30\n",
      "Per residue ROC AUC: 0.26\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 1EWY_A : 0.30\n",
      "Per residue ROC AUC: 0.52\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2Y5B_A : 0.38\n",
      "Per residue ROC AUC: 0.52\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 1JTD_B : 0.44\n",
      "Per residue ROC AUC: 0.82\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4ETP_A : 0.79\n",
      "Per residue ROC AUC: 0.65\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3FHC_A : 0.57\n",
      "Per residue ROC AUC: 0.59\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 1WDW_H : 0.55\n",
      "Per residue ROC AUC: 0.46\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3H6G_B : 0.47\n",
      "Per residue ROC AUC: 0.47\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4V0O_F : 0.52\n",
      "Per residue ROC AUC: 0.58\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4C9B_B : 0.60\n",
      "Per residue ROC AUC: 0.49\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4BKX_B : 0.47\n",
      "Per residue ROC AUC: 0.51\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2QXV_A : 0.41\n",
      "Per residue ROC AUC: 0.41\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2AYO_A : 0.36\n",
      "Per residue ROC AUC: 0.38\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 1I2M_B : 0.40\n",
      "Per residue ROC AUC: 0.26\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3SJA_I : 0.44\n",
      "Per residue ROC AUC: 0.59\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2XWT_C : 0.51\n",
      "Per residue ROC AUC: 0.74\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3WN7_A : 0.56\n",
      "Per residue ROC AUC: 0.47\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2I3T_A : 0.56\n",
      "Per residue ROC AUC: 0.62\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 5J57_A : 0.62\n",
      "Per residue ROC AUC: 0.37\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3NVN_A : 0.58\n",
      "Per residue ROC AUC: 0.35\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4M0W_A : 0.37\n",
      "Per residue ROC AUC: 0.60\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4YC7_B : 0.52\n",
      "Per residue ROC AUC: 0.81\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4ETP_B : 0.84\n",
      "Per residue ROC AUC: 0.46\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 4LVN_A : 0.48\n",
      "Per residue ROC AUC: 0.40\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 3ZWL_B : 0.38\n",
      "Per residue ROC AUC: 0.56\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 1W1W_B : 0.49\n",
      "Per residue ROC AUC: 0.91\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2F4M_A : 0.83\n",
      "Per residue ROC AUC: 0.42\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2C0L_A : 0.38\n",
      "Per residue ROC AUC: 0.45\n",
      "intpred_pred/intpred_pdbs/ Per point ROC AUC 2PMS_A : 0.40\n",
      "Number of proteins: 52\n",
      "IntPred Global ROC AUC: 0.5027\n",
      "IntPred Global ROC AUC per residue: 0.5382\n",
      "IntPred Median ROC AUC per protein : 0.4914\n",
      "IntPred Median ROC AUC per residue per protein: 0.4896\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x130b97f50>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Benchmark IntPred\n",
    "   \n",
    "all_pred = []\n",
    "all_labels = []\n",
    "\n",
    "all_pred_res = []\n",
    "all_labels_res = []\n",
    "\n",
    "all_roc_auc_per_prot_intpred = []\n",
    "all_res_roc_auc_per_prot_intpred = []\n",
    "\n",
    "for fn in os.listdir(intpred_pred_dir):  \n",
    "    \n",
    "    pdbid_chain = fn.split('.')[0]\n",
    "    pdbid = pdbid_chain.split('_')[0]\n",
    "    if pdbid not in transient_pdbs: \n",
    "        continue\n",
    "\n",
    "    l, p, res_l, res_p, roc_auc, res_roc_auc = compute_roc_auc_sppider_intpred_psiver(intpred_pred_dir, pdbid_chain)\n",
    "\n",
    "    all_pred.append(p)\n",
    "    all_labels.append(l)\n",
    "    \n",
    "    all_pred_res.append(res_p)\n",
    "    all_labels_res.append(res_l)\n",
    "    \n",
    "    all_roc_auc_per_prot_intpred.append(roc_auc)\n",
    "    all_res_roc_auc_per_prot_intpred.append(res_roc_auc)\n",
    "    \n",
    "print(\"Number of proteins: {}\".format(len(all_pred)))\n",
    "\n",
    "all_pred = np.concatenate(all_pred, axis=0)\n",
    "all_labels = np.concatenate(all_labels, axis=0)\n",
    "\n",
    "all_pred_by_point_intpred = all_pred.copy()\n",
    "all_labels_by_point_intpred = all_labels.copy()\n",
    "\n",
    "#print(metrics.roc_auc_score(all_my_labels_per_res, all_my_pred_per_res))\n",
    "\n",
    "all_labels_res = np.concatenate(all_labels_res, axis=0)\n",
    "\n",
    "all_pred_res = np.concatenate(all_pred_res, axis=0)\n",
    "\n",
    "\n",
    "all_roc_auc = sklearn.metrics.roc_auc_score(all_labels, all_pred)\n",
    "all_roc_auc_res = sklearn.metrics.roc_auc_score(all_labels_res, all_pred_res)\n",
    "\n",
    "\n",
    "\n",
    "print('IntPred Global ROC AUC: {:.4f}'.format(all_roc_auc))\n",
    "print('IntPred Global ROC AUC per residue: {:.4f}'.format(all_roc_auc_res))\n",
    "\n",
    "print('IntPred Median ROC AUC per protein : {:.4f}'.format(np.median(all_roc_auc_per_prot_intpred)))\n",
    "print('IntPred Median ROC AUC per residue per protein: {:.4f}'.format(np.median(all_res_roc_auc_per_prot_intpred)))\n",
    "\n",
    "pos_label = np.where(all_labels_res == 1)[0]\n",
    "pos = all_pred_res[pos_label]\n",
    "neg_label = np.where(all_labels_res == 0)[0]\n",
    "neg = all_pred_res[neg_label]\n",
    "sns.distplot(pos, hist=True, color='r', kde_kws={\"shade\": True}, norm_hist=True, label='true positives')\n",
    "sns.distplot(neg, hist=True, color='b', kde_kws={\"shade\": True}, norm_hist=True, label='true negatives')\n",
    "plt.legend('true positives', 'true negatives')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Per residue ROC AUC: 0.82\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1XQS_A : 0.81\n",
      "Per residue ROC AUC: 0.76\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4XL5_C : 0.73\n",
      "Per residue ROC AUC: 0.76\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4HDO_A : 0.70\n",
      "Per residue ROC AUC: 0.79\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3BH6_B : 0.81\n",
      "Per residue ROC AUC: 0.69\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4Y61_B : 0.68\n",
      "Per residue ROC AUC: 0.33\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 5TIH_A : 0.33\n",
      "Per residue ROC AUC: 0.71\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3RT0_A : 0.61\n",
      "Per residue ROC AUC: 0.64\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3MCA_B : 0.57\n",
      "Per residue ROC AUC: 0.54\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4X33_B : 0.40\n",
      "Per residue ROC AUC: 0.62\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3O5T_A : 0.52\n",
      "Per residue ROC AUC: 0.67\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4ZGY_A : 0.57\n",
      "Per residue ROC AUC: 0.45\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4ZRJ_A : 0.45\n",
      "Per residue ROC AUC: 0.77\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1XG2_A : 0.74\n",
      "Per residue ROC AUC: 0.62\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4DVG_B : 0.62\n",
      "Per residue ROC AUC: 0.58\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3VYR_B : 0.55\n",
      "Per residue ROC AUC: 0.66\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2WUS_A : 0.61\n",
      "Per residue ROC AUC: 0.77\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3QML_D : 0.83\n",
      "Per residue ROC AUC: 0.48\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2V3B_A : 0.41\n",
      "Per residue ROC AUC: 0.59\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4FZV_A : 0.54\n",
      "Per residue ROC AUC: 0.51\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3VV2_A : 0.54\n",
      "Per residue ROC AUC: 0.63\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1F6M_A : 0.56\n",
      "Per residue ROC AUC: 0.65\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1ZHH_A : 0.54\n",
      "Per residue ROC AUC: 0.54\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4YEB_A : 0.55\n",
      "Per residue ROC AUC: 0.55\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2V9T_B : 0.57\n",
      "Per residue ROC AUC: 0.69\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1EWY_A : 0.67\n",
      "Per residue ROC AUC: 0.32\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1NCA_N : 0.38\n",
      "Per residue ROC AUC: 0.62\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2Y5B_A : 0.59\n",
      "Per residue ROC AUC: 0.91\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1JTD_B : 0.83\n",
      "Per residue ROC AUC: 0.41\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4ETP_A : 0.35\n",
      "Per residue ROC AUC: 0.60\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3FHC_A : 0.68\n",
      "Per residue ROC AUC: 0.52\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1WDW_H : 0.50\n",
      "Per residue ROC AUC: 0.46\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3H6G_B : 0.39\n",
      "Per residue ROC AUC: 0.53\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4V0O_F : 0.52\n",
      "Per residue ROC AUC: 0.70\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4C9B_B : 0.67\n",
      "Per residue ROC AUC: 0.61\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4BKX_B : 0.58\n",
      "Per residue ROC AUC: 0.53\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2QXV_A : 0.54\n",
      "Per residue ROC AUC: 0.63\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2AYO_A : 0.63\n",
      "Per residue ROC AUC: 0.76\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1I2M_B : 0.71\n",
      "Per residue ROC AUC: 0.84\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3SJA_I : 0.78\n",
      "Per residue ROC AUC: 0.55\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2XWT_C : 0.56\n",
      "Per residue ROC AUC: 0.70\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3WN7_A : 0.69\n",
      "Per residue ROC AUC: 0.41\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2I3T_A : 0.40\n",
      "Per residue ROC AUC: 0.57\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 5J57_A : 0.49\n",
      "Per residue ROC AUC: 0.65\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3NVN_A : 0.55\n",
      "Per residue ROC AUC: 0.60\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4M0W_A : 0.61\n",
      "Per residue ROC AUC: 0.57\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4YC7_B : 0.54\n",
      "Per residue ROC AUC: 0.46\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4ETP_B : 0.39\n",
      "Per residue ROC AUC: 0.45\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 4LVN_A : 0.44\n",
      "Per residue ROC AUC: 0.49\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 3ZWL_B : 0.48\n",
      "Per residue ROC AUC: 0.50\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 1W1W_B : 0.44\n",
      "Per residue ROC AUC: 0.50\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2F4M_A : 0.56\n",
      "Per residue ROC AUC: 0.69\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2C0L_A : 0.63\n",
      "Per residue ROC AUC: 0.45\n",
      "psiver_pred/psiver_pdbs/ Per point ROC AUC 2PMS_A : 0.47\n",
      "Number of proteins: 53\n",
      "psiver Global ROC AUC: 0.58\n",
      "psiver Global ROC AUC per residue: 0.61\n",
      "psiver Median ROC AUC per protein : 0.56\n",
      "psiver Median ROC AUC per residue per protein: 0.60\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x129cfd390>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAD8CAYAAABw1c+bAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvIxREBQAAIABJREFUeJzt3Xd8VFX6+PHPSU8IJIQESCOhF+mE3kEREMEulkVdFdvq6u5+/eoW3dXv7vqzrWV1FRXrKk1UEBBlqSotQAgklECAEJJACAESCKnn98cZlgApk2Rm7iTzvF+veU1y5869z4XkyZlzz3mO0lojhBDCc3hZHYAQQgjXksQvhBAeRhK/EEJ4GEn8QgjhYSTxCyGEh5HEL4QQHkYSvxBCeBhJ/EII4WEk8QshhIfxsTqAqoSHh+v4+HirwxBCiEZjy5Ytx7XWEfbs65aJPz4+nsTERKvDEEKIRkMpdcjefaWrRwghPIwkfiGE8DCS+IUQwsPU2sevlJoNTAGOaa17VvH6/wB3VDpedyBCa31CKXUQKADKgTKtdYKjAhdCCFE/9rT4PwImVvei1volrXVfrXVf4Glgjdb6RKVdxtpel6QvhBBuoNbEr7VeC5yobT+b24AvGhSREEIIp3JYH79SKgjzyeDLSps18L1SaotSamYt75+plEpUSiXm5uY6KiwhhBCXcOTN3WuBny7p5hmhte4PTAIeUUqNqu7NWutZWusErXVCRIRdcxCEEELUgyMT/3Qu6ebRWh+xPR8DvgIGOfB8Qggh6sEhM3eVUiHAaODOStuaAV5a6wLb1xOA5xxxPtH4zZpV/Wsza+wUFEI0lD3DOb8AxgDhSqlM4FnAF0Br/Y5tt+uB77XWZyq9tQ3wlVLq/Hk+11p/57jQhRBC1EetiV9rfZsd+3yEGfZZeVs60Ke+gQkhhHAOmbkrhBAeRhK/EEJ4GEn8QgjhYdyyHr9wHzL6RoimR1r8wmVKSiA3FyoqrI5ECM8mLX7hcCUlkJQEGzfC5s3mkZUFp0+b14ODITwc4uJgzBiIirI0XCE8jiR+0SBaw5EjsGmTeaxfb57PnTOvBwdDmzbQsSMEBoKfH5w6BcePw88/w9q1MHAgXHMNtG1b6cA19THVRvqghKiRJH5RJ2fPwoEDkJ4Oixeb1vzRo+Y1Hx/o1AmmTDGJPzISmjev/lhFRZCYCNu2wZYtMH06jKq2mpMQwlEk8XuI+t6kLSmBvXshJQV27YKcHNPKB2jfHvr2hS5doHt306r38zOvLV5ce0yBgTByJAwYAN99B//+N2RkwF13gb/9lyaEqCNJ/OIypaWwfDl89hksXGi+9/GB6GgYOtT0ybdteyHJA+zbZx71ERQE111nun7WrYMrr4Rvb/IlJLDUMRckhLiIJH7xX+fOwbvvwgsvmJZ9SAj06GFa8tHR4OvrvHN7ecGIERARYf7oXHn4Gpb/ehlhzYqdd1IhPJQM5xRoDR9/bPrnH3/ctOb/+ldYsADGj4f4eOcm/cq6doXnnoMdmWGMfWUKx04HuObEQngQSfwerrgYZsyAu+82LfxXXoF//AOGDTPdO1YYOhTevm45e4+GMOaVa8k5FWhNIEI0UZL4PdjRo/D3v5ubqvfcA2++Cf37Wx2VMTzuCLOuX8bBvGDGvjKFbEn+QjiMJH4PlZcHL78MJ0/Ciy+aVr+Xm/00DIrN5r3rl5FxIpixr1xL1skgq0MSoklws1914QqFhfD662ao5s03Q0KC1RFVLyEmh/dvWEpmfhCjXr6WA8drmBgghLCLJH4PU1ICb71lZs5OnWpKJ7i7/tFHmX3jEo4X+DPixamkZoVaHZIQjZoM5/Qwc+aYWbdTpkBMjNlmz2QrV1q8GEKSu122/Y6+O5m/owejXp7K0keXMah9rgXRCdH4SYvfg+zaBT/9ZGrjdOlidTR11yb4LP++9RuCfEoY++oUluyItTokIRolSfwe4tw5+PRTaNkShgyxOpr6axdawOfTv6F9y5NMe/tqPvixq9UhCdHoSOL3EN98Y0byTJjguslYzhLRrIiPb/6Woe2OcN+no/m/Jf3+Wz9ICFG7WhO/Umq2UuqYUmpnNa+PUUqdUkol2R7PVHptolJqj1Jqn1LqKUcGLuy3cSOsWmUKqkVHWx1Nw8xN7sbc5G4s2d2RcR0O0qftUf60aCDjXr2G8gpldXhCNAr2tPg/AibWss86rXVf2+M5AKWUN/AWMAnoAdymlOrRkGBF/Tz9tCmENmKE1ZE4lreX5ror9jA87jCr90Zz2/vjKC6VD7FC1KbWUT1a67VKqfh6HHsQsE9rnQ6glJoDTANS63EsUU8rV5rW/pgxF1fTbCq8FEzofIBmfiXM39KR1OyWPKgvv1ZZm0WICxzVPBqqlNqulFqmlLrCti0aOFxpn0zbtioppWYqpRKVUom5uTJMzxG0hj/9CVq3ht69rY7GTloTc3QLcSe20T4vkfZ5ifiWF9X6tuFxR5jafS+pWS15880LK4AJIS7niHH8W4E4rXWhUmoy8DXQua4H0VrPAmYBJCQkyK06B/juO1Pj/oknQLl597d3eQl998xh7OaXiMy7+HbSOZ9mJEVNZnPMdeQHxVR7jAHROfh6VbAwtRuvvw6PPWYWexFCXKzBiV9rfbrS10uVUm8rpcKBI0DlgdYxtm3CBc639iMjYdIk80fAXbXPXMcdS++gZeFhskO6MjfhJc6dPkeF8sa/vIjeeasZeHghQzLms77dLazo9ADaq+of3d6Rxyjp0I0lS+C11+DXvzb3N4QQFzQ48Sul2gJHtdZaKTUI032UB5wEOiul2mMS/nTg9oaeT9jnu+/MOrZPPunewzd7pS3kjqW3kx8UzXsjPmJ327GgFCHpW/+7z76QAQSX3sOorHkMzZhHm8L9fNnzWYr8Qqo8ZufOZmbyt99eSP5CiAvsGc75BbAe6KqUylRK3auUelAp9aBtl5uAnUqp7cAbwHRtlAG/ApYDu4B5WusU51yGuNTrr5s6PFddZXUk1Ru6/V/MWHwTR0Kv4J9jF7I7cly1fVKFvmEsjXuQb+IepV3+du7b9ADhhQerPXanTqYWUWYmvPoqHDvmpIsQohFS2g1nviQkJOjExESrw2h0zi+onpMDzz5rFlNx11m6Q5Lf5eYVD5ISOZ5Ph7xNqc/FnfGVW/yXij6zl1v2/x2tvPlw4FucCmx70euneo/679cHDpjaP+3bw4oV0K7d5cer70L0QrgTpdQWrbVdtXZl0HMTtGoVeHu770ie2JzNXL/yMXa3Gc1Hw967LOnX5kizLvy787P4lRdxx7bfEVhystp927eHG2+E7Gzzh3DXroZGL0TjJ9U5m5izZ81Inm7drL+pGZK89rJtgSWnuHvj/RT6hLC47b00P5hcr2MfC4xnTsffc2fan7k96Sk+GfAPSr2r/gMSHW2Wk3zySfMJ6JNPYNq0ep1WiCZBEn8T8/PPpuZ+375WR1IFXcH1Kf9HcEkeH3X9O0U+LRp0uIzmV/Blh99y8/4XuX7n/zGv9/Ogqv4Q26kTvP02/OUvcN11Zjbz88+bT0YOVVO/UW2kX0m4iHT1NCEVFaabJzoa2rSxOprLDcmYT6e8TSyPvY+sZnWe6lGlPaFD+D7mHrrl/siIg/+ucd+2bc1N7ylTzFrDQ4aYOkZCeBpJ/E3Inj1mZS13bO03P5fL6PQP2RuSwJbwqx167E2tp7AjbBRj939Ah7zNNe7r5we//a2Z43Dw4IWunxMnHBqSEG5NunqakJ9/Bn9/6NjR6kguNyHtLbwqyvku9j7HTyNWim/bPUzrokPcsOM5Dva/kfyQ+BrfMm6cSfqffgrz5sH69Wbt4atabaFd2JlKe+52bKznlZSYvzxHj5oV70+ehLVrITgYmjUzM+8GDYIBA8z3QjiQJP4m4tQp2LYNevQAHzf7X22fl8gVR1exOvI2Tvq3rf0N9VDqHcC8Dk9x/+7fcfeiG3hz+k+U+dY8WigoCB54wOTabdsgKQk2lQygc+uTjOuaRZ+Y444LUGvIyoLNm2H3bsjIgPLyC683a2bWxCwqguJic5cezE2IQYPgvvtg+nTr79iLJsHNUoSor/nzobTUJH534l1RwqQ9r5HnH8lPba936rnyAyL5Kv5xbtv/V25c+QhzJ3xg16eLFi1g9GjzCSDt+/1sOhzNu+t6EBZ0jgDfCmYMTcPbq57zXQoLYd06czMhO9vEExkJ/ftDVBRERJik7+0N11574X0nT5qxp6mp8OOPcO+98JvfmOc//tEspSZEPUnibyI++ghatTI3MN3J4IwFhJ89zL87PUu5l/PrQqeFDuSH7o9xVcobHIwcysbe99v9Xn9/U+VzaLsj7M1txdqDsfzykzG88kNvXrppI5N6Hq71GP+VkwP/+Y/pQyotNXfcx40zix3b02oPDYWhQ83jl7+EHTvMMmqvvWb6p158EWbMAC+5TSfqThJ/E5CWZhZRHznSvapw+pcVMvzg56S1GMD+kH4uO+/yK56g3Ykkblj5K7Ja9+Vw24F1er+Xgm6t8+gakUdoYDGv/TSQyW9O4tfjdvDijRvx86mo/s15eWaq8IYNJil3725a9+Hh9b8gpcxsvN694bbbTPK/5x744AP4/HOIlUXnRd1I4m8CPv74Qo5xJ0MOzSewrIBV0a6tzaeVN58NfpMnVlzDPd9cxxu3b2Dx4ronR6VgYpcDjOt4iJfXDub1lb1Yn96GufevID688OKdz5yBJUtg9Wrzff/+MHCg4/vkO3WCN94wVfjeftvc/J03z6y0I4Sd5HNiI1dRYRJ/QoK5SekugoryGJIxj9TQoeQEuX6Y0Vn/lnwwYjZ+JQXct3ASAeeqL+tQGz/vCn4/dj2vT/mBXdmhDPt/09h3zDb5rKLCfNx65hmz3Fn37qZrZvRo592I9fKCyZNN4g8KgiuvNFOT3bDulnBP0uJv5DZuNBUoZ8xwr1Wnxia+hF95EWuibrMshpyQbnw0bBb3r5vBPYuuZ9YN31Hu41/v413d5QAdwk5y1/wpjH/1Gtbe8S5x375lhmVGRZkpwa1b1z/gxYvr/p4pU2D5cnPjNzMTXn7Zvfr7hFuSFn8jt2CBGb45dKjVkVzQ/EwOI7a9wY6wUeQGVlEO04X2tR7O3ISX6ZS5mjuW3Yl3WXGDjtc5PJ/Z1y7k1GkY98/r2XckgFVX/IpZ3f/BrJypzEoewqxkF5ZE9fc3o4FuuMHUn5458+JhokJUQVr8jZjWJvG7WzfP2M0v4l1ewtrIW60OBYCtcdcTXJzHtO3PEXz2KB9N/Yqzga3qfiBdQYtdm5i8biHLy99gvFrFlf7r+J82a/BVNdzwdTal4Fe/MsNC33/fDCH95BP3XoFHWEpa/I1YYqKZBzR6tNWRXBBUlMeQ5HfZFjuVEwFRVofzX2u73MenQ/5Ju+xNPPbFUMLz0+x639zkbsxN7sZPa8sI/uANIr/7kFwdzs5uN3Fthx0cKgxnYZob1L9WytxbmDkT5syBu+6Slr+olrT4G7Hz3TzDh1sdyQXDk97Cv+wsq7o9DHlnan+DCyXFTuVkYBT3/Hwfv/20Dz/1eYTVA/+HwqCq++WVLqdz7noGH/6S9vlbOe0bxtfxvyY5bDQoL7qTw5DW+1l5uDNdWh6jX+ssF19RFW6z3VOZNcvc+J01S8b6i8tI4m9kzlf91Rpmz4aYmAsjCK3mW3qWEdveJCVyPDkhXQnJq34VLascDE/gtfHfMjHlZUZvfZVh299ma/c7yQnvyfGQjvjm76d1YTptC9LocCKR0HNHOeUXzoroGWyOmHRZzf+rolM4XBjGx6kDiW3+A+GBZy26skpuu82UfvjgA5P8X39dbviKi0jib6QOHzaVOPu5bl5UrQbtnE3wueOs6vaQ1aHUKL9ZDF8Meo0V3R/lqtQ36J/6Kf7lRRftc9anOVlBnfk+6i72hA5Gq6oL9/t4aW7ukMi/do3h010JPN5vrXvk2HvuMcO83nzT1KT4v/+zOiLhRiTxN1Jbt5pGXKdOVkdieJWXMibxZQ60GsCB8EFWh2OX3OYd+Xzw66A1zYtzaVWYQdjhJI4FtqPAt5XdreSW/me5MiqVJYf7sPloLIPa1qG0g7MoBQ89ZIq9/fWvprbPb39rdVTCTUjib6S2bTMz9QPrtlyt0/TdO4+wgkN8NfwZq0OpO6UoCGhNQUBr8k/Xrz88IeIgSSdimbenL1e0ynFwgHaobg5A586mPtDvfgfJyVXfEJKVvzxOrT/lSqnZSqljSqmd1bx+h1IqWSm1Qyn1s1KqT6XXDtq2JymlEh0ZuCfLyzM1wNq3tzoSG60Zk/gyOS26sCtyvNXRWMJLwbXttnOm1I+v9/WyOpwLvLxg0iSIizPF3bZssToi4Qbsad58BEys4fUDwGitdS/geeDSRUfHaq37aq0T6heiuFRKinl2l8TfMXMN0blJrO18L7qaNW89QWTQaYa0SWftkY5syLJ24tpFvL1h6lRTDvr992H7dqsjEhar9bdUa70WqHZhOq31z1rrfNu3G4AYB8UmqrFzp7lf5y4l2UdveZUC/1ZsiXNuvf3GYEzkbpr7FvHEmmnuVTrH1xeuv96UlHj3XVPnX3gsRzfP7gWWVfpeA98rpbYopWrsSFRKzVRKJSqlEnNzcx0cVtNRVmYWcIqPd48ReuH5aXRP/5b1He+kzDvA6nAs5+9dzvioXWzIjmfuHjdb/Njf3yT/sDBT4G3XLqsjEhZxWOJXSo3FJP7/rbR5hNa6PzAJeEQpNaq692utZ2mtE7TWCREREY4Kq8nZv9+szBcfb3Ukxsitr1Ph5ctPHWdYHYrb6NPqMD1Cs/jfdVMoKnWz8ROBgXDjjRASYoZ6JiVZHZGwgEMSv1KqN/A+ME1rnXd+u9b6iO35GPAV0DjG+bmxnTvN/bp2btCFHFh0goEpH7K13XUUBsgf6/O8FDzZ5zsyClry+rZq2zrWCQqCm2823T7vvGPqeguP0uDEr5RqBywEfqG13ltpezOlVPPzXwMTgCpHBgn77dxpVvHzc/4qhrUasuM9/MvOsqbLfVaH4naGtD7A+KhU/rZxPEfPuFEFvfPOt/xjY+Huu+H//T+p5+9Bav0cqpT6AhgDhCulMoFnAV8ArfU7wDNAK+BtZTqdy2wjeNoAX9m2+QCfa62/c8I1eIzMTMjKMkssukpI8toqt3tVlDFq88ukN+9NUd5ZtyzPYLXf9f6ea5c/yrPrr+adK7+0OpzL+fmZNQR27ICnnjKTQz74wFT5tNesSwfx1YHMH7BMrYlfa13jShpa6/uAy5p8Wut0oM/l7xD1tXy5eXaH/v3ux9bQovg4S2IftDoUtzQ33YxeHhB+iFnJQ2kbVEBU8GkAZvbeYGVoF/PxgT/9yUz0eu89c8N34ULo6PpV04TruNmdJwHVN6JmzTJ19xuybrejDM6Yz3H/KNJa9Lc6FLc2OmoP20/EsiCtN4/1+9HqcKqmlCns1rGjqenTp49ZyeuBB9xj6JhwOM+dbdPIaA1paaYap9W/izEndxJzehebWk8BD56wZY9mPiWMaruHlLxIUvMasCyjKwwaZFr93bqZOj8TJsChQ1ZHJZxAfmsbiWPH4PRpk/itNvjwAoq8m7G91VirQ2kUBrc+QEu/MyxI60N5hZu3oNu0gZdegscfN4vI9+gBf/ubGUMsmgxJ/I3Evn3mOTra2jhanDtK96Nr2BZ+1WW16UXVfLwqmBCTwpHCUNZkNoK+c6Vg2jSz4ENCAvzhD9CzJyxdanVkwkEk8TcSaWlmBF5YmLVxDDq8EIBNra+xNpBGpntoNp1aHOWb/T3JLmxudTj2adsW/vIXePFFKCmBa64xfxAOHLA6MtFAkvgbib17TWvfyv59v7Kz9D+ymF0th3LaTyZs1YVSMDl2B2UVXvxu7VSrw6mbgQNNcbeZM+GHHy50/5SVWR2ZqCcZ1dMI5OebUsy9LK722y9rCQFlZ1jfZpq1gTRSrQLOMLxtGp/v7s+9PTcyrt2+y/aZlTyk2vdbOgzU19eM/LnySnjrLdP9s3QpTJ7sHsPMRJ1Ii78RSEszz1b276uKMgZnzOdQcA+ymnWxLpBGbmTbNGKbneDBFTdxptQNpl/XVUQE/PnP8Pvfmzo/zz8PibLURmMjLf5GIC3NTLK0snZd92NrCT13lO863mtdEHYKSXffWcS+XhU8n/A196y5m1+vuo73J8yzOqTqV++qzfTpptX/3ntmyNm4cY6NSziNtPgbgbQ0iIoyxdksoTXDDn3Bcf8o9obIejoNNaT1Ae7vto4Pdg5m/t7eVodTfyEhpthbp04wd66M+mlEJPG7ucJCyM62dvx+3MntRBXsZUObaTJhy0EevWIlvcMOc//3t3DotJusqFMfPj4wZYqZ9PXNN/D111ZHJOwgv8Vuzh3G7w85NJczPi1IbjXGuiCaGF+vCl4ePJ/yCsXNi2c0zv7+886v69urFyxbBhvcqBaRqJIkfje3b59ZMrVNG2vO37pwP12P/8zmiMmUeflbE0QT1S44nxcGLSTxaAy3LbmTsopG/OuoFIwfbz6afvYZHDlidUSiBo34J80zHDpkbur6WHQbfvjBzyn2CpQJW05yZfQu/thvCYvTr+CxVdc17pL4Xl5mkpefn1ng5dw5qyMS1ZDE78YqKkzit6q13yp/H1fkrGRLxNWc82kks00boTs6beL+rmv51/bhfJveo3En/2bNTPLPzYVPPpHFXdyUJH43dvSoqY3Vtq015x+b+CIVypsNbRrZTNNG6IleK7g+fivfHriCb/b3bNz5MiYGhg+HLVtg+3aroxFVkMTvxs5XxLWixR9SkMnAlI/YFn4lhb4WFwjyAF5K89eErxkQfpBlB7uzIK13407+AwZAq1awYAGUllodjbiEJH43dvCgmSlvRWG20VteQekKfm5znetP7qG8lObadtsZ3Ho/KzK68tnuAVQ01uTv7Q2jR5sun5UrrY5GXEISvxs7eBBat3b9xK3mZ3IYmvwuW9tdxyl/i24weCilYFLMTka23cuPRzowa8dQShvraJ/4eOjQAZYsMTN7hdtopD9RTV95uVlc3YpunnGbXsC7vIQV3R9z/ckFSpnRPlfH7GDbsRje3DaCc2WNtLrKqFGmq+ebb6yORFRiV+JXSs1WSh1TSu2s5nWllHpDKbVPKZWslOpf6bW7lFJptsddjgq8qcvKMr8vrr6xG1KQydDkd0iMu4njzdu79uTiIsPapHND/BbS8iN4deto8oqCrA6p7sLCoG9fs5pXTo7V0Qgbe1v8HwETa3h9EtDZ9pgJ/AtAKRUGPAsMBgYBzyqlGvH8dNc5eNA8u7rFf+XGv6J0BT/0kNa+O+jTKpNbO27mSEEIY+Y93HgWcals0CDTX7lihdWRCBu7Er/Wei1wooZdpgGfaGMDEKqUigSuBn7QWp/QWucDP1DzHxBhc+gQBARAaKjrzhl26gCDd77PxvbTyW8W67oTixp1C83hjk4bSD8Vxsi5v2p8tX2CgsziLevXS1+/m3BUH380cLjS95m2bdVtF7U4f2PXlStuXbXheSrwZkX3R113UmGXDi2OM3vURxw724zRcx/mwKlGNsR2wACzYteaNVZHInCjevxKqZmYbiLatWtncTTWKi01pU4SXFgBuU1eKgmpH7Ou0z2cDrRoxpiHmJtev//Yvq0y+XD0R9y79i5Gz32YlTf/i04t8xwcnZOEhZkRPqtXw9VXm7IOwjKOavEfASr3DcTYtlW3/TJa61la6wStdUKElSuOuIHMTFOuwZX9+9es+1+KfZrxH2ntu7WeLbP4ePSHFJb4MXreI6TlN6JlDwcMMHXGpXqn5RyV+BcBM2yje4YAp7TW2cByYIJSqqXtpu4E2zZRg4wM8+yqxN8xYxVXpH/Lf7o9whn/RtaF4IG6hebw0ZgPOVfmw5h5Dzee5B8TY36oV6wwLRthGXuHc34BrAe6KqUylVL3KqUeVEo9aNtlKZAO7APeAx4G0FqfAJ4HNtsez9m2iRpkZpobu81dMIBD6QquXfs78oOiWNf5l84/oXCIriFH+Wj07MaV/JWC/v1NEardu62OxqPZO6rnNq11pNbaV2sdo7X+QGv9jtb6HdvrWmv9iNa6o9a6l9Y6sdJ7Z2utO9keHzrrQpqSzEwID3fNjd1+u78g9thWlvZ8kjLvAOefUDhMl5BjjS/5d+4M/v5mhI+wjMzcdTMVFebGbrgLfod9SouY9OPvOdyyF9vaSU2exqhLyDE+HG26fcbOf4j9J1tZHVLNfHyga1fYtk2GdlpIEr+bOXTIlGJ2ReIfs+UVwgoyWNz7D2hZS7fR6hpylA9HfcjZEl/GznuI9JNufp+mRw8zdG3BAqsj8Vjy2+5mkpPNs7MHNoUWHGbcpr+zPXoS+1sPc+7JhNN1DT3Kh6M/oqDEn7HzH+bgKTee5BUZCS1bmoVahCXcZhy/MM4n/lZO/sQ+Ze2TeOkKFvf5o3NPJBymtvH/t3ZIZPaoj7hn7d2Mmfcwa259m7gW+S6Krg6UMq3+NWvgwAFoLzWhXE1a/G5m+3ZTpsGZ81vaZ66j3545rOrygJRmaELmpiewIz+a6R02cfRsMAM+e5wXNo1lVvIQq0O7XPfu5g/Ap59aHYlHksTvZrZvd243j6oo57pVv+ZkYCSruj3kvBMJy0Q1O8WMzus5U+rPq1vGkH/ODUdrtWgB/frJurwWka4eN3LmDOzfD0Mc3EALSV7736/7Zy4iJncbX7b/DYEZuwl07KmEm4hudpJfdF7PJ3uH8uqWMczosYXI4AKrw7rYhAnwwgtmaOcwuc/kStLidyMpKabx46wWf0Dpacbtf49DwVeQ0nKkc04i3EZMs3zu7Lyek8UBjJv/EEfPBFsd0sVGjDBri375pdWReBxp8buR8zd2nTWUc+z+DwgoLWRZp/tdW/ZTWKZdcD53dtrAp/uG0v+z3/DbAasJ9iu5aJ+ZvS2qndOsmanf8+WX8PLL8jPpQtLidyPJyaZ0eUiI44/dpiCNAZmLSIyYyLGgeMefQLituOYnuL3jRo6dDeYfW0dxptTX6pAuGDnSTF5JSrI6Eo9MU6KCAAAgAElEQVQiid+NJCebkW0Ob/hozaTdr1PkE8zqqNsdfHDRGHRocZzpHTeRfaYFr28bRZG7rOE7fLhZnWvhQqsj8SiS+N2E1mZET8eOjj92z6MraHdqByuj7uScj5v18wqX6RxyjFs6JHK4IJS3kkZQUu4Gv/4hIdCnjyR+F3OD/3kBpjDbyZNmrQpH8is9w/i0d8kK6kBS+HjHHlw0Ot1Cc7ghfiv7TobzbvIwyircoF995EhITZWKnS4kid9N7Nxpnh09iXHs5hcJKc5leex9aOXt2IOLRqlX2BGmtNvOzrxIPkwZRLnVyX/ECPP81VfWxuFBJPG7idRU8xwX57hjtjx9iLGbX2RHy5EcDu7huAOLRi8h4hAToneSeLQdT6yeZu0cqogIU8JBuntcRhK/m0hNNcuSOnJEz5S1T6JR/CdmhuMOKpqM4W33M7T1Pt5MGsnLiWOsDWbECEhMvLD8nHAqSfxuIiXFsa399kd+pO/eeazs9hCn/Tx7DWNRvQkxKUyO3cGT667ls9T+1gUy0jah8JtvrIvBg7jJmC7PprVp8V95peMOeO2a33EysC2ruzxAUMYuBx1YNDVeCgaEHyQ1vy13L7+N5OORdArN++/rLpvcFRMDsbGwdCk8+qhrzunBpMXvBjIzoaDAcS3+3mlfEpezke+u+B2lPlKNR9TMx6uC6R03E+p3ln9tH8bxoiBrAhk8GFatMkWrhFNJ4ncD52/sxsc3/Fhe5aVM/vFpskO6khh/Y8MPKDxCoE8pt3faSFmFF29vH865MgtGgA0ZYpafW7XK9ef2MJL43YAjE//QHbOIOLmPb3s9LcM3RZ2EBxRyS4dEsgtbMDtlsOtH+vTqBYGBsGSJi0/seexK/EqpiUqpPUqpfUqpp6p4/R9KqSTbY69S6mSl18orvbbIkcE3FSkpZiW6ho7o8S8+zVXr/8K+iKHsbjvWMcEJj9KxRS4TYlLYnhvN94e6uPbkfn6maNvSpVKj38lqvbmrlPIG3gKuAjKBzUqpRVrr1PP7aK2fqLT/o0C/Soco0lr3dVzITcOsWRe+XrUKgoNh8eKGHXP01n/QvCiXD4bNlkqHot6GtE7n8Jkwvt7Xi7WZHRgVk+66kw8eDD/+aFpDPXu67rwexp5RPYOAfVrrdACl1BxgGpBazf63Ac86JrymT2vIyoKuXRt2nMBz+Yza8irJ0RM5HNbHMcEJj6QUTI1LIudsC2799hdsvfNV5yziUlVLp8B2nmefhauvrvn9M2c6PiYPYU/ijwYOV/o+Exhc1Y5KqTigPbCy0uYApVQiUAa8oLX+up6xNkknT8K5czUvrl55Ba3qjNn/AYElp/m5xURC0rc6MELhiQK8y7i142Zm7xnBncvu4Ieb3sVLuaD7pXlzM5N3x47aE7+oN0ff3J0OLNBal1faFqe1TgBuB15TSlVZf1IpNVMplaiUSszNzXVwWO4rO9s815T4axNYcorBGQtIDR0mtfaFw7QJLOBP/Zaw8nBnXto8xnUnbt/erEF69qzrzulh7En8R4DYSt/H2LZVZTrwReUNWusjtud0YDUX9/9X3m+W1jpBa50Q4czVxt1MVpZ5bkjiH5oxF7/yItZE3eqYoISwuSF+KxNjdvLHnyexOSe29jc4Qvv2UFFxYbibcDh7Ev9moLNSqr1Syg+T3C8bnaOU6ga0BNZX2tZSKeVv+zocGE719wY8Una2GcEWVM85M4ElJxl0+EtSWg4nN9CBNR+EAOYdSKBvqwya+RRzzVf38sbWEcxKHsKs5CHOO2lkJAQESOJ3oloTv9a6DPgVsBzYBczTWqcopZ5TSk2ttOt0YI7WF43D6g4kKqW2A6swffzyv1lJVlZDW/vz8C0vZm2ktPaFcwT6lHJj/BaOFwUzb68LBuh5eZnyDSkpMqzTSeyq1aO1XgosvWTbM5d8/+cq3vcz0KsB8TVpWpsWf+fO9Xu/X9kZEjK/ZlfoUI4HuuhjuPBIcc1PMKJtGuuyutAn4gh9IrKde8L4eEhLM78gUVHOPZcHkpm7Fjp9GoqK6t/iH5C5iICyM/zU9gbHBiZEFcZE7qFN4Ck+25VAYYmfc092fhp7Sopzz+OhJPFbqCEjerwrShiSMZ/05r3JbtbJsYEJUQUfrwpuiN/KmVI//r17gHN7YZo3N78Y0s/vFJL4LXQ+8YeF1f29vbO/p3lJHj+1lUJswnXaBp1mbNQuth6L4fPdTq7f366d6e4pKXHueTyQJH4LZWeDvz80a1a39yldzrCDX5AV1IEDzXs7JzghqjG8zT5im+Xx6MrryS5s7rwTxcdDaSns2+e8c3goSfwWys42rf26ltXpeuxHWhVlmta+1OQRLual4Lr4JM6W+fLAipud1+UTEwPe3tLP7wSS+C10PvHX1dCMeZzwb8vuUCeOpRaiBuEBhTzecwWL06/gs10DnHMSX1+IjpbE7wSS+C1y5oypR1XXG7uRp3cTe2onmyKukXr7wlIzOq+nf/ghHl15PUcKWjjnJPHxpoWUn++c43soSfwWqe+N3cEZCyj2CiQpfLzjgxKiDryV5m8JCyku9+b+H25xTpfP+fVIZXSPQ0nit0hOjnmuS+IPLs7jiqOrSAofT4m3ReuiClFJfPMT/KbXDyw72J3ZOwc5/gTh4WaxCkn8DiWJ3yLZ2eDjAy3q8Al5QOY3eOlyNkVc47zAhKijOzttZFBEOk+snsah0y0de3ClzLDO1FRTuE04hCR+i5y/setl5/+Ad0UJAzIXkRYygPyASOcGJ4Qd5qYnMDc9gfkHBjCszT6Ky72ZuPB+3tnu4EEHcXGmRHNGhmOP68Ek8VskO9uss2uvK3JWElyaz6bWU5wXlBD11NK/iKtjUth9og2rDjt4Jrn08zucJH4LnDkDJ07UbUTPwMyvyA2IIb25LKso3NOA8EN0CclhYVpvUo63cdyBg4KgTRtJ/A4kid8Cu3ebZ3tv7Eae3kP06d0kRkyUCVvCbSkF0+KS8Pcu485ld1BS7sDhxnFxZlWuoiLHHdODSeK3wK5d5tneFv+AzG8o8fInudVY5wUlhAME+xYzLW4bSbnR/OmniY47cFycubm7d6/jjunBJPFbIDXV3NQNDa1934DiU/TM+Q87w0ZR7F3Hoj5CWKBr6FFu6bCZlxLHsOJQPRebuFRUlJnJK7N4HUISvwV27TJJ39uOT8IJqZ/gV3GOLeFXOz8wIRzk6T7L6NDiOL9YdjvHzgY3/IDe3mZVLunndwhJ/BZITbWzf19rhm5/hyNBnaXmvmhUAn1KeXXIPE6cC+Su76ZToR1wbyouDnJzzUM0iCR+FysuNveo7Onf73BkHW1PpLIlQlr7ovHpGnKU3/ddxncHu/NK4uiGH/D8qlzS6m8wu9bcFY6zZw+Ul9uX+Iduf4ci3xbsDBvp/MCEcKC56Qm2rzQ9Qo/w1LprOF7UjI6heczsvaF+Bw0NhZAQ088/2gF/SDyYtPhd7Py9qdoSf1BRHr3TvmRL3A2Uefk7PzAhnEApmBafRIh/EbN2DGnYWr1Kme6e3buhrMxxQXoguxK/UmqiUmqPUmqfUuqpKl6/WymVq5RKsj3uq/TaXUqpNNvjLkcG3xjt3GnuU9U2a3fArs/wqShhQ/vbXBOYEE4S4F3GLR02U1Diz+yUQQ3r72/f3vSXyqpcDVJr4ldKeQNvAZOAHsBtSqkeVew6V2vd1/Z43/beMOBZYDAwCHhWKeXgKk6NS0qKGZzgU1Mnm9YM3vE+GWF9yA7t7rLYhHCWqKBTTIrdSUpeJC9sGlf/A8XGmpbTzp2OC84D2dPiHwTs01qna61LgDnANDuPfzXwg9b6hNY6H/gBcOCsjsZn584LpUeq0y5nE5F5O9nYfrprghLCBRLCD9KzZSbP/Hw1Px2Jr99B/PzMqlyS+BvEnsQfDRyu9H2mbdulblRKJSulFiilYuv4Xo9QVATp6RcGJ1Rn8I73KfYJYlusvX9fhXB/SsG1cduJCjrF7Uvv5ERRYP0OdH5VrsOHa91VVM1RN3cXA/Fa696YVv3HdT2AUmqmUipRKZWY20TH6e7eDVrXnPj9Swrot+cLkmKupdjXARNfhHAjAd5lvDpkHlmFLbivvqt2tW9vnpctc2hsnsSexH8EiK30fYxt239prfO01sW2b98HBtj73krHmKW1TtBaJ0RERNgTe6Nz/tNpTYm/7565+JeeYWMH6eYRTVOvsCP8ptf3fLWvN7N21KN2f1gYNG8uib8B7En8m4HOSqn2Sik/YDqwqPIOSqnKK4NMBWxlyFgOTFBKtbTd1J1g2+aRUlLMTd2YmOr3GbzjfXJadOFQWH/XBSaEi93dZT1DW+/nd2uurfuqXUqZ1tOKFVBS4pT4mrpaE7/Wugz4FSZh7wLmaa1TlFLPKaWm2nZ7TCmVopTaDjwG3G177wngecwfj83Ac7ZtHiklxawiV92InjbHU4jL2cjG9rdK+WXRpHkpzfMJX1OhFff/cHPdu3zat4fCQvjpJ6fE19TZNXNXa70UWHrJtmcqff008HQ1750NzG5AjE3DrFmkrJ9O9/BcIhf/h5DkbpftMnLvW5Qrb9J0J0LSt1oQpBDOd2FWL4yN2s2SQ32467vbGBF9wP5ZvedbUMuWwVgpV15XMnPXRc4U+3AgrwWdWlX9gceropTe2d+zN2QgZ31DXBydENZICD9IfHAu8/f2If9cHUb5+PlB796weLHzgmvCJPG7yK5sU3y/U6v8Kl/vcnw9zUpPsi38SleGJYSlvGwlHcoqvFi4r1fd3jxsmBkqJ4uz1JkkfhfZmWXqMFeX+PsdWcJp31bsb9HPlWEJYbkw/7MMbbOfTTlxbMhqZ/8bhw83z9LqrzOpzukiKVkt8fMup13o6ctea34ul455m/ip7Q1o5cB1SoVoJEa2TWNbXjtuW3onTw1cWeXYhsv6/9u2hY4d4Ztv4Le/dU2gTYS0+F0kJbsl7cNO4uN1+fCFPtnf4UUFSa3GWxCZENbz9y7jyqhdHDzdis1HY2t/w3nDhpmRPcePOy+4JkgSv4skZbSiW3je5S9oTd+spRwMvoL8gMjLXxfCQ/RplUFk0Em+TOtNSbmdn3yHDzeLsC9dWvu+4r8k8bvA0aOQfboZ3Vtfnvjj8pMIK8qSm7rC43kpuDpmJyeLg1h3pL19b+rSBcLDTXePsJskfhfYts08d299+cfRfllLOOfdjF0th7k4KiHcT/vmecQHH2f5wW6UltuRnpQy3T3Ll8O5c84PsImQxO8CSUnmuXvExS3+gNICuh9bw46wkbLKlhA2Y6L2cKokkB+z7Gz1Dx8OZ87AypXODawJkcTvAtu2QUzIaVoEXFxXpGfOCnwrStgWfpVFkQnhfuKDj9Mu+DjfHexGaYUdKapvXwgKku6eOpDE7wJbt17e2gfTzZMd1IGcoI4WRCWEe1IKxkTu5WRxED9nxdf+Bj8/GDQIvvpK1uK1kyR+JysogP37L+/fb3t6L5EFaWxrJTd1hbhUh+a5xDbL47sD3SivsKNg4dixkJsLq1c7PbamQBK/kyUnm8VXul3S4u+XtYRS5cfOsFEWRSaE+1LKTOo6UdyMLcdqqGN+3uDB0KwZzJnj/OCaAEn8TnZ+RE+PSkM5VWkxvXJ+YFfLoZzzkVW2hKhK55CjhAcU8MOhLrWXbfb3N6N7Fi6UGv12kMTvZNu2QWgotAk+899tzfckElB2hq3hEyyMTAj35qVgaOv9ZBSEkXYyvPY3jB0L+fnwww/OD66Rk8TvZElJ0KnTxeuqhCavJTcghozgHtYFJkQj0KfVYYJ8ivnhUJfad05IMEsyzp3r/MAaOUn8TlRaatbZ7Vhp0I7/sQwCcw6yJWKirLIlRC18vSoYGHGAHcej2JtfS6vf1xdGjICvv5bJXLWQxO9Eqammu7Fz5wvbQpPXUeHtS3LYGMviEqIxGRRxEG9VwT+2jK5953HjzFA6WYi9RpL4nej8jd3ziV+VnKPFro0UxHaXm7pC2CnYt5jeYZl8nJrAiaJaVunq1w9atoQvvnBNcI2UJH4n2rYNAgIgOtp832LXRrxKiznZYYC1gQnRyAxps5+iMj/e3zmk5h29vWHMGFi0CE5UvcypkMTvVN98YwoHLl0Kc7d3RW/cRE5gPB+fnGp1aEI0Km0CCxgckc4/tw2nrLYyDpMnQ3ExfPaZa4JrhOxK/EqpiUqpPUqpfUqpp6p4/TdKqVSlVLJS6j9KqbhKr5UrpZJsj0WODN6dFRbC4cMXWvuxp3YQWbiPxIhJclNXiHqY0Xk9hwtb8vW+njXv2KkTdOsGs2ZR+wQAz1Rr4ldKeQNvAZOAHsBtSqlLxyFuAxK01r2BBcCLlV4r0lr3tT08pqm7caNZHyIqynw/OGMBRd7BJLcaY2lcQjRWY6L2ENvsBK9vHVn7zpMnQ0qK+UUUl7GnxT8I2Ke1TtdalwBzgGmVd9Bar9Jan7V9uwGwY4510/bTT+Y5KgpanjpIt2Pr2BJxtZRfFqKevJXm9k4b+TGrA1uPRte88/jxEBgI77/vmuAaGXsSfzRwuNL3mbZt1bkXqDyWKkAplaiU2qCUuq4eMTZKP/4IERFmJvmIpH8CsDliksVRCdG43Ri/lSCfYl7bWkuNq6AgM5N3zhwzvFNcxKE3d5VSdwIJwEuVNsdprROA24HXlFJV1iBWSs20/YFIzM3NdWRYLldWBuvXm9a+f0kBg3e8T2rL4RT42THtXAhRrRZ+57ghfitz9vTlSK5fzTtfc41ZoEUKt13GnsR/BKi87H2MbdtFlFJXAn8Apmqti89v11ofsT2nA6uBflWdRGs9S2udoLVOiIiIsPsC3FFysrm5Gx0NCSkfE1hyio2tp1gdlhBNwl2d11OuvXjzq1p6lLt3hw4d4N135SbvJexJ/JuBzkqp9kopP2A6cNHoHKVUP+BdTNI/Vml7S6WUv+3rcGA4kOqo4N3V+f796LbljNz2OofC+nEkuKu1QQnRRMQG53NVdCrvLo6ksMi7+h2VgmnTYMsWWLPGdQE2ArUmfq11GfArYDmwC5intU5RSj2nlDo/SuclIBiYf8mwze5AolJqO7AKeEFr3eQT/48/Qps2MOLol0Sc3MeaLvdZHZIQTco9XX7iZKEvs5e2rXnHiRMhLAxeeME1gTUSSrvhR6CEhASdmJhodRj1orXp4uneXfNpch98is/y4tUraHFgu9WhCdGkvHtwIsWliud/eYCHpmVXv+O//21G92zbZtbnbaKUUlts91NrJTN3HezQIcjOhqFByUQd38EP3R9Fqxo+jgoh6mVA1wLyTvuRtK+WulfTppnVuV58seb9PIiP1QE0NevWmefJ2//G8eA4kmI9Zs6aEC7V/9xP/Ow/nqWrm1HR+Vu8VA29F926mdE9PXuaOiozZ7ouUDckLX4HW7IEWgUXM/jwAlZ0e5QKL/nbKoQzeCkYHbmHzMJQvtl3Rc079+9vbvbK6lyAJH6HKi6GpUs113gvoyQski1x11sdkhBNWs+wI7TyL+DP66+mQtdQA6t5c+jRw4y8yMurfj8PIYnfgVatgoICxS2n3iNj/D1UePlaHZIQTZq30oyO3Evy8ajai7cNHWqev/rK+YG5OUn8DvT1l+U0U2cY0uYAOYOn1f4GIUSD9QrLpH3zXP6yfkLtrf6EBNi82Uyt92CS+B2kogK+mVPEJL2UzOseRXtL374QruCl4OHuq0k+HsX8vX1q3nngQAgOhieeML+0HkoSv4Ns+u4EOYXBXBm9m/xuQ60ORwiPMrndDrqHZvE/a6ZwprSGGj6+vjB8uCnX7ME1fCTxO8hXT2/Ch1I63tLf6lCE8DjeSvPHfks4XNiSv28aV/POPXpA167w5JNw6pRrAnQzkvgdQG/YyFfJHRjacjc+7WqpEy6EcIoB4RlMbZfES5vHsi+/VfU7KgWPPQY5OfDoo64L0I1I4m+owkJ23fpn0uhCwshAq6MRwqP9rvdyfL3KeXx1LUt/9OgBd94Jn37qkV0+kvgb6je/4a2Ma/H1LmfEgHNWRyOER2sdWMgjPVay5EAPvthdZQX4C37xC/MH4MEHISPDNQG6CUn8DbFoEcfe+5rZXvdxVcJJwlqUWR2REB5vRucN9A8/xMwfbqq5y8fbG37/eygthRkzzApKHkISf31lZcF99/HPsGco1r7cOvZY7e8RQjidj1cFLw+ej7eq4NYlv6C4rIYiidHRpr9/zRp44AGPWbBFEn99nDoFkydTWKD557n7Gd7zFO3aFNf+PiGES0QFneLvA79i67FYnlxXy+p3V19tWvyzZ8Mf/uCaAC0mib+uzp2D666DlBRmT5hD/ll/po+T1r4Q7mZc1G7u6vwzb2wbxRtbR9S88913w5Qp8Pe/w+uvuyQ+K8n00rooLzc3hFavpuR//8QrHw6nV4dCrog/a3VkQni0uelVrz8S1/w4V0Wn8OvV1xPoU8r9vTdWfQCl4PHHzaf5xx+Hs2fhqafM9iZIWvz2KiyEm26CBQvgoYf43/33k3EsgDuvOmp1ZEKIangrzStD5jO67R4eWHETn6QOqGFnb/jjH2H8eHPTd8YM8wm/CZLEb48DB0xlv0WL4JFHWNj2YV5bEMsNI3MZ1K3A6uiEEDXw8yrn9WFzGNL6AHd9dztPr5tMWXk1LXk/P9PP/8tfwmefwbhx5ve/iZHEXxOtTQnXQYPMmoovvMD+oXfyyxe70T3uDA9OzbI6QiGEHQK8y3hnxGfc0mEzL2wez1W/7U3OiWpq+ihlunT//GdISoIrrjB9/yUlLo3ZmSTxVyclBa66Cm64AVq0gLfeIqPdCG54picVFfDMjEP4+njG0C8hmgJ/7zKeG7CIFwZ+yfrUFnS/ayAvz43lXEk1aXD0aPjoI1PR8/e/Nwu1z5ljxv03ckrbMW5VKTUReB3wBt7XWr9wyev+wCfAACAPuFVrfdD22tPAvUA58JjWenlt50tISNCJiYl1uxJHqKiAlSvhgw9g/nwICjJ3+6dN46uf23Dvi10pLvXi2bsP2tXFs/jnC5NHQtK3OjFwIURdpIcPYs32UA7mBBLX5hxP3Z7B9HHHCA2uZhLX+vXwr3/B4cNm7P8jj8Dtt0NcnGsDr4FSaovWuuq73JfuW1viV0p5A3uBq4BMYDNwm9Y6tdI+DwO9tdYPKqWmA9drrW9VSvUAvgAGAVHACqCL1rq8pnO6NPHn55sV0tesgS+/NF06LVrAhAlU3H4n6w61Y9a3UXz+nzZ0iz3DH39xiOgI+z7ySeIXwj2d6mCq6Ea1Kua9JVHsORyEv28F14/MZeqwPEb1Pnn573lFBWzaZPLE+fzUsydcey2MHAkDBkDr1i6+kgscnfiHAn/WWl9t+/5pAK313yvts9y2z3qllA+QA0QAT1Xet/J+NZ2z3om/osJ8DCspMQvgFhWZ0ThnzsCJE5CbC8eOmboce/eid++hNP0w5/DnpG9rTnYeSFafiexqMZhdmS1YvjmMjGMBBPqXc93w49wzKadO3TuS+IVwT+cT/7XD8tAa9mYG8t2mMFZubcnps2aUe/vIInp3OEPnmLN0iiqiTVgpESEltAopI/hkJkFJPxOYuA7flCS8K0pQAFFR0K2b+SQQHw9t20JYGLRqBSEh0KyZ6UkIDDQ3kv38wMfHjChq4NDRuiR+e8bxRwOHK32fCQyubh+tdZlS6hTQyrZ9wyXvdV7d4uBgk+xrsc5vPONKllFGpTVxS4FU2wMIbV5Gz45FzLwpj9H9CggM0EANCzxUISD4wj+vr7/cThHCXZz/3Wwe7g9AQkQFCf2O81T5cdIyAtiyO4jtewPZeSiYZZvCKCm99Pd3EHDDRVu8VTne2RV4ZZejtFndS6F5gad4lH/WHpRS5g9HZmZDL69WbjOBSyk1E5hp+7ZQKbXHaScr+Q+1JfGTBfBjknnUQzhwvF7vdC9N4TqawjWAXIdj/Wye/vZhvd5d5TWUa3Mj81KP2R610hqOHGlIy9/uGw72JP4jQGyl72Ns26raJ9PW1ROCuclrz3sB0FrPAmbZF7Z7U0ol2vuRy501hetoCtcAch3upClcgz39D5uBzkqp9kopP2A6sOiSfRYBd9m+vglYqc3Ng0XAdKWUv1KqPdAZ2OSY0IUQQtRHrS1+W5/9r4DlmOGcs7XWKUqp54BErfUi4APgU6XUPuAE5o8Dtv3mYXrOy4BHahvRI4QQwrns6uPXWi8Fll6y7ZlKX58Dbq7mvX8F/tqAGBujJtFlRdO4jqZwDSDX4U4a/TXYNYFLCCFE0yFjDIUQwsNI4m8ApdREpdQepdQ+pdRTVbzur5Saa3t9o1Iq3vVR1syOa/iNUipVKZWslPqPUsp95qhXUtt1VNrvRqWUVkq55agMe65DKXWL7f8kRSn1uatjrI0dP1PtlFKrlFLbbD9Xk62IszZKqdlKqWNKqZ3VvK6UUm/YrjNZKdXf1THWm9ZaHvV4YG507wc6YCYFbAd6XLLPw8A7tq+nA3Otjrse1zAWCLJ9/ZC7XYO912HbrzmwFjOpMMHquOv5/9EZ2Aa0tH3f2uq463ENs4CHbF/3AA5aHXc11zIK6A/srOb1ycAyQAFDgI1Wx2zvQ1r89TcI2Ke1TtdalwBzgGmX7DMN+Nj29QJgvFJutaRPrdegtV6ltT6/xNgGzFwMd2PP/wXA88D/A9x1dQ17ruN+4C2tdT6A1trd1v205xo00ML2dQjglvXNtdZrMaMUqzMN+EQbG4BQpVSka6JrGEn89VdVKYtLy1FcVMoCOF/Kwl3Ycw2V3Ytp4bibWq/D9jE8Vmu9xJWB1ZE9/x9dgC5KqZ+UUhtslXPdiT3X8GfgTqVUJma04KOuCc3h6vr74zbcpmSDcG9KqTuBBGC01bHUlVLKC3gVuNviUBzBB9PdMwbz6WutUqqX1vqkpVHVzW3AR1rrV4vAzpwAAAFwSURBVGxFID9VSvXU2lbgRjidtPjrry6lLLiklIW7sKukhlLqSuAPwFStdbGLYquL2q6jOdATWK2UOojpj13khjd47fn/yAQWaa1LtdYHMCXTO7soPnvYcw33AvMAtKnUG4Cpf9PY2F2Sxt1I4q+/hpSycBe1XoNSqh/wLibpu1t/8nk1XofW+pTWOlxrHa+1jsfcq5iqtbZgtZ8a2fMz9TWmtY9SKhzT9ZPuyiBrYc81ZADjAZRS3TGJP9elUTrGImCGbXTPEOCU1jrb6qDsIV099aQbUMrCXdh5DS8BwcB8233pDK31VMuCroKd1+H27LyO5cAEpVQqphjk/2it3eZTpJ3X8FvgPaXUE5gbvXe7WYMIAKXUF5g/suG2+xHPgqnlrrV+B3N/YjKwDzgL3GNNpHUnM3eFEMLDSFePEEJ4GEn8QgjhYSTxCyGEh5HEL4QQHkYSvxBCeBhJ/EII4WEk8QshhIeRxC+EEB7m/wNCanVpDm0FmwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Benchmark PSIVER\n",
    "   \n",
    "all_pred = []\n",
    "all_labels = []\n",
    "\n",
    "all_pred_res = []\n",
    "all_labels_res = []\n",
    "\n",
    "all_roc_auc_per_prot_psiver = []\n",
    "all_res_roc_auc_per_prot_psiver = []\n",
    "\n",
    "for fn in os.listdir(psiver_pred_dir):  \n",
    "    \n",
    "    pdbid_chain = fn.split('.')[0]\n",
    "    pdbid = pdbid_chain.split('_')[0]\n",
    "    if pdbid not in transient_pdbs: \n",
    "        continue\n",
    "\n",
    "    l, p, res_l, res_p, roc_auc, res_roc_auc = compute_roc_auc_sppider_intpred_psiver(psiver_pred_dir, pdbid_chain)\n",
    "\n",
    "    all_pred.append(p)\n",
    "    all_labels.append(l)\n",
    "    \n",
    "    all_pred_res.append(res_p)\n",
    "    all_labels_res.append(res_l)\n",
    "    \n",
    "    all_roc_auc_per_prot_psiver.append(roc_auc)\n",
    "    all_res_roc_auc_per_prot_psiver.append(res_roc_auc)\n",
    "    \n",
    "print(\"Number of proteins: {}\".format(len(all_pred)))\n",
    "\n",
    "all_pred = np.concatenate(all_pred, axis=0)\n",
    "all_labels = np.concatenate(all_labels, axis=0)\n",
    "\n",
    "all_pred_by_point_psiver = all_pred.copy()\n",
    "all_labels_by_point_psiver = all_labels.copy()\n",
    "\n",
    "#print(metrics.roc_auc_score(all_my_labels_per_res, all_my_pred_per_res))\n",
    "\n",
    "all_labels_res = np.concatenate(all_labels_res, axis=0)\n",
    "\n",
    "all_pred_res = np.concatenate(all_pred_res, axis=0)\n",
    "\n",
    "\n",
    "all_roc_auc = sklearn.metrics.roc_auc_score(all_labels, all_pred)\n",
    "all_roc_auc_res = sklearn.metrics.roc_auc_score(all_labels_res, all_pred_res)\n",
    "\n",
    "\n",
    "\n",
    "print('psiver Global ROC AUC: {:.2f}'.format(all_roc_auc))\n",
    "print('psiver Global ROC AUC per residue: {:.2f}'.format(all_roc_auc_res))\n",
    "\n",
    "print('psiver Median ROC AUC per protein : {:.2f}'.format(np.median(all_roc_auc_per_prot_psiver)))\n",
    "print('psiver Median ROC AUC per residue per protein: {:.2f}'.format(np.median(all_res_roc_auc_per_prot_psiver)))\n",
    "\n",
    "pos_label = np.where(all_labels_res == 1)[0]\n",
    "pos = all_pred_res[pos_label]\n",
    "neg_label = np.where(all_labels_res == 0)[0]\n",
    "neg = all_pred_res[neg_label]\n",
    "sns.distplot(pos, hist=True, color='r', kde_kws={\"shade\": True}, norm_hist=True)\n",
    "sns.distplot(neg, hist=True, color='b', kde_kws={\"shade\": True}, norm_hist=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1XQS\n",
      "Per residue ROC AUC: 0.58\n",
      "sppider_pred/ Per point ROC AUC 1XQS_A : 0.52\n",
      "4XL5\n",
      "Per residue ROC AUC: 0.74\n",
      "sppider_pred/ Per point ROC AUC 4XL5_C : 0.75\n",
      "4HDO\n",
      "Per residue ROC AUC: 0.74\n",
      "sppider_pred/ Per point ROC AUC 4HDO_A : 0.70\n",
      "3BH6\n",
      "Per residue ROC AUC: 0.69\n",
      "sppider_pred/ Per point ROC AUC 3BH6_B : 0.63\n",
      "4Y61\n",
      "Per residue ROC AUC: 0.66\n",
      "sppider_pred/ Per point ROC AUC 4Y61_B : 0.52\n",
      "5TIH\n",
      "Per residue ROC AUC: 0.39\n",
      "sppider_pred/ Per point ROC AUC 5TIH_A : 0.30\n",
      "3RT0\n",
      "Per residue ROC AUC: 0.58\n",
      "sppider_pred/ Per point ROC AUC 3RT0_A : 0.57\n",
      "3MCA\n",
      "Per residue ROC AUC: 0.73\n",
      "sppider_pred/ Per point ROC AUC 3MCA_B : 0.73\n",
      "4X33\n",
      "Per residue ROC AUC: 0.64\n",
      "sppider_pred/ Per point ROC AUC 4X33_B : 0.57\n",
      "3O5T\n",
      "Per residue ROC AUC: 0.87\n",
      "sppider_pred/ Per point ROC AUC 3O5T_A : 0.87\n",
      "4ZGY\n",
      "Per residue ROC AUC: 0.83\n",
      "sppider_pred/ Per point ROC AUC 4ZGY_A : 0.82\n",
      "4ZRJ\n",
      "Per residue ROC AUC: 0.65\n",
      "sppider_pred/ Per point ROC AUC 4ZRJ_A : 0.64\n",
      "1XG2\n",
      "Per residue ROC AUC: 0.45\n",
      "sppider_pred/ Per point ROC AUC 1XG2_A : 0.34\n",
      "2MTA\n",
      "4DVG\n",
      "Per residue ROC AUC: 0.55\n",
      "sppider_pred/ Per point ROC AUC 4DVG_B : 0.56\n",
      "3VYR\n",
      "Per residue ROC AUC: 0.77\n",
      "sppider_pred/ Per point ROC AUC 3VYR_B : 0.74\n",
      "2WUS\n",
      "Per residue ROC AUC: 0.44\n",
      "sppider_pred/ Per point ROC AUC 2WUS_A : 0.40\n",
      "3QML\n",
      "Per residue ROC AUC: 0.59\n",
      "sppider_pred/ Per point ROC AUC 3QML_D : 0.66\n",
      "2V3B\n",
      "Per residue ROC AUC: 0.78\n",
      "sppider_pred/ Per point ROC AUC 2V3B_A : 0.70\n",
      "4FZV\n",
      "Per residue ROC AUC: 0.69\n",
      "sppider_pred/ Per point ROC AUC 4FZV_A : 0.63\n",
      "3VV2\n",
      "Per residue ROC AUC: 0.66\n",
      "sppider_pred/ Per point ROC AUC 3VV2_A : 0.67\n",
      "2OOR\n",
      "1F6M\n",
      "Per residue ROC AUC: 0.76\n",
      "sppider_pred/ Per point ROC AUC 1F6M_A : 0.76\n",
      "1ZHH\n",
      "Per residue ROC AUC: 0.82\n",
      "sppider_pred/ Per point ROC AUC 1ZHH_A : 0.79\n",
      "4YEB\n",
      "Per residue ROC AUC: 0.50\n",
      "sppider_pred/ Per point ROC AUC 4YEB_A : 0.45\n",
      "2V9T\n",
      "Per residue ROC AUC: 0.42\n",
      "sppider_pred/ Per point ROC AUC 2V9T_B : 0.30\n",
      "1EWY\n",
      "Per residue ROC AUC: 0.70\n",
      "sppider_pred/ Per point ROC AUC 1EWY_A : 0.64\n",
      "1NCA\n",
      "Per residue ROC AUC: 0.47\n",
      "sppider_pred/ Per point ROC AUC 1NCA_N : 0.38\n",
      "2Y5B\n",
      "Per residue ROC AUC: 0.53\n",
      "sppider_pred/ Per point ROC AUC 2Y5B_A : 0.42\n",
      "1JTD\n",
      "Per residue ROC AUC: 0.74\n",
      "sppider_pred/ Per point ROC AUC 1JTD_B : 0.70\n",
      "4ETP\n",
      "Per residue ROC AUC: 0.85\n",
      "sppider_pred/ Per point ROC AUC 4ETP_A : 0.86\n",
      "3FHC\n",
      "Per residue ROC AUC: 0.40\n",
      "sppider_pred/ Per point ROC AUC 3FHC_A : 0.31\n",
      "1WDW\n",
      "Per residue ROC AUC: 0.88\n",
      "sppider_pred/ Per point ROC AUC 1WDW_H : 0.82\n",
      "3H6G\n",
      "Per residue ROC AUC: 0.74\n",
      "sppider_pred/ Per point ROC AUC 3H6G_B : 0.71\n",
      "4V0O\n",
      "Per residue ROC AUC: 0.53\n",
      "sppider_pred/ Per point ROC AUC 4V0O_F : 0.45\n",
      "4C9B\n",
      "Per residue ROC AUC: 0.70\n",
      "sppider_pred/ Per point ROC AUC 4C9B_B : 0.71\n",
      "4BKX\n",
      "Per residue ROC AUC: 0.65\n",
      "sppider_pred/ Per point ROC AUC 4BKX_B : 0.65\n",
      "2QXV\n",
      "Per residue ROC AUC: 0.59\n",
      "sppider_pred/ Per point ROC AUC 2QXV_A : 0.51\n",
      "2AYO\n",
      "Per residue ROC AUC: 0.63\n",
      "sppider_pred/ Per point ROC AUC 2AYO_A : 0.54\n",
      "1I2M\n",
      "Per residue ROC AUC: 0.60\n",
      "sppider_pred/ Per point ROC AUC 1I2M_B : 0.51\n",
      "3SJA\n",
      "Per residue ROC AUC: 0.46\n",
      "sppider_pred/ Per point ROC AUC 3SJA_I : 0.34\n",
      "2XWT\n",
      "Per residue ROC AUC: 0.56\n",
      "sppider_pred/ Per point ROC AUC 2XWT_C : 0.45\n",
      "3WN7\n",
      "Per residue ROC AUC: 0.40\n",
      "sppider_pred/ Per point ROC AUC 3WN7_A : 0.31\n",
      "2I3T\n",
      "Per residue ROC AUC: 0.80\n",
      "sppider_pred/ Per point ROC AUC 2I3T_A : 0.70\n",
      "5J57\n",
      "Per residue ROC AUC: 0.57\n",
      "sppider_pred/ Per point ROC AUC 5J57_A : 0.42\n",
      "3NVN\n",
      "Per residue ROC AUC: 0.65\n",
      "sppider_pred/ Per point ROC AUC 3NVN_A : 0.69\n",
      "1GP2\n",
      "4M0W\n",
      "Per residue ROC AUC: 0.66\n",
      "sppider_pred/ Per point ROC AUC 4M0W_A : 0.58\n",
      "4YC7\n",
      "Per residue ROC AUC: 0.65\n",
      "sppider_pred/ Per point ROC AUC 4YC7_B : 0.64\n",
      "4ETP\n",
      "Per residue ROC AUC: 0.93\n",
      "sppider_pred/ Per point ROC AUC 4ETP_B : 0.89\n",
      "4LVN\n",
      "Per residue ROC AUC: 0.71\n",
      "sppider_pred/ Per point ROC AUC 4LVN_A : 0.74\n",
      "3LVK\n",
      "3ZWL\n",
      "Per residue ROC AUC: 0.64\n",
      "sppider_pred/ Per point ROC AUC 3ZWL_B : 0.55\n",
      "3R9A\n",
      "1W1W\n",
      "Per residue ROC AUC: 0.82\n",
      "sppider_pred/ Per point ROC AUC 1W1W_B : 0.81\n",
      "2F4M\n",
      "Per residue ROC AUC: 0.81\n",
      "sppider_pred/ Per point ROC AUC 2F4M_A : 0.84\n",
      "2C0L\n",
      "Per residue ROC AUC: 0.49\n",
      "sppider_pred/ Per point ROC AUC 2C0L_A : 0.48\n",
      "3BP8\n",
      "2PMS\n",
      "Per residue ROC AUC: 0.69\n",
      "sppider_pred/ Per point ROC AUC 2PMS_A : 0.61\n",
      "Number of proteins: 53\n",
      "Global ROC AUC: 0.63\n",
      "Global ROC AUC per residue: 0.66\n",
      "Median ROC AUC per protein : 0.63\n",
      "Median ROC AUC per residue per protein: 0.65\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x1174c5a50>"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Benchmark SPPIDER\n",
    "   \n",
    "all_pred = []\n",
    "all_labels = []\n",
    "\n",
    "all_pred_res = []\n",
    "all_labels_res = []\n",
    "\n",
    "all_roc_auc_per_prot_sppider = []\n",
    "all_res_roc_auc_per_prot_sppider = []\n",
    "\n",
    "for fn in os.listdir(sppider_pred_dir):  \n",
    "    pdbid_chain = fn.split('.')[0]\n",
    "    pdbid = pdbid_chain.split('_')[0]\n",
    "    print(pdbid)\n",
    "    \n",
    "    if pdbid not in transient_pdbs: \n",
    "        continue\n",
    "\n",
    "    l, p, res_l, res_p, roc_auc, res_roc_auc = compute_roc_auc_sppider_intpred_psiver(sppider_pred_dir, pdbid_chain)\n",
    "\n",
    "    all_pred.append(p)\n",
    "    all_labels.append(l)\n",
    "    \n",
    "    all_pred_res.append(res_p)\n",
    "    all_labels_res.append(res_l)\n",
    "    \n",
    "    all_roc_auc_per_prot_sppider.append(roc_auc)\n",
    "    all_res_roc_auc_per_prot_sppider.append(res_roc_auc)\n",
    "    \n",
    "print(\"Number of proteins: {}\".format(len(all_pred)))\n",
    "\n",
    "all_pred = np.concatenate(all_pred, axis=0)\n",
    "all_labels = np.concatenate(all_labels, axis=0)\n",
    "\n",
    "all_pred_by_point_sppider = all_pred.copy()\n",
    "all_labels_by_point_sppider = all_labels.copy()\n",
    "\n",
    "#print(metrics.roc_auc_score(all_my_labels_per_res, all_my_pred_per_res))\n",
    "\n",
    "all_labels_res = np.concatenate(all_labels_res, axis=0)\n",
    "\n",
    "all_pred_res = np.concatenate(all_pred_res, axis=0)\n",
    "\n",
    "\n",
    "all_roc_auc = sklearn.metrics.roc_auc_score(all_labels, all_pred)\n",
    "all_roc_auc_res = sklearn.metrics.roc_auc_score(all_labels_res, all_pred_res)\n",
    "\n",
    "\n",
    "\n",
    "print('Global ROC AUC: {:.2f}'.format(all_roc_auc))\n",
    "print('Global ROC AUC per residue: {:.2f}'.format(all_roc_auc_res))\n",
    "\n",
    "print('Median ROC AUC per protein : {:.2f}'.format(np.median(all_roc_auc_per_prot_sppider)))\n",
    "print('Median ROC AUC per residue per protein: {:.2f}'.format(np.median(all_res_roc_auc_per_prot_sppider)))\n",
    "\n",
    "pos_label = np.where(all_labels_res == 1)[0]\n",
    "pos = all_pred_res[pos_label]\n",
    "neg_label = np.where(all_labels_res == 0)[0]\n",
    "neg = all_pred_res[neg_label]\n",
    "sns.distplot(pos, hist=True, color='r', kde_kws={\"shade\": True}, norm_hist=True)\n",
    "sns.distplot(neg, hist=True, color='b', kde_kws={\"shade\": True}, norm_hist=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compute the MASIF predictions per residue.\n",
    "# Params: myid: PDBID_CHAIN format\n",
    "# scores: the predicted scores for this one. \n",
    "def compute_pred_labels_per_residue(myid, scores):\n",
    "    # We use the same pdb that sppider predicted. \n",
    "    parser = PDBParser()\n",
    "    # Use the same pdb that sppider used. \n",
    "    struct = parser.get_structure(sppider_pred_dir+myid+'.pdb',sppider_pred_dir+myid+'.pdb')\n",
    "    atoms = Selection.unfold_entities(struct, 'A')\n",
    "    residues = Selection.unfold_entities(struct, 'R')\n",
    "    \n",
    "    ground_truth_mesh = pymesh.load_mesh(ground_truth_dir+myid+'.ply')\n",
    "    \n",
    "    res_labels = assign_labels_per_residue(atoms, residues, ground_truth_mesh)\n",
    "    \n",
    "    # Now assign predictions to each vert    \n",
    "    verts = ground_truth_mesh.vertices\n",
    "    ground_truth_mesh.add_attribute(\"vertex_area\")\n",
    "    vertex_areas = ground_truth_mesh.get_attribute(\"vertex_area\")\n",
    "    \n",
    "    at_coord = np.array([at.get_coord() for at in atoms]).astype(float)\n",
    "    verts = ground_truth_mesh.vertices.astype(float)\n",
    "    \n",
    "    \n",
    "    residues = [x for x in verts]\n",
    "    flann = pyflann.FLANN()\n",
    "    r,d = flann.nn(at_coord, verts)\n",
    "    d = np.sqrt(d)\n",
    "    n= len(r)\n",
    "    \n",
    "    # For each residue, assign the mean score. \n",
    "    res_scores = {}\n",
    "    res_areas = {}\n",
    "    for ix, v in enumerate(verts): \n",
    "        pred = scores[ix]\n",
    "        res_id = atoms[r[ix]].get_parent().get_id()\n",
    "        if res_id not in res_scores:\n",
    "            res_scores[res_id] = []\n",
    "            res_areas[res_id] = []\n",
    "        res_scores[res_id].append(pred)\n",
    "        res_areas[res_id].append(pred)\n",
    "    print('Total number of residues {}'.format(len(res_scores.keys())))\n",
    "\n",
    "    for key in res_scores:\n",
    "        res_scores[key] =  np.max(res_scores[key])  \n",
    "\n",
    "    pred_per_res = []\n",
    "    labels_per_res = []\n",
    "    for key in res_labels: \n",
    "        if key in res_scores and key in res_labels:\n",
    "            pred_per_res.append(res_scores[key])\n",
    "            labels_per_res.append(res_labels[key])\n",
    "\n",
    "    \n",
    "    return pred_per_res, labels_per_res "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compute per residue and per point ROC AUC scores for MASIF> \n",
    "def compute_masif_scores():\n",
    "    all_my_scores = []\n",
    "    all_my_labels = []\n",
    "    all_my_auc_per_protein = {}\n",
    "    all_my_auc_per_res = []\n",
    "    all_my_names = []\n",
    "    \n",
    "    all_my_pred_per_res = []\n",
    "    all_my_labels_per_res = []\n",
    "    count = 0\n",
    "\n",
    "    for name in os.listdir(masif_pred_dir): \n",
    "\n",
    "        myid = name.split('.')[0]\n",
    "        pdbid = myid.split('_')[0]\n",
    "        print(pdbid)\n",
    "    \n",
    "        if pdbid not in transient_pdbs: \n",
    "            continue\n",
    "        # Load interface for file. \n",
    "        mesh = pymesh.load_mesh(masif_pred_dir+'/'+name)\n",
    "        masif_pred = mesh.get_attribute('vertex_iface')\n",
    "        all_my_scores.append(masif_pred)\n",
    "        \n",
    "        # Load groundtruth for file. \n",
    "        mesh = pymesh.load_mesh(ground_truth_dir+'/'+name)\n",
    "        groundtruth = mesh.get_attribute('vertex_iface')\n",
    "        all_my_labels.append(groundtruth)\n",
    "        \n",
    "        auc = metrics.roc_auc_score(groundtruth, masif_pred)\n",
    "        print (auc)\n",
    "        all_my_auc_per_protein[myid] = auc\n",
    "\n",
    "        \n",
    "        count = count+1\n",
    "            \n",
    "        print('{} ROC AUC: {}'.format(myid, auc))\n",
    "            \n",
    "#        # Compute the prediction per residue.\n",
    "        pred_per_res, labels_per_res = compute_pred_labels_per_residue(myid, masif_pred)\n",
    "        auc_per_res = metrics.roc_auc_score(labels_per_res, pred_per_res)\n",
    "        all_my_auc_per_res.append(auc_per_res)\n",
    "\n",
    "\n",
    "        all_my_pred_per_res.append(pred_per_res)\n",
    "        all_my_labels_per_res.append(labels_per_res)\n",
    "        print('{} ROC AUC per res: {}'.format(myid, auc_per_res))\n",
    "\n",
    "    all_my_scores = np.concatenate(all_my_scores, axis=0)\n",
    "    all_my_labels = np.concatenate(all_my_labels, axis=0)\n",
    "    print('Number of proteins: {}'.format(count))\n",
    "    print('Median AUC per res: {}'.format(np.median(all_my_auc_per_res)))\n",
    "            \n",
    "    return all_my_scores, all_my_labels, all_my_auc_per_protein, \\\n",
    "                    all_my_names, \\\n",
    "                    all_my_labels_per_res, all_my_pred_per_res, \\\n",
    "                    all_my_auc_per_res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4YEB\n",
      "0.30842007374672264\n",
      "4YEB_A ROC AUC: 0.308420073747\n",
      "Total number of residues 205\n",
      "4YEB_A ROC AUC per res: 0.326315789474\n",
      "3BH6\n",
      "0.8637100744471168\n",
      "3BH6_B ROC AUC: 0.863710074447\n",
      "Total number of residues 266\n",
      "3BH6_B ROC AUC per res: 0.851736111111\n",
      "2QXV\n",
      "0.8590961176367115\n",
      "2QXV_A ROC AUC: 0.859096117637\n",
      "Total number of residues 284\n",
      "2QXV_A ROC AUC per res: 0.875518341308\n",
      "4ETP\n",
      "0.9569008182323422\n",
      "4ETP_A ROC AUC: 0.956900818232\n",
      "Total number of residues 326\n",
      "4ETP_A ROC AUC per res: 0.958293384468\n",
      "1XG2\n",
      "0.7721965497926699\n",
      "1XG2_A ROC AUC: 0.772196549793\n",
      "Total number of residues 244\n",
      "1XG2_A ROC AUC per res: 0.748348017621\n",
      "2AYO\n",
      "0.871198177086431\n",
      "2AYO_A ROC AUC: 0.871198177086\n",
      "Total number of residues 310\n",
      "2AYO_A ROC AUC per res: 0.807037037037\n",
      "3VV2\n",
      "0.7720700157350882\n",
      "3VV2_A ROC AUC: 0.772070015735\n",
      "Total number of residues 228\n",
      "3VV2_A ROC AUC per res: 0.694776486\n",
      "1W1W\n",
      "0.8232751674152109\n",
      "1W1W_B ROC AUC: 0.823275167415\n",
      "Total number of residues 283\n",
      "1W1W_B ROC AUC per res: 0.822968262704\n",
      "1I2M\n",
      "0.705224134429956\n",
      "1I2M_B ROC AUC: 0.70522413443\n",
      "Total number of residues 313\n",
      "1I2M_B ROC AUC per res: 0.679513744992\n",
      "1XQS\n",
      "0.8668554287697527\n",
      "1XQS_A ROC AUC: 0.86685542877\n",
      "Total number of residues 217\n",
      "1XQS_A ROC AUC per res: 0.865039281706\n",
      "2Y5B\n",
      "0.8242885023855366\n",
      "2Y5B_A ROC AUC: 0.824288502386\n",
      "Total number of residues 270\n",
      "2Y5B_A ROC AUC per res: 0.813438438438\n",
      "5TIH\n",
      "0.25242319246692924\n",
      "5TIH_A ROC AUC: 0.252423192467\n",
      "Total number of residues 281\n",
      "5TIH_A ROC AUC per res: 0.234905660377\n",
      "1EWY\n",
      "0.7542497591856523\n",
      "1EWY_A ROC AUC: 0.754249759186\n",
      "Total number of residues 260\n",
      "1EWY_A ROC AUC per res: 0.831884057971\n",
      "3QML\n",
      "0.656069492653465\n",
      "3QML_D ROC AUC: 0.656069492653\n",
      "Total number of residues 234\n",
      "3QML_D ROC AUC per res: 0.60873015873\n",
      "4X33\n",
      "0.9098260710844205\n",
      "4X33_B ROC AUC: 0.909826071084\n",
      "Total number of residues 286\n",
      "4X33_B ROC AUC per res: 0.928641829292\n",
      "1WDW\n",
      "0.8077133203299501\n",
      "1WDW_H ROC AUC: 0.80771332033\n",
      "Total number of residues 302\n",
      "1WDW_H ROC AUC per res: 0.844311749705\n",
      "3LVK\n",
      "1NCA\n",
      "0.6073050323050323\n",
      "1NCA_N ROC AUC: 0.607305032305\n",
      "Total number of residues 289\n",
      "1NCA_N ROC AUC per res: 0.560699588477\n",
      "2V9T\n",
      "0.9448898107337309\n",
      "2V9T_B ROC AUC: 0.944889810734\n",
      "Total number of residues 179\n",
      "2V9T_B ROC AUC per res: 0.91975308642\n",
      "1F6M\n",
      "0.8256860771353657\n",
      "1F6M_A ROC AUC: 0.825686077135\n",
      "Total number of residues 280\n",
      "1F6M_A ROC AUC per res: 0.826040781648\n",
      "3O5T\n",
      "0.8389708668613335\n",
      "3O5T_A ROC AUC: 0.838970866861\n",
      "Total number of residues 220\n",
      "3O5T_A ROC AUC per res: 0.780339277947\n",
      "4C9B\n",
      "0.532132486967701\n",
      "4C9B_B ROC AUC: 0.532132486968\n",
      "Total number of residues 232\n",
      "4C9B_B ROC AUC per res: 0.459656871072\n",
      "3VYR\n",
      "0.891829367647864\n",
      "3VYR_B ROC AUC: 0.891829367648\n",
      "Total number of residues 297\n",
      "3VYR_B ROC AUC per res: 0.909852941176\n",
      "2WUS\n",
      "0.8458786638253328\n",
      "2WUS_A ROC AUC: 0.845878663825\n",
      "Total number of residues 274\n",
      "2WUS_A ROC AUC per res: 0.846055979644\n",
      "2F4M\n",
      "0.9363165766588748\n",
      "2F4M_A ROC AUC: 0.936316576659\n",
      "Total number of residues 259\n",
      "2F4M_A ROC AUC per res: 0.960932944606\n",
      "3BP8\n",
      "2I3T\n",
      "0.8429039439739575\n",
      "2I3T_A ROC AUC: 0.842903943974\n",
      "Total number of residues 283\n",
      "2I3T_A ROC AUC per res: 0.780950626382\n",
      "4ZRJ\n",
      "0.9290031805481243\n",
      "4ZRJ_A ROC AUC: 0.929003180548\n",
      "Total number of residues 261\n",
      "4ZRJ_A ROC AUC per res: 0.916503587736\n",
      "4XL5\n",
      "0.9480246716989467\n",
      "4XL5_C ROC AUC: 0.948024671699\n",
      "Total number of residues 207\n",
      "4XL5_C ROC AUC per res: 0.929311621021\n",
      "2C0L\n",
      "0.6976427210560338\n",
      "2C0L_A ROC AUC: 0.697642721056\n",
      "Total number of residues 250\n",
      "2C0L_A ROC AUC per res: 0.626481481481\n",
      "3FHC\n",
      "0.41368560905796037\n",
      "3FHC_A ROC AUC: 0.413685609058\n",
      "Total number of residues 327\n",
      "3FHC_A ROC AUC per res: 0.403692093347\n",
      "2PMS\n",
      "0.8302527402669955\n",
      "2PMS_A ROC AUC: 0.830252740267\n",
      "Total number of residues 279\n",
      "2PMS_A ROC AUC per res: 0.79923574368\n",
      "4YC7\n",
      "0.743368099170046\n",
      "4YC7_B ROC AUC: 0.74336809917\n",
      "Total number of residues 243\n",
      "4YC7_B ROC AUC per res: 0.699178332319\n",
      "4BKX\n",
      "0.7848349612058485\n",
      "4BKX_B ROC AUC: 0.784834961206\n",
      "Total number of residues 266\n",
      "4BKX_B ROC AUC per res: 0.752865144334\n",
      "3SJA\n",
      "0.23343990330419095\n",
      "3SJA_I ROC AUC: 0.233439903304\n",
      "Total number of residues 311\n",
      "3SJA_I ROC AUC per res: 0.195723684211\n",
      "4LVN\n",
      "0.8935504560164754\n",
      "4LVN_A ROC AUC: 0.893550456016\n",
      "Total number of residues 244\n",
      "4LVN_A ROC AUC per res: 0.922895622896\n",
      "3NVN\n",
      "0.865961738949481\n",
      "3NVN_A ROC AUC: 0.865961738949\n",
      "Total number of residues 340\n",
      "3NVN_A ROC AUC per res: 0.812056091787\n",
      "4Y61\n",
      "0.7199084334989589\n",
      "4Y61_B ROC AUC: 0.719908433499\n",
      "Total number of residues 204\n",
      "4Y61_B ROC AUC per res: 0.673482227115\n",
      "1JTD\n",
      "0.9281327325620278\n",
      "1JTD_B ROC AUC: 0.928132732562\n",
      "Total number of residues 237\n",
      "1JTD_B ROC AUC per res: 0.970194003527\n",
      "3H6G\n",
      "0.9412925799563507\n",
      "3H6G_B ROC AUC: 0.941292579956\n",
      "Total number of residues 306\n",
      "3H6G_B ROC AUC per res: 0.964972527473\n",
      "3WN7\n",
      "0.7291930553444462\n",
      "3WN7_A ROC AUC: 0.729193055344\n",
      "Total number of residues 247\n",
      "3WN7_A ROC AUC per res: 0.618586789555\n",
      "3RT0\n",
      "0.63753981838064\n",
      "3RT0_A ROC AUC: 0.637539818381\n",
      "Total number of residues 263\n",
      "3RT0_A ROC AUC per res: 0.559108527132\n",
      "4DVG\n",
      "0.762499970064971\n",
      "4DVG_B ROC AUC: 0.762499970065\n",
      "Total number of residues 275\n",
      "4DVG_B ROC AUC per res: 0.704278812974\n",
      "2V3B\n",
      "0.9304992977868178\n",
      "2V3B_A ROC AUC: 0.930499297787\n",
      "Total number of residues 335\n",
      "2V3B_A ROC AUC per res: 0.979890933879\n",
      "1ZHH\n",
      "0.7383194850112892\n",
      "1ZHH_A ROC AUC: 0.738319485011\n",
      "Total number of residues 286\n",
      "1ZHH_A ROC AUC per res: 0.654366543665\n",
      "2MTA\n",
      "2OOR\n",
      "4ZGY\n",
      "0.8519036885346535\n",
      "4ZGY_A ROC AUC: 0.851903688535\n",
      "Total number of residues 328\n",
      "4ZGY_A ROC AUC per res: 0.803563791875\n",
      "4M0W\n",
      "0.8937126928750877\n",
      "4M0W_A ROC AUC: 0.893712692875\n",
      "Total number of residues 271\n",
      "4M0W_A ROC AUC per res: 0.867286115007\n",
      "3ZWL\n",
      "0.7810445204717154\n",
      "3ZWL_B ROC AUC: 0.781044520472\n",
      "Total number of residues 291\n",
      "3ZWL_B ROC AUC per res: 0.742958673335\n",
      "5J57\n",
      "0.29827168476651966\n",
      "5J57_A ROC AUC: 0.298271684767\n",
      "Total number of residues 225\n",
      "5J57_A ROC AUC per res: 0.325899843505\n",
      "4HDO\n",
      "0.7855665998869883\n",
      "4HDO_A ROC AUC: 0.785566599887\n",
      "Total number of residues 282\n",
      "4HDO_A ROC AUC per res: 0.758088235294\n",
      "4V0O\n",
      "0.8509997517349461\n",
      "4V0O_F ROC AUC: 0.850999751735\n",
      "Total number of residues 270\n",
      "4V0O_F ROC AUC per res: 0.879948680352\n",
      "1GP2\n",
      "4FZV\n",
      "0.8661856852398598\n",
      "4FZV_A ROC AUC: 0.86618568524\n",
      "Total number of residues 300\n",
      "4FZV_A ROC AUC per res: 0.829930698633\n",
      "4ETP\n",
      "0.9325052852950115\n",
      "4ETP_B ROC AUC: 0.932505285295\n",
      "Total number of residues 253\n",
      "4ETP_B ROC AUC per res: 0.926126736699\n",
      "3R9A\n",
      "2XWT\n",
      "0.9399650020191143\n",
      "2XWT_C ROC AUC: 0.939965002019\n",
      "Total number of residues 193\n",
      "2XWT_C ROC AUC per res: 0.946103896104\n",
      "3MCA\n",
      "0.8610093830060466\n",
      "3MCA_B ROC AUC: 0.861009383006\n",
      "Total number of residues 306\n",
      "3MCA_B ROC AUC per res: 0.760416666667\n",
      "Number of proteins: 53\n",
      "Median AUC per res: 0.807037037037\n",
      "ROC AUC for all points : 0.787\n",
      "ROC AUC for all residues: 0.760\n",
      "Median AUC per protein at the residue level: 0.807\n",
      "0.8302527402669955\n",
      "Median AUC per protein at the point level: 0.830\n"
     ]
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "import os\n",
    "\n",
    "all_scores_masif, all_labels_masif, all_auc_per_protein_masif, \\\n",
    "    all_names_masif, \\\n",
    "    all_my_labels_per_res_masif, all_my_pred_per_res_masif, \\\n",
    "    all_auc_per_protein_per_res_masif = compute_masif_scores()\n",
    "\n",
    "all_my_labels_per_res_masif = np.concatenate(all_my_labels_per_res_masif, axis = 0)\n",
    "all_my_pred_per_res_masif = np.concatenate(all_my_pred_per_res_masif, axis = 0)    \n",
    "\n",
    "print('ROC AUC for all points : {:.3f}'.format(metrics.roc_auc_score(all_labels_masif, all_scores_masif)))\n",
    "print('ROC AUC for all residues: {:.3f}'.format(metrics.roc_auc_score(all_my_labels_per_res_masif, all_my_pred_per_res_masif)))\n",
    "print('Median AUC per protein at the residue level: {:.3f}'.format(np.median(all_auc_per_protein_per_res_masif)))\n",
    "\n",
    "all_auc_per_protein_masif_list = [all_auc_per_protein_masif[x] for x in all_auc_per_protein_masif.keys() ]\n",
    "print(np.median(all_auc_per_protein_masif_list))\n",
    "\n",
    "print('Median AUC per protein at the point level: {:.3f}'.format(np.median(all_auc_per_protein_masif_list)))\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.807037037037037, 0.6541125541125542, 0.6038418079096045, 0.4895926851165151]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "sns.set(style=\"white\", context=\"talk\")\n",
    "\n",
    "xdata = ['MaSIF-site', 'SPPIDER', 'PSIVER', 'IntPred']\n",
    "ydata = [np.median(all_auc_per_protein_per_res_masif), \\\n",
    "        np.median(all_res_roc_auc_per_prot_sppider), \\\n",
    "        np.median(all_res_roc_auc_per_prot_psiver), \\\n",
    "        np.median(all_res_roc_auc_per_prot_intpred) ]\n",
    "\n",
    "f, (ax1) = plt.subplots(1, 1, figsize=(8, 4), sharex=False)\n",
    "\n",
    "plt.ylim([0.5,1.0])\n",
    "\n",
    "sns.barplot(x=xdata, y=ydata, ax=ax1, palette=['lightgray', 'blue', 'lightgreen', 'pink'])\n",
    "\n",
    "ax1.set_ylim([0,1])\n",
    "plt.text(-0.2,ydata[0],'{:.2f}'.format(ydata[0]))\n",
    "plt.text(0.8,ydata[1],'{:.2f}'.format(ydata[1]))\n",
    "plt.text(1.8,ydata[2],'{:.2f}'.format(ydata[2]))\n",
    "plt.text(2.8,ydata[3],'{:.2f}'.format(ydata[3]))\n",
    "#ax1.set_xticks([])\n",
    "\n",
    "ax1.spines['right'].set_visible(False)\n",
    "ax1.spines['top'].set_visible(False)\n",
    "#axes[0].spines['right'].set_visible(False)\n",
    "#axes[0].spines['top'].set_visible(False)\n",
    "\n",
    "\n",
    "print(ydata)\n",
    "\n",
    "plt.savefig('sppider_bars_w_intpred.pdf', type='pdf')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.807037037037037, 0.6541125541125542, 0.6038418079096045]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "sns.set(style=\"white\", context=\"talk\")\n",
    "\n",
    "xdata = ['MaSIF-site', 'SPPIDER', 'PSIVER']\n",
    "ydata = [np.median(all_auc_per_protein_per_res_masif), \\\n",
    "        np.median(all_res_roc_auc_per_prot_sppider), \\\n",
    "        np.median(all_res_roc_auc_per_prot_psiver) ]\n",
    "\n",
    "f, (ax1) = plt.subplots(1, 1, figsize=(8, 4), sharex=False)\n",
    "\n",
    "plt.ylim([0.5,1.0])\n",
    "\n",
    "sns.barplot(x=xdata, y=ydata, ax=ax1, palette=['lightgray', 'blue', 'lightgreen'])\n",
    "\n",
    "ax1.set_ylim([0,1])\n",
    "#plt.text(-0.2,ydata[0],'{:.2f}'.format(ydata[0]))\n",
    "#plt.text(0.8,ydata[1],'{:.2f}'.format(ydata[1]))\n",
    "#plt.text(1.8,ydata[2],'{:.2f}'.format(ydata[2]))\n",
    "ax1.set_xticks([])\n",
    "\n",
    "ax1.spines['right'].set_visible(False)\n",
    "ax1.spines['top'].set_visible(False)\n",
    "#axes[0].spines['right'].set_visible(False)\n",
    "#axes[0].spines['top'].set_visible(False)\n",
    "\n",
    "\n",
    "print(ydata)\n",
    "\n",
    "plt.savefig('sppider_psiver_bars.pdf', type='pdf')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "f, (ax1) = plt.subplots(1, 1, figsize=(8, 4), sharex=False)\n",
    "sns.distplot(all_auc_per_protein_per_res_masif, color='gray')\n",
    "sns.distplot(all_res_roc_auc_per_prot_sppider, color='blue' )\n",
    "sns.distplot(all_res_roc_auc_per_prot_psiver, color='green')\n",
    "ax1.spines['right'].set_visible(False)\n",
    "ax1.spines['top'].set_visible(False)\n",
    "ax1.legend(['MaSIF-site', 'SPPIDER', 'PSIVER'])\n",
    "ax1.set_xlabel('ROC AUC per protein (residue level)')\n",
    "ax1.set_ylabel('Frequency')\n",
    "plt.savefig('sppider_psiver_histo.pdf', type='pdf')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.760292729678894\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot per residue for MsSIF\n",
    "import seaborn as sns\n",
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline\n",
    "#sns.set(style=\"white\", context=\"talk\")\n",
    "\n",
    "\n",
    "pos_label = np.where(all_my_labels_per_res_masif == 1)[0]\n",
    "pos = all_my_pred_per_res_masif[pos_label]\n",
    "neg_label = np.where(all_my_labels_per_res_masif == 0)[0]\n",
    "neg = all_my_pred_per_res_masif[neg_label]\n",
    "\n",
    "\n",
    "f, ax = plt.subplots(1, 1, figsize=(8, 4), sharex=False)\n",
    "sns.distplot(pos, hist=True, color='r', kde_kws={\"shade\": True},ax=ax,  norm_hist=False)\n",
    "sns.distplot(neg, hist=True, color='b', kde_kws={\"shade\": True},ax=ax, norm_hist=False)\n",
    "plt.xlim([0,1])\n",
    "ax.set_xticks([])\n",
    "ax.set_yticks([])\n",
    "ax.spines['right'].set_visible(False)\n",
    "ax.spines['top'].set_visible(False)\n",
    "print(metrics.roc_auc_score(all_my_labels_per_res_masif, all_my_pred_per_res_masif))\n",
    "#plt.savefig('dist_plot_ours.pdf', type='pdf')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7866842622794995\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot per point for MsSIF\n",
    "import seaborn as sns\n",
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline\n",
    "#sns.set(style=\"white\", context=\"talk\")\n",
    "\n",
    "\n",
    "pos_label = np.where(all_labels_masif == 1)[0]\n",
    "pos = all_scores_masif[pos_label]\n",
    "neg_label = np.where(all_labels_masif == 0)[0]\n",
    "neg = all_scores_masif[neg_label]\n",
    "\n",
    "\n",
    "f, ax = plt.subplots(1, 1, figsize=(8, 4), sharex=False)\n",
    "sns.distplot(pos, hist=True, color='r', kde_kws={\"shade\": True},ax=ax,  norm_hist=True)\n",
    "sns.distplot(neg, hist=True, color='b', kde_kws={\"shade\": True},ax=ax, norm_hist=True)\n",
    "plt.xlim([0,1])\n",
    "ax.set_xticks([])\n",
    "ax.set_yticks([])\n",
    "ax.spines['right'].set_visible(False)\n",
    "ax.spines['top'].set_visible(False)\n",
    "print(metrics.roc_auc_score(all_labels_masif, all_scores_masif))\n",
    "plt.savefig('dist_plot_masif_all_points_transient.pdf', type='pdf')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.628998980373133\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# SPPIDER dist plot -- all points.\n",
    "pos_label = np.where(all_labels_by_point_sppider == 1)[0]\n",
    "pos = all_pred_by_point_sppider[pos_label]\n",
    "neg_label = np.where(all_labels_by_point_sppider == 0)[0]\n",
    "neg = all_pred_by_point_sppider[neg_label]\n",
    "f, ax = plt.subplots(1, 1, figsize=(8, 4), sharex=False)\n",
    "sns.distplot(pos, hist=True, color='r', kde_kws={\"shade\": True},ax=ax,  norm_hist=True)\n",
    "sns.distplot(neg, hist=True, color='b', kde_kws={\"shade\": True},ax=ax, norm_hist=True)\n",
    "plt.xlim([0,1])\n",
    "ax.set_xticks([])\n",
    "ax.set_yticks([])\n",
    "ax.spines['right'].set_visible(False)\n",
    "ax.spines['top'].set_visible(False)\n",
    "#axes[1].spines['top'].set_visible(False)\n",
    "#axes[0].spines['right'].set_visible(False)\n",
    "#axes[0].spines['top'].set_visible(False)\n",
    "print(metrics.roc_auc_score(all_labels_by_point_sppider, all_pred_by_point_sppider))\n",
    "plt.savefig('dist_plot_sppider.pdf', type='pdf')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}