ML/activark_benchmark.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Import libraries and the activark script (prepareTestData)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import prepareTestData as activark\n",
    "import numpy as np\n",
    "from sklearn.metrics import accuracy_score, matthews_corrcoef, roc_auc_score"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Create a class to store mutation information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Mutation:\n",
    "    def __init__(self, mutation, mutationType):\n",
    "        self.mutation = mutation\n",
    "        self.mutationType = [mutationType]\n",
    "        self.results = {}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Specificy the input file (assuming the first col is in mechismo-like format)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "INPUT_FILE = 'final_ADRN.txt'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Read the input file and store information in a dic using the class defined above as template. Save the extracted mutations in a new file called input.txt, which will used to give as input to activark as input in the next step."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "223\n"
     ]
    }
   ],
   "source": [
    "dic_mutations = {}\n",
    "for line in open(INPUT_FILE, 'r'):\n",
    "    mutation, mutationType = line.strip().split('\\t')\n",
    "    if mutation not in dic_mutations:\n",
    "        dic_mutations[mutation] = Mutation(mutation, mutationType)\n",
    "    else:\n",
    "        if mutationType not in dic_mutations[mutation].mutationType:\n",
    "            dic_mutations[mutation].mutationType.append(mutationType)\n",
    "\n",
    "l = ''\n",
    "for mutation in dic_mutations:\n",
    "    # if 'Activating' in dic_mutations[mutation].mutationType:\n",
    "        l += mutation + '\\n'\n",
    "open('input.txt', 'w').write(l)\n",
    "print (len(l.split('\\n')[:-1]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Run activark to make predictions\n",
    "The results return a JSON object with 3 keys:\n",
    "1. predictions: mutation-wise predictions\n",
    "2. overEntries: number of entries(mutations) that were ignored (by defualt first 100 mutations are consdered)\n",
    "3. entries_not_found: entries (mutations) that were ignored because they were not found in the database."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/224 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 67%|██████▋   | 150/224 [05:42<02:59,  2.43s/it]"
     ]
    }
   ],
   "source": [
    "results_gbc = activark.predict(5, 'input.txt', algo='GBC')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gzip\n",
    "# get all the predictions from the PolyPhen2 output file\n",
    "with gzip.open('../data/polyphen_output_new_test_set.tsv.gz', 'rt') as f:\n",
    "    for line in f:\n",
    "        if line.startswith('#'):\n",
    "            continue\n",
    "        else:\n",
    "            line = line.strip().split('\\t')\n",
    "            acc = line[0].strip()\n",
    "            pos = line[1].strip()\n",
    "            wtAA = line[2].strip()\n",
    "            mutAA = line[3].strip()\n",
    "            mutation = acc + '/' + wtAA+ pos + mutAA\n",
    "            prediction = line[9].strip()\n",
    "            prob = line[10].strip()\n",
    "            # print (mutation, acc, pos, wtAA, mutAA, prediction)\n",
    "            if mutation in dic_mutations:\n",
    "                if 'polyphen2' not in dic_mutations[mutation].results:\n",
    "                    dic_mutations[mutation].results['polyphen2'] = prob\n",
    "                # dic_mutations[mutation].results['polyphen2'] = prediction\n",
    "                # dic_mutations[mutation].prob = prob\n",
    "            # else:\n",
    "            #     print (f'{mutation} not found in the DB')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gzip\n",
    "# get all the predictions from the PolyPhen2 output file\n",
    "with gzip.open('../data/pmut_output.tsv.gz', 'rt') as f:\n",
    "    for line in f:\n",
    "        if line.startswith('#'):\n",
    "            continue\n",
    "        mutation = line.split()[0] + '/' + line.split()[1]\n",
    "        prob = line.split()[2].split('(')[1].split(')')[0]\n",
    "        # print (mutation, acc, pos, wtAA, mutAA, prediction)\n",
    "        if mutation in dic_mutations:\n",
    "            if 'pmut' not in dic_mutations[mutation].results:\n",
    "                dic_mutations[mutation].results['pmut'] = prob\n",
    "            # dic_mutations[mutation].results['polyphen2'] = prediction\n",
    "            # dic_mutations[mutation].prob = prob\n",
    "        # else:\n",
    "        #     print (f'{mutation} not found in the DB')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'polyphen2': '0.01', 'pmut': '0.1775'}\n",
      "{'acc': 'Q96QP1', 'gene': 'ALPK1', 'uniprot_id': 'ALPK1_HUMAN', 'protein_name': 'Alpha-protein kinase 1', 'mutation': 'V1092A', 'adjacentSites': 'TAQHYvEFNKR', 'hmmPos': '-', 'alnPos': '-', 'region': '-', 'ptmType': '-', 'mutType': '-', 'N': 'NA\\t', 'D': 'NA\\t', 'A': 'NA\\t', 'AILDvN': 'NA\\t', 'AILDRvN': 'NA\\t', 'RvN': 'NA\\t', 'AIvLD': 'NA\\t', 'AIvNLD': 'NA\\t', 'LDvNAI': 'NA\\t', 'AIvN': 'NA\\t', 'LDvN': 'NA\\t'}\n"
     ]
    }
   ],
   "source": [
    "print (dic_mutations['O96017/E501K'].results)\n",
    "print (results_gbc['predictions']['Q96QP1/V1092A'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot ROC curve\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import roc_curve, auc\n",
    "import matplotlib as mpl\n",
    "mpl.rcParams['font.size'] = 12\n",
    "\n",
    "for activark_model in ['AIvLD', 'AIvNLD', 'LDvNAI', 'RvN', 'AILDvN']:\n",
    "    # Clear the plot\n",
    "    plt.clf()\n",
    "    y_true = []\n",
    "    y_pred = {'Activark': [], 'PolyPhen2': [], 'PMUT': []}\n",
    "    for mutation in dic_mutations:\n",
    "        mutationType = dic_mutations[mutation].mutationType\n",
    "        if activark_model == 'AIvLD':\n",
    "            if 'Activating' not in mutationType and 'Deactivating' not in mutationType:\n",
    "                continue\n",
    "        elif activark_model in ['AIvNLD', 'LDvNAI']:\n",
    "            if 'Activating' not in mutationType and 'Deactivating' not in mutationType and 'Neutral' not in mutationType:\n",
    "                continue\n",
    "        elif activark_model == 'RvN':\n",
    "            if 'Resistance' not in mutationType and 'Neutral' not in mutationType:\n",
    "                continue\n",
    "        elif activark_model == 'AILDvN':\n",
    "            if 'Activating' not in mutationType and 'Deactivating' not in mutationType and 'Neutral' not in mutationType:\n",
    "                continue\n",
    "        else:\n",
    "            continue\n",
    "        if mutation not in results_gbc['predictions']:\n",
    "            continue\n",
    "        if results_gbc['predictions'][mutation][activark_model].strip() == 'NA':\n",
    "            continue\n",
    "        activark_prob = float(results_gbc['predictions'][mutation][activark_model])\n",
    "        dic_mutations[mutation].results['activark'] = activark_prob\n",
    "        if 'polyphen2' not in dic_mutations[mutation].results:\n",
    "            continue\n",
    "        dic_mutations[mutation].results['polyphen2'] = float(dic_mutations[mutation].results['polyphen2'])\n",
    "        if 'pmut' not in dic_mutations[mutation].results:\n",
    "            continue\n",
    "        dic_mutations[mutation].results['pmut'] = float(dic_mutations[mutation].results['pmut'])\n",
    "        y_pred['Activark'].append(dic_mutations[mutation].results['activark'])\n",
    "        y_pred['PolyPhen2'].append(dic_mutations[mutation].results['polyphen2'])\n",
    "        y_pred['PMUT'].append(dic_mutations[mutation].results['pmut'])\n",
    "        if activark_model == 'AIvLD':\n",
    "            y_true.append(1 if 'Activating' in dic_mutations[mutation].mutationType else 0)\n",
    "        elif activark_model in ['AIvNLD']:\n",
    "            y_true.append(1 if 'Activating' in dic_mutations[mutation].mutationType else 0)\n",
    "        elif activark_model in ['LDvNAI']:\n",
    "            y_true.append(1 if 'Deactivating' in dic_mutations[mutation].mutationType else 0)\n",
    "        elif activark_model == 'RvN':\n",
    "            y_true.append(1 if 'Resistance' in dic_mutations[mutation].mutationType else 0)\n",
    "        elif activark_model == 'AILDvN':\n",
    "            y_true.append(1 if 'Activating' in dic_mutations[mutation].mutationType or 'Deactivating' in dic_mutations[mutation].mutationType else 0)\n",
    "\n",
    "    # print (len(y_true))\n",
    "    # print (y_pred['Activark'])\n",
    "\n",
    "    for model in y_pred:\n",
    "        fpr, tpr, thresholds = roc_curve(y_true, y_pred[model])\n",
    "        roc_auc = auc(fpr, tpr)\n",
    "        label_name = f'{model} (AUC = {round(roc_auc, 3)})'\n",
    "        plt.plot(fpr, tpr, label=label_name)\n",
    "        plt.plot([0, 1], [0, 1], 'k--')\n",
    "        plt.xlim([0.0, 1.0])\n",
    "        plt.ylim([0.0, 1.05])\n",
    "    plt.xlabel('False Positive Rate')\n",
    "    plt.ylabel('True Positive Rate')\n",
    "    plt.legend(loc=\"lower right\")\n",
    "    plt.title(f'Activark {activark_model} vs PolyPhen2 vs PMUT on Test set n={len(y_true)}')\n",
    "    plt.savefig(f'comparisons/roc_{activark_model}.svg', format='svg', dpi=600)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['JAK2', 'O60674', 'H608N', 'Activating', 0.299, 0.22, 0.48, 0.665, 0.422, 0.998, 0.4058], ['JAK2', 'O60674', 'H608Y', 'Activating', 0.44, 0.207, 0.353, 0.745, 0.176, 1.0, 0.6641], ['JAK2', 'O60674', 'L604F', 'Activating', 0.493, 0.227, 0.28, 0.548, 0.767, 0.998, 0.2704], ['PAK4', 'O96013', 'I417T', 'Activating', 0.294, 0.357, 0.348, 0.551, 0.486, 0.994, 0.7833], ['EGFR', 'P00533', 'R776L', 'Activating', 0.045, 0.083, 0.872, 0.795, 0.936, 1.0, 0.8732], ['EGFR', 'P00533', 'S645C', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.999, 0.7094], ['MET', 'P08581', 'Y1234E', 'Activating', 0.009, 0.071, 0.92, 0.958, 0.941, 1.0, 'NA'], ['KIT', 'P10721', 'K509I', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.6164], ['FGFR1', 'P11362', 'D647N', 'Activating', 0.032, 0.021, 0.947, 0.95, 0.983, 1.0, 0.5541], ['BRAF', 'P15056', 'L525R', 'Activating', 0.252, 0.338, 0.409, 0.563, 0.8, 0.76, 0.7118], ['BRAF', 'P15056', 'L597R', 'Activating', 0.03, 0.566, 0.403, 0.528, 0.824, 1.0, 0.8915], ['BRAF', 'P15056', 'V600G', 'Activating', 0.007, 0.007, 0.986, 0.985, 0.993, 0.98, 0.804], ['PDGFRA', 'P16234', 'D842N', 'Activating', 0.023, 0.015, 0.962, 0.965, 0.983, 1.0, 0.7134], ['FGFR2', 'P21802', 'K659E', 'Activating', 0.023, 0.06, 0.917, 0.931, 0.611, 0.999, 0.7923], ['FGFR2', 'P21802', 'K659Q', 'Activating', 0.113, 0.072, 0.815, 0.899, 0.422, 0.999, 0.6614], ['FGFR2', 'P21802', 'K659T', 'Activating', 0.039, 0.104, 0.856, 0.892, 0.758, 0.998, 0.6724], ['FGFR2', 'P21802', 'Y375C', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.998, 0.8519], ['FGFR3', 'P22607', 'N540K', 'Activating', 0.061, 0.045, 0.893, 0.904, 0.967, 1.0, 0.4387], ['FGFR3', 'P22607', 'S249C', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.8989], ['MAPK1', 'P28482', 'E322K', 'Activating', 0.174, 0.42, 0.406, 0.63, 0.453, 1.0, 0.855], ['AXL', 'P30530', 'W451C', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.727], ['CDK6', 'Q00534', 'R31C', 'Activating', 0.188, 0.175, 0.637, 0.74, 0.196, 0.999, 0.5793], ['ACVR1', 'Q04771', 'G328R', 'Activating', 0.325, 0.254, 0.421, 0.635, 0.439, 1.0, 0.7338], ['ERBB4', 'Q15303', 'E715K', 'Activating', 0.233, 0.182, 0.585, 0.694, 0.157, 0.163, 0.7652], ['ERBB4', 'Q15303', 'R687K', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.671, 0.4732], ['TNNI3K', 'Q59H18', 'H592Y', 'Activating', 0.234, 0.574, 0.192, 0.259, 0.206, 0.652, 0.4322], ['TNNI3K', 'Q59H18', 'I512T', 'Activating', 0.089, 0.188, 0.724, 0.678, 0.777, 1.0, 0.418], ['LRRK2', 'Q5S007', 'F1700L', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.964, 0.3178], ['KSR1', 'Q8IVT5', 'S406A', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 'NA', 'NA'], ['MARK4', 'Q96L34', 'F202L', 'Activating', 0.074, 0.309, 0.617, 0.627, 0.574, 1.0, 0.4164], ['ALPK1', 'Q96QP1', 'T237M', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 'NA'], ['ALPK1', 'Q96QP1', 'V1092A', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.999, 'NA'], ['PINK1', 'Q9BXM7', 'G411A', 'Activating', 0.731, 0.159, 0.11, 0.495, 0.065, 0.059, 0.0977], ['PINK1', 'Q9BXM7', 'W403A', 'Activating', 0.542, 0.219, 0.239, 0.338, 0.195, 0.06, 0.5516], ['WNK3', 'Q9BYP7', 'E314A', 'Activating', 0.185, 0.296, 0.519, 0.738, 0.464, 1.0, 0.8814], ['WNK3', 'Q9BYP7', 'E314Q', 'Activating', 0.235, 0.219, 0.546, 0.787, 0.321, 1.0, 0.7936], ['TBK1', 'Q9UHD2', 'R573G', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.4085], ['TBK1', 'Q9UHD2', 'S511E', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.214, 0.1879], ['ALK', 'Q9UM73', 'I1171S', 'Activating', 0.301, 0.291, 0.408, 0.539, 0.772, 1.0, 0.5671], ['WNK2', 'Q9Y3S1', 'A1607V', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.995, 'NA'], ['WNK2', 'Q9Y3S1', 'P702L', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.993, 'NA'], ['WNK2', 'Q9Y3S1', 'V2053I', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.999, 'NA'], ['FGFR4', 'P22455', 'C477A', 'Activating', 0.602, 0.114, 0.284, 0.782, 0.372, 1.0, 0.4458], ['FGFR4', 'P22455', 'C552A', 'Activating', 0.231, 0.245, 0.524, 0.626, 0.972, 0.911, 0.61], ['RAF1', 'P04049', 'R231Q', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.298, 0.2594], ['PDGFRB', 'P09619', 'N666S', 'Activating', 0.055, 0.057, 0.888, 0.909, 0.984, 1.0, 0.7797], ['TRIB3', 'Q96RU7', 'Q84R', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.0, 'NA'], ['MAP3K3', 'Q99759', 'I441M', 'Activating', 0.438, 0.297, 0.265, 0.59, 0.355, 1.0, 'NA'], ['MAPK1', 'P28482', 'D321N', 'Activating', 0.411, 0.263, 0.326, 0.469, 0.423, 0.999, 0.683], ['EGFR', 'P00533', 'V774M', 'Activating', 0.622, 0.152, 0.225, 0.523, 0.318, 1.0, 0.8692], ['PDGFRB', 'P09619', 'N666K', 'Activating', 0.06, 0.053, 0.888, 0.909, 0.977, 1.0, 0.556], ['FGFR2', 'P21802', 'P253R', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.134, 0.6995], ['MAP2K1', 'Q02750', 'N122D', 'Activating', 0.716, 0.178, 0.106, 0.378, 0.63, 0.97, 0.1125], ['MAP2K1', 'Q02750', 'N345K', 'Activating', 0.452, 0.106, 0.441, 0.746, 0.188, 0.483, 0.2082], ['MAP2K1', 'Q02750', 'P124Q', 'Activating', 0.439, 0.142, 0.419, 0.693, 0.466, 1.0, 0.2536], ['JAK2', 'O60674', 'V617L', 'Activating', 0.521, 0.084, 0.395, 0.854, 0.098, 0.083, 0.2816], ['RET', 'P07949', 'C630Y', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.999, 0.6906], ['ROS1', 'P08922', 'D2113N', 'Activating', 0.221, 0.399, 0.38, 0.566, 0.294, 0.999, 0.7165], ['PDGFRB', 'P09619', 'S548Y', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.982, 0.6767], ['ALK', 'Q9UM73', 'D1225N', 'Activating', 0.724, 0.118, 0.158, 0.636, 0.245, 1.0, 0.4569], ['BRAF', 'P15056', 'V600R', 'Activating', 0.007, 0.007, 0.986, 0.989, 0.995, 0.55, 0.8748], ['FGFR2', 'P21802', 'S252W', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.999, 0.9054], ['FGFR3', 'P22607', 'N262Q', 'Activating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.999, 0.6813], ['BMPR1B', 'O00238', 'E244K', 'Deactivating', 0.157, 0.654, 0.19, 0.117, 0.396, 1.0, 0.8783], ['MAP3K7', 'O43318', 'T187I', 'Deactivating', 0.166, 0.346, 0.488, 0.378, 0.318, 0.988, 0.607], ['ULK1', 'O75385', 'S467A', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.072, 0.366], ['MAP4K4', 'O95819', 'K54R', 'Deactivating', 0.015, 0.96, 0.025, 0.04, 0.517, 0.955, 'NA'], ['CHEK2', 'O96017', 'E321A', 'Deactivating', 0.076, 0.175, 0.749, 0.702, 0.408, 1.0, 0.9373], ['CHEK2', 'O96017', 'T476M', 'Deactivating', 0.518, 0.331, 0.151, 0.261, 0.197, 1.0, 0.8967], ['RAF1', 'P04049', 'T543M', 'Deactivating', 0.111, 0.403, 0.486, 0.539, 0.411, 0.997, 0.743], ['LCK', 'P06239', 'P440S', 'Deactivating', 0.256, 0.534, 0.21, 0.39, 0.566, 0.996, 'NA'], ['RET', 'P07949', 'S811F', 'Deactivating', 0.071, 0.659, 0.269, 0.389, 0.972, 1.0, 0.9211], ['MET', 'P08581', 'S1016A', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.986, 'NA'], ['FGFR1', 'P11362', 'E670K', 'Deactivating', 0.025, 0.924, 0.05, 0.077, 0.584, 1.0, 0.9212], ['BRAF', 'P15056', 'D594G', 'Deactivating', 0.03, 0.914, 0.056, 0.117, 0.722, 0.999, 0.8915], ['BRAF', 'P15056', 'D594N', 'Deactivating', 0.042, 0.892, 0.067, 0.135, 0.683, 1.0, 0.8875], ['NPR2', 'P20594', 'G123V', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.735, 0.5464], ['NPR2', 'P20594', 'L314R', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.0, 0.2328], ['NPR2', 'P20594', 'L51P', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.975, 0.8112], ['NPR2', 'P20594', 'R388Q', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.8139], ['NPR2', 'P20594', 'R932C', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.958, 0.5868], ['EPHA1', 'P21709', 'P460L', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.999, 0.5071], ['JAK1', 'P23458', 'V666G', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.2617], ['EPHA2', 'P29317', 'P817H', 'Deactivating', 0.259, 0.437, 0.304, 0.438, 0.572, 0.999, 0.8915], ['EPHB2', 'P29323', 'S706F', 'Deactivating', 0.084, 0.695, 0.221, 0.315, 0.96, 0.996, 'NA'], ['AKT1', 'P31749', 'G159V', 'Deactivating', 0.034, 0.855, 0.111, 0.16, 0.757, 0.943, 0.8986], ['AKT1', 'P31749', 'G336D', 'Deactivating', 0.196, 0.543, 0.261, 0.239, 0.236, 1.0, 0.8986], ['AKT1', 'P31749', 'H265Y', 'Deactivating', 0.126, 0.702, 0.172, 0.325, 0.337, 0.988, 0.8986], ['AKT1', 'P31749', 'R273Q', 'Deactivating', 0.057, 0.462, 0.481, 0.558, 0.282, 0.998, 0.8986], ['AKT2', 'P31751', 'P50T', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.001, 0.2746], ['ACVRL1', 'P37023', 'I256N', 'Deactivating', 0.123, 0.678, 0.199, 0.318, 0.694, 1.0, 0.9403], ['ACVRL1', 'P37023', 'L285P', 'Deactivating', 0.045, 0.904, 0.051, 0.094, 0.673, 1.0, 0.9403], ['ACVRL1', 'P37023', 'P424L', 'Deactivating', 0.475, 0.418, 0.107, 0.268, 0.311, 1.0, 0.9403], ['ACVRL1', 'P37023', 'R67Q', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.154, 0.31], ['ZAP70', 'P43403', 'D521N', 'Deactivating', 0.27, 0.551, 0.18, 0.216, 0.513, 1.0, 0.8693], ['ZAP70', 'P43403', 'W163C', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.8915], ['GSK3B', 'P49841', 'Y216E', 'Deactivating', 0.014, 0.671, 0.315, 0.246, 0.348, 1.0, 0.6299], ['GSK3B', 'P49841', 'Y216F', 'Deactivating', 0.005, 0.891, 0.104, 0.08, 0.585, 1.0, 0.9048], ['MAPK12', 'P53778', 'D171A', 'Deactivating', 0.021, 0.949, 0.03, 0.042, 0.561, 1.0, 0.7865], ['CDK6', 'Q00534', 'K43M', 'Deactivating', 0.008, 0.98, 0.012, 0.017, 0.533, 1.0, 0.855], ['MAP2K1', 'Q02750', 'K97M', 'Deactivating', 0.007, 0.981, 0.013, 0.023, 0.44, 1.0, 0.7749], ['GUCY2D', 'Q02846', 'R838S', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.7754], ['PRKCD', 'Q05655', 'G510S', 'Deactivating', 0.249, 0.479, 0.272, 0.456, 0.469, 1.0, 'NA'], ['ATM', 'Q13315', 'N358I', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.017, 'NA'], ['ATM', 'Q13315', 'N796H', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.985, 'NA'], ['RIPK1', 'Q13546', 'D138N', 'Deactivating', 0.04, 0.767, 0.193, 0.185, 0.754, 0.999, 0.8555], ['BMPR2', 'Q13873', 'Q721R', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.201, 0.1988], ['PDK1', 'Q15118', 'G380D', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 'NA'], ['TNNI3K', 'Q59H18', 'P742L', 'Deactivating', 0.693, 0.15, 0.157, 0.626, 0.232, 0.906, 0.2822], ['LRRK2', 'Q5S007', 'W1295A', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.5654], ['STRADA', 'Q7RTN6', 'S264R', 'Deactivating', 0.127, 0.701, 0.171, 0.221, 0.54, 1.0, 'NA'], ['ACVR1C', 'Q8NER5', 'I195T', 'Deactivating', 0.26, 0.461, 0.279, 0.518, 0.285, 0.997, 0.725], ['ACVR1C', 'Q8NER5', 'N150H', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.003, 0.2184], ['WNK4', 'Q96J92', 'D321A', 'Deactivating', 0.016, 0.963, 0.021, 0.027, 0.531, 1.0, 'NA'], ['TP53RK', 'Q96S44', 'R249G', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 'NA'], ['PINK1', 'Q9BXM7', 'G409V', 'Deactivating', 0.124, 0.737, 0.139, 0.217, 0.602, 1.0, 0.9074], ['IRAK4', 'Q9NWZ3', 'D320Y', 'Deactivating', 0.256, 0.49, 0.254, 0.355, 0.566, 1.0, 0.8606], ['IRAK4', 'Q9NWZ3', 'Q29P', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.109, 0.2551], ['TBK1', 'Q9UHD2', 'S511A', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.012, 0.1969], ['TLK1', 'Q9UKI8', 'M566T', 'Deactivating', 0.116, 0.304, 0.58, 0.658, 0.387, 0.997, 'NA'], ['TLK1', 'Q9UKI8', 'Q479E', 'Deactivating', 0.298, 0.164, 0.538, 0.808, 0.164, 0.886, 'NA'], ['ALK', 'Q9UM73', 'F1174S', 'Deactivating', 0.253, 0.395, 0.352, 0.562, 0.945, 1.0, 0.8396], ['TNNI3K', 'Q59H18', 'L577F', 'Deactivating', 0.136, 0.422, 0.442, 0.444, 0.902, 1.0, 0.6574], ['EGFR', 'P00533', 'Y891D', 'Deactivating', 0.051, 0.6, 0.35, 0.425, 0.621, 1.0, 0.8732], ['KIT', 'P10721', 'P665L', 'Deactivating', 0.52, 0.25, 0.23, 0.495, 0.405, 1.0, 0.9181], ['ZAP70', 'P43403', 'R170C', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.8758], ['ERBB4', 'Q15303', 'N706D', 'Deactivating', 0.187, 0.392, 0.422, 0.557, 0.456, 0.997, 0.3973], ['ROCK2', 'O75116', 'K1007R', 'Deactivating', 'NA', 'NA', 'NA', 'NA', 'NA', 0.006, 'NA'], ['JAK2', 'O60674', 'G993A', 'Resistance', 0.708, 0.074, 0.218, 0.677, 0.272, 1.0, 0.2969], ['EGFR', 'P00533', 'G724S', 'Resistance', 0.137, 0.298, 0.566, 0.664, 0.681, 1.0, 0.8526], ['EGFR', 'P00533', 'L718M', 'Resistance', 0.526, 0.123, 0.351, 0.58, 0.787, 1.0, 0.7844], ['EGFR', 'P00533', 'N540K', 'Resistance', 'NA', 'NA', 'NA', 'NA', 'NA', 0.989, 0.78], ['ERBB2', 'P04626', 'L755S', 'Resistance', 0.193, 0.348, 0.459, 0.607, 0.843, 1.0, 0.8566], ['ERBB2', 'P04626', 'R784G', 'Resistance', 0.053, 0.098, 0.848, 0.798, 0.907, 1.0, 0.8732], ['CSF1R', 'P07333', 'G795A', 'Resistance', 0.591, 0.124, 0.285, 0.67, 0.479, 0.986, 0.2617], ['RET', 'P07949', 'G810C', 'Resistance', 0.205, 0.28, 0.516, 0.563, 0.848, 1.0, 0.9212], ['RET', 'P07949', 'G810D', 'Resistance', 0.137, 0.261, 0.603, 0.482, 0.923, 1.0, 0.9212], ['RET', 'P07949', 'G810R', 'Resistance', 0.184, 0.287, 0.53, 0.435, 0.901, 1.0, 0.9212], ['RET', 'P07949', 'G810S', 'Resistance', 0.205, 0.269, 0.526, 0.563, 0.923, 1.0, 0.9172], ['MET', 'P08581', 'F1200I', 'Resistance', 0.251, 0.39, 0.36, 0.461, 0.547, 1.0, 'NA'], ['MET', 'P08581', 'F1200L', 'Resistance', 0.245, 0.457, 0.298, 0.345, 0.574, 1.0, 'NA'], ['MET', 'P08581', 'G1163R', 'Resistance', 0.2, 0.26, 0.54, 0.458, 0.823, 1.0, 'NA'], ['MET', 'P08581', 'L1195V', 'Resistance', 0.168, 0.338, 0.494, 0.554, 0.876, 0.998, 'NA'], ['MET', 'P08581', 'M1250T', 'Resistance', 0.161, 0.489, 0.351, 0.427, 0.436, 0.999, 'NA'], ['ROS1', 'P08922', 'D1988N', 'Resistance', 0.577, 0.23, 0.194, 0.578, 0.19, 0.993, 0.5553], ['ROS1', 'P08922', 'D2033N', 'Resistance', 0.302, 0.322, 0.376, 0.42, 0.959, 0.996, 0.2895], ['ROS1', 'P08922', 'F2004V', 'Resistance', 0.284, 0.419, 0.297, 0.488, 0.82, 1.0, 0.7188], ['ROS1', 'P08922', 'G1957A', 'Resistance', 0.109, 0.681, 0.209, 0.35, 0.274, 0.998, 0.8624], ['ROS1', 'P08922', 'G2032R', 'Resistance', 0.211, 0.237, 0.552, 0.482, 0.786, 1.0, 0.8733], ['ROS1', 'P08922', 'L1982V', 'Resistance', 0.184, 0.401, 0.415, 0.621, 0.706, 0.989, 0.3726], ['ROS1', 'P08922', 'L2010M', 'Resistance', 0.547, 0.088, 0.365, 0.738, 0.234, 1.0, 0.212], ['ROS1', 'P08922', 'L2026M', 'Resistance', 0.8, 0.03, 0.17, 0.563, 0.956, 1.0, 0.203], ['ROS1', 'P08922', 'L2086F', 'Resistance', 0.454, 0.309, 0.237, 0.329, 0.387, 1.0, 0.8693], ['ROS1', 'P08922', 'S1986F', 'Resistance', 0.284, 0.415, 0.302, 0.504, 0.805, 0.999, 0.298], ['ROS1', 'P08922', 'S1986Y', 'Resistance', 0.293, 0.376, 0.331, 0.576, 0.83, 0.999, 0.2137], ['KIT', 'P10721', 'N828K', 'Resistance', 0.424, 0.272, 0.304, 0.472, 0.403, 1.0, 0.6336], ['KIT', 'P10721', 'V560G', 'Resistance', 0.106, 0.717, 0.177, 0.374, 0.523, 0.888, 0.5873], ['BRAF', 'P15056', 'K601E', 'Resistance', 0.044, 0.072, 0.884, 0.844, 0.587, 0.913, 0.3449], ['ERBB3', 'P21860', 'G337R', 'Resistance', 'NA', 'NA', 'NA', 'NA', 'NA', 0.976, 0.6342], ['FGFR4', 'P22455', 'N535K', 'Resistance', 0.072, 0.046, 0.882, 0.904, 0.967, 1.0, 0.4202], ['FGFR3', 'P22607', 'M528I', 'Resistance', 0.2, 0.058, 0.742, 0.842, 0.699, 0.975, 0.3191], ['FGFR3', 'P22607', 'V553M', 'Resistance', 0.83, 0.058, 0.112, 0.669, 0.125, 1.0, 0.3491], ['FGFR3', 'P22607', 'V555M', 'Resistance', 0.7, 0.057, 0.243, 0.782, 0.954, 0.999, 0.3323], ['FLT3', 'P36888', 'D698N', 'Resistance', 0.606, 0.11, 0.284, 0.613, 0.928, 0.516, 0.3062], ['FLT3', 'P36888', 'N676K', 'Resistance', 0.051, 0.052, 0.897, 0.909, 0.979, 1.0, 0.5537], ['FLT3', 'P36888', 'N701K', 'Resistance', 0.741, 0.091, 0.168, 0.642, 0.34, 0.998, 0.6915], ['CDK9', 'P50750', 'L156F', 'Resistance', 0.683, 0.124, 0.193, 0.406, 0.175, 1.0, 0.5803], ['BTK', 'Q06187', 'C481F', 'Resistance', 0.105, 0.732, 0.163, 0.167, 0.982, 0.999, 0.8396], ['BTK', 'Q06187', 'L528W', 'Resistance', 0.18, 0.588, 0.233, 0.182, 0.68, 0.993, 0.8555], ['BTK', 'Q06187', 'T474I', 'Resistance', 0.226, 0.265, 0.509, 0.518, 0.949, 1.0, 0.2516], ['NTRK3', 'Q16288', 'F617L', 'Resistance', 0.304, 0.161, 0.535, 0.576, 0.974, 1.0, 0.3092], ['NTRK3', 'Q16288', 'G623E', 'Resistance', 0.105, 0.185, 0.71, 0.658, 0.88, 1.0, 0.8335], ['NTRK3', 'Q16288', 'G623R', 'Resistance', 0.183, 0.226, 0.592, 0.561, 0.884, 1.0, 0.8335], ['PINK1', 'Q9BXM7', 'M318G', 'Resistance', 0.077, 0.444, 0.479, 0.337, 0.967, 0.993, 0.6745], ['ALK', 'Q9UM73', 'D1203N', 'Resistance', 0.635, 0.121, 0.245, 0.548, 0.876, 0.865, 0.2578], ['ALK', 'Q9UM73', 'F1174C', 'Resistance', 0.41, 0.22, 0.37, 0.767, 0.874, 1.0, 0.5797], ['ALK', 'Q9UM73', 'I1171T', 'Resistance', 0.329, 0.241, 0.43, 0.445, 0.707, 1.0, 0.5618], ['ALK', 'Q9UM73', 'L1256F', 'Resistance', 0.333, 0.346, 0.322, 0.367, 0.51, 1.0, 0.8335], ['ALK', 'Q9UM73', 'V1180L', 'Resistance', 0.457, 0.137, 0.407, 0.754, 0.154, 0.951, 0.3288], ['MET', 'P08581', 'L1195F', 'Resistance', 0.164, 0.316, 0.521, 0.535, 0.86, 1.0, 'NA'], ['KIT', 'P10721', 'L783V', 'Resistance', 0.23, 0.191, 0.578, 0.63, 0.894, 0.996, 0.9141], ['LCK', 'P06239', 'T316I', 'Resistance', 0.228, 0.215, 0.557, 0.533, 0.95, 1.0, 'NA'], ['AURKA', 'O14965', 'S104L', 'Neutral', 0.509, 0.359, 0.133, 0.306, 0.168, 0.044, 0.2093], ['GAK', 'O14976', 'S1168N', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.104, 'NA'], ['MUSK', 'O15146', 'L629F', 'Neutral', 0.282, 0.081, 0.636, 0.708, 0.413, 0.989, 0.47], ['MAST3', 'O60307', 'G883S', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 'NA'], ['CHEK2', 'O96017', 'E501K', 'Neutral', 0.91, 0.029, 0.061, 0.736, 0.077, 0.01, 0.1775], ['MOS', 'P00540', 'S300P', 'Neutral', 0.806, 0.082, 0.113, 0.621, 0.086, 0.873, 'NA'], ['FGR', 'P09769', 'S130R', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.607, 0.7255], ['MAK', 'P20794', 'N384S', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.0, 0.3546], ['EPHA5', 'P54756', 'S235A', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.009, 0.2602], ['CDKL1', 'Q00532', 'L67P', 'Neutral', 0.057, 0.392, 0.551, 0.568, 0.65, 0.993, 0.7199], ['TEK', 'Q02763', 'L634F', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.001, 0.4045], ['RNASEL', 'Q05823', 'S406F', 'Neutral', 0.24, 0.425, 0.335, 0.402, 0.5, 0.993, 0.6682], ['ATM', 'Q13315', 'S788R', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.068, 'NA'], ['PKN2', 'Q16513', 'A197E', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 0.2517], ['TNNI3K', 'Q59H18', 'D151H', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.026, 0.2682], ['CDKL4', 'Q5MAI5', 'R307C', 'Neutral', 0.43, 0.247, 0.323, 0.492, 0.125, 0.001, 0.4517], ['TTBK2', 'Q6IQ55', 'E842G', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.877, 0.2073], ['ALPK2', 'Q86TB3', 'A29V', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.022, 'NA'], ['VRK2', 'Q86Y07', 'I157M', 'Neutral', 0.262, 0.169, 0.57, 0.671, 0.707, 0.975, 0.158], ['VRK2', 'Q86Y07', 'N211S', 'Neutral', 0.193, 0.581, 0.226, 0.332, 0.283, 0.93, 0.2869], ['MAPKAPK5', 'Q8IW41', 'M67I', 'Neutral', 0.704, 0.132, 0.164, 0.479, 0.263, 0.176, 0.1844], ['MYO3A', 'Q8NEV4', 'V1195A', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.0, 0.1349], ['TRPM7', 'Q96QT4', 'A1254V', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.4, 'NA'], ['PASK', 'Q96RG2', 'T512A', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.0, 0.0467], ['RIOK2', 'Q9BVS4', 'R155H', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 1.0, 'NA'], ['STK31', 'Q9BXU1', 'T987S', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.013, 'NA'], ['POMK', 'Q9H5K3', 'M301T', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.063, 0.5226], ['EIF2AK4', 'Q9P2K8', 'Q1406H', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.077, 0.2982], ['MAP3K2', 'Q9Y2U5', 'I110V', 'Neutral', 'NA', 'NA', 'NA', 'NA', 'NA', 0.0, 'NA'], ['IRAK3', 'Q9Y616', 'S288L', 'Neutral', 0.528, 0.203, 0.269, 0.688, 0.172, 0.968, 'NA']]\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "data = []\n",
    "for mutation in dic_mutations:\n",
    "    for mutationType in dic_mutations[mutation].mutationType:\n",
    "        if mutation not in results_gbc['predictions']:\n",
    "            continue\n",
    "        row = []\n",
    "        gene_name = results_gbc['predictions'][mutation]['gene']\n",
    "        row.append(gene_name)\n",
    "        acc = results_gbc['predictions'][mutation]['acc']\n",
    "        row.append(acc)\n",
    "        mut = results_gbc['predictions'][mutation]['mutation']\n",
    "        row.append(mut)\n",
    "        row.append(mutationType)\n",
    "        \n",
    "        activark_models = ['N', 'D', 'A', 'AIvLD', 'RvN']\n",
    "        for activark_model in activark_models:\n",
    "            if results_gbc['predictions'][mutation][activark_model].strip() == 'NA':\n",
    "                activark_prob = 'NA'\n",
    "            else:\n",
    "                activark_prob = float(results_gbc['predictions'][mutation][activark_model])\n",
    "            row.append(activark_prob)\n",
    "        if 'polyphen2' not in dic_mutations[mutation].results:\n",
    "            row.append('NA')\n",
    "        else:\n",
    "            row.append(float(dic_mutations[mutation].results['polyphen2']))\n",
    "        if 'pmut' not in dic_mutations[mutation].results:\n",
    "            row.append('NA')\n",
    "        else:\n",
    "            row.append(float(dic_mutations[mutation].results['pmut']))\n",
    "        data.append(row)\n",
    "print (data)\n",
    "df = pd.DataFrame(data, columns=['Gene', 'UniProt_Acc', 'Mutation', 'MutationType',  'N', 'D', 'A', 'AIvLD', 'RvN', 'PolyPhen2', 'PMUT'])\n",
    "df.to_csv('comparisons.tsv', sep='\\t', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'dic_mutations' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[1], line 9\u001b[0m\n\u001b[1;32m      7\u001b[0m activark_models \u001b[38;5;241m=\u001b[39m [\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mAILDvN\u001b[39m\u001b[38;5;124m'\u001b[39m]\n\u001b[1;32m      8\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m activark_model \u001b[38;5;129;01min\u001b[39;00m activark_models:\n\u001b[0;32m----> 9\u001b[0m     \u001b[38;5;28;01mfor\u001b[39;00m mutation \u001b[38;5;129;01min\u001b[39;00m \u001b[43mdic_mutations\u001b[49m:\n\u001b[1;32m     10\u001b[0m         \u001b[38;5;28;01mfor\u001b[39;00m mutationType \u001b[38;5;129;01min\u001b[39;00m dic_mutations[mutation]\u001b[38;5;241m.\u001b[39mmutationType:\n\u001b[1;32m     11\u001b[0m             \u001b[38;5;66;03m# print (mutation, mutationType)\u001b[39;00m\n\u001b[1;32m     12\u001b[0m             \u001b[38;5;28;01mif\u001b[39;00m mutation \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;129;01min\u001b[39;00m results_gbc[\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mpredictions\u001b[39m\u001b[38;5;124m'\u001b[39m]:\n",
      "\u001b[0;31mNameError\u001b[0m: name 'dic_mutations' is not defined"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from sklearn.metrics import roc_auc_score, matthews_corrcoef, f1_score\n",
    "\n",
    "y_pred_prob = []\n",
    "y_pred_label = []\n",
    "y_true = []\n",
    "activark_models = ['AILDvN']\n",
    "for activark_model in activark_models:\n",
    "    for mutation in dic_mutations:\n",
    "        for mutationType in dic_mutations[mutation].mutationType:\n",
    "            # print (mutation, mutationType)\n",
    "            if mutation not in results_gbc['predictions']:\n",
    "                continue\n",
    "            if results_gbc['predictions'][mutation][activark_model].strip() == 'NA':\n",
    "                continue\n",
    "            # if mutationType == 'Resistance':\n",
    "            # if mutationType == 'Activating':\n",
    "            # if mutationType == 'Deactivating':\n",
    "            # if mutationType == 'Activating' or mutationType == 'Deactivating':\n",
    "            if mutationType == 'Activating' or mutationType == 'Deactivating' or mutationType == 'Resistance':\n",
    "                y_true.append(1)\n",
    "            elif mutationType == 'Neutral':\n",
    "            # elif mutationType == 'Deactivating':\n",
    "                y_true.append(0)\n",
    "            else:\n",
    "                continue\n",
    "            y_pred_prob.append(float(results_gbc['predictions'][mutation][activark_model]))\n",
    "            y_pred_label.append(1 if float(results_gbc['predictions'][mutation][activark_model]) > 0.5 else 0)\n",
    "print (len(y_true), len(y_pred_prob))\n",
    "print ('auc', roc_auc_score(y_true, y_pred_prob))\n",
    "print ('mcc', matthews_corrcoef(y_true, y_pred_label))\n",
    "print ('f1', f1_score(y_true, y_pred_label))\n",
    "# print count of 1 in y_true and count of 0 in y_true\n",
    "print (y_true.count(1), y_true.count(0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'Activark': [0.72, 0.503, 0.743, 0.838, 0.946, 0.954, 0.923, 0.999, 0.985, 0.962, 0.994, 0.988, 0.997, 0.969, 0.968, 0.951, 0.404, 0.768, 0.888, 0.86, 0.989, 0.164, 0.681, 0.89, 0.833, 0.841, 0.769, 0.879, 0.967, 0.683, 0.122, 0.972, 0.27, 0.416, 0.332, 0.047, 0.823, 0.127, 0.993, 0.954, 0.944, 0.901, 0.246, 0.963, 0.973, 0.986, 0.981, 0.981, 0.907, 0.994, 0.888, 0.897, 0.981, 0.874, 0.974, 0.21, 0.77, 0.997, 0.999, 0.97, 0.996, 0.997, 0.993, 0.106, 0.767, 0.965, 0.822, 0.979, 0.973, 0.978, 0.344, 0.887, 0.492, 0.633, 0.021, 0.972, 0.766, 0.427, 0.8, 0.838, 0.189], 'PolyPhen2': [0.998, 1.0, 0.998, 0.994, 1.0, 1.0, 0.76, 1.0, 0.98, 1.0, 0.999, 0.999, 0.998, 1.0, 1.0, 0.999, 1.0, 0.163, 0.652, 1.0, 1.0, 0.059, 0.06, 1.0, 1.0, 1.0, 1.0, 0.911, 1.0, 0.999, 1.0, 1.0, 0.97, 0.483, 1.0, 0.083, 0.999, 1.0, 0.55, 1.0, 0.988, 1.0, 1.0, 0.997, 1.0, 1.0, 0.999, 1.0, 0.999, 0.943, 1.0, 0.988, 0.998, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.999, 0.906, 0.997, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.997, 0.044, 0.989, 0.01, 0.993, 0.993, 0.001, 0.975, 0.93, 0.176], 'PMUT': [0.4058, 0.6641, 0.2704, 0.7833, 0.8732, 0.5541, 0.7118, 0.8915, 0.804, 0.7134, 0.7923, 0.6614, 0.6724, 0.4387, 0.855, 0.5793, 0.7338, 0.7652, 0.4322, 0.418, 0.4164, 0.0977, 0.5516, 0.8814, 0.7936, 0.5671, 0.4458, 0.61, 0.7797, 0.683, 0.8692, 0.556, 0.1125, 0.2082, 0.2536, 0.2816, 0.7165, 0.4569, 0.8748, 0.8783, 0.607, 0.9373, 0.8967, 0.743, 0.9211, 0.9212, 0.8915, 0.8875, 0.8915, 0.8986, 0.8986, 0.8986, 0.8986, 0.9403, 0.9403, 0.9403, 0.8693, 0.6299, 0.9048, 0.7865, 0.855, 0.7749, 0.8555, 0.2822, 0.725, 0.9074, 0.8606, 0.8396, 0.6574, 0.8732, 0.9181, 0.3973, 0.2093, 0.47, 0.1775, 0.7199, 0.6682, 0.4517, 0.158, 0.2869, 0.1844]}\n",
      "auc 0.8307749161407697\n",
      "mcc 0.4196865125339547\n",
      "f1 0.6442165214081723\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from sklearn.metrics import roc_auc_score, matthews_corrcoef, f1_score\n",
    "\n",
    "y_pred_prob = []\n",
    "y_pred_label = []\n",
    "y_true = []\n",
    "y_mcc_true = []\n",
    "y_mcc_label = []\n",
    "for mutation in dic_mutations:\n",
    "    for mutationType in dic_mutations[mutation].mutationType:\n",
    "        if mutationType == 'Resistance':\n",
    "            continue\n",
    "        if mutation not in results_gbc['predictions']:\n",
    "            continue\n",
    "        if results_gbc['predictions'][mutation][activark_model].strip() == 'NA':\n",
    "            continue\n",
    "        row_true = []\n",
    "        if mutationType == 'Neutral':\n",
    "            row_true.append(1)\n",
    "            y_mcc_true.append(0)\n",
    "        else:\n",
    "            row_true.append(0)\n",
    "        if mutationType == 'Activating':\n",
    "            row_true.append(1)\n",
    "            y_mcc_true.append(1)\n",
    "        else:\n",
    "            row_true.append(0)\n",
    "        if mutationType == 'Deactivating':\n",
    "            row_true.append(1)\n",
    "            y_mcc_true.append(2)\n",
    "        else:\n",
    "            row_true.append(0)\n",
    "        activar_N = float(results_gbc['predictions'][mutation]['N'])\n",
    "        activar_A = float(results_gbc['predictions'][mutation]['A'])\n",
    "        activar_D = float(results_gbc['predictions'][mutation]['D'])\n",
    "        row_preds = [activar_N, activar_A, activar_D]\n",
    "        y_true.append(row_true)\n",
    "        y_pred_prob.append(row_preds)\n",
    "        if activar_N > activar_A and activar_N > activar_D:\n",
    "            y_mcc_label.append(0)\n",
    "        elif activar_A > activar_N and activar_A > activar_D:\n",
    "            y_mcc_label.append(1)\n",
    "        else:\n",
    "            y_mcc_label.append(2)\n",
    "        \n",
    "print (y_pred)\n",
    "print ('auc', roc_auc_score(y_true, y_pred_prob, multi_class='ovr'))\n",
    "print ('mcc', matthews_corrcoef(y_mcc_true, y_mcc_label))\n",
    "print ('f1', f1_score(y_mcc_true, y_mcc_label, average='weighted'))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "activark",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}