notebooks/bte_sleeve.ipynb

{
 "cells": [
  {
   "cell_type": "raw",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "UZZqPBS0UppM",
    "outputId": "281abfa6-b935-4d99-a267-e237900aa5dc"
   },
   "source": [
    "!pip install numpy pandas requests requests_cache SetSimilaritySearch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "uG7JfetZjcXH"
   },
   "outputs": [],
   "source": [
    "from copy import copy,deepcopy\n",
    "from itertools import combinations\n",
    "import json\n",
    "import textwrap\n",
    "\n",
    "\n",
    "from fisher import pvalue_npy\n",
    "import matplotlib.pyplot as plt\n",
    "import networkx as nx\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "id": "rzoUUSDeVz0a"
   },
   "outputs": [],
   "source": [
    "import requests\n",
    "import requests_cache\n",
    "\n",
    "\n",
    "requests_cache.install_cache(\n",
    "    \"pfocr_cache\",\n",
    "    allowable_methods=['GET', 'POST'],\n",
    "    stale_if_error=False,\n",
    ")\n",
    "#requests_cache.remove_expired_responses()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "j9DiSC9NU_Fa"
   },
   "source": [
    "# BTE Sleeve\n",
    "\n",
    "TLDR: skip ahead to the \"View TRAPI results and figures\" section to see figures by TRAPI result and TRAPI results by figure.\n",
    "\n",
    "Some queries result in an explosion of the number of results. This is bad for performance of the server and also not very useful to the user. This notebook explores using PFOCR as a \"compression sleeve\" to dampen these types of explosions. I don't have a CC-licensed image of a compression sleeve, so instead here's an image of a heat-shrink tube.\n",
    "\n",
    "<a title=\"oomlout, CC BY-SA 2.0 &lt;https://creativecommons.org/licenses/by-sa/2.0&gt;, via Wikimedia Commons\" href=\"https://commons.wikimedia.org/wiki/File:HESH-06-L-STAN-01_(9601242883).jpg\"><img width=\"512\" alt=\"HESH-06-L-STAN-01 (9601242883)\" src=\"https://upload.wikimedia.org/wikipedia/commons/thumb/d/d3/HESH-06-L-STAN-01_%289601242883%29.jpg/512px-HESH-06-L-STAN-01_%289601242883%29.jpg\"></a>\n",
    "\n",
    "Note BTE Sleeve currently only works on queries with nodes of categories supported by PFOCR, e.g., `biolink:Gene`, `biolink:SmallMolecule`, `biolink:Disease`. TODO: update the Fisher's Exact test to work with node categories not in PFOCR."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "TxW41muraAFL"
   },
   "source": [
    "## Get BTE TRAPI Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "N71NYHWeVCj9",
    "outputId": "d8c2c9ae-9750-455c-ff61-b357b4ff0b2d",
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "200\n",
      "got response from cache\n"
     ]
    }
   ],
   "source": [
    "query = {\n",
    "    \"message\": {\n",
    "        \"query_graph\": {\n",
    "            \"edges\": {\n",
    "                \"e0\": {\n",
    "                    \"subject\": \"n0\",\n",
    "                    \"object\": \"n1\",\n",
    "                    \"predicates\": [\n",
    "                        \"biolink:entity_regulates_entity\",\n",
    "                        \"biolink:genetically_interacts_with\"\n",
    "                    ]\n",
    "                },\n",
    "                \"e1\": {\n",
    "                    \"subject\": \"n1\",\n",
    "                    \"object\": \"n2\",\n",
    "                    \"predicates\": [\n",
    "                        \"biolink:related_to\"\n",
    "                    ]\n",
    "                }\n",
    "            },\n",
    "            \"nodes\": {\n",
    "                \"n0\": {\n",
    "                    \"ids\": [\n",
    "                        \"NCBIGene:5872\"\n",
    "                    ],\n",
    "                    \"categories\": [\n",
    "                        \"biolink:Gene\"\n",
    "                    ]\n",
    "                },\n",
    "                \"n1\": {\n",
    "                    \"categories\": [\n",
    "                        \"biolink:Gene\"\n",
    "                    ],\n",
    "                },\n",
    "                \"n2\": {\n",
    "                    \"categories\": [\n",
    "                        \"biolink:SmallMolecule\"\n",
    "                    ]\n",
    "                }\n",
    "            }\n",
    "        }\n",
    "    }\n",
    "}\n",
    "\n",
    "trapi_response = requests.post(\"http://localhost:3000/v1/query\", json=query)\n",
    "#trapi_response = requests.post(\"https://api.bte.ncats.io/v1/query\", json=query)\n",
    "#trapi_response = requests.post(\"https://arax.ncats.io/api/rtxkg2/v1.2/query\", json=query)\n",
    "#trapi_response = requests.post(\"https://arax.ncats.io/api/arax/v1.2/query\", json=query)\n",
    "print(trapi_response.status_code)\n",
    "if trapi_response.status_code != 200:\n",
    "    raise Exception(trapi_response.text)\n",
    "if trapi_response.from_cache:\n",
    "    print(f\"got response from cache\")\n",
    "\n",
    "trapi_message = trapi_response.json()[\"message\"]\n",
    "\n",
    "if len(trapi_message[\"results\"]) == 0:\n",
    "    raise Exception(f'No results returned')"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "print(json.dumps(query))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Analyze this GraphML file in Cytoscape: ./bte_sleeve_query_template.graphml\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{('n0',\n",
       "  'n1'): Text(0.48768536731014417, 0.49969156770226864, 'biolink:entity_regulates_entity\\nbiolink:genetically_interacts_with'),\n",
       " ('n1', 'n2'): Text(-0.4879853217817037, -0.5, 'biolink:related_to')}"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeQAAAEuCAYAAABf3WErAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8qNh9FAAAACXBIWXMAAAsTAAALEwEAmpwYAABSiUlEQVR4nO3dd3yN5//H8dfJlkFiJBFCiL1iftXeQUl/RW21S41SarViFFWUarVUa9RsrBoltto1axMrIkaEEEGCSE7O74/LORkSQpOck+TzfDzyaHKfe1znKO98rvu6r0uj0+l0CCGEEMKozIzdACGEEEJIIAshhBAmQQJZCCGEMAESyEIIIYQJkEAWQgghTIAEshBCCGECJJCFEEKkgQOAHaDJAl92wMG0/XhSQSPPIQshhPhvTgJ1gUhjNyQN2QP7gYoZdkUJZCGEEP/BVaAq8MjYDUkHjsAxoFiGXE0CWQghxDuKRVWQAUCccZuSLsyAMsApwDxDriaEEEK8g2nAdbJmGIN6X0Go95n+JJCFEEK8gwBgEhCV4VfetQtKlQJbW2jQAIKD0/NqUcBE1PtNXxLIQggh3lIs0BZ4nuFXvn8fWreGiRMhPByqVoX27dP7qs+BdoA2Xa8igSyEEOItTUN15abfECQPD5g+HSpUgFy5VOg+fw5r10LZstC2LdjYwPjxcPo0XLyYbk1Bvc/077qWQBZCCPEWAlFd1U/T/UqrVsHWrRAUBGfOwKJFcP48eHnF72NnB56eanv60nddB6bbFSSQhRBCvIXpQEyGXGnQIHBzg9y5wccHTp2CyEhVMSeUKxc8eZIRLYoBZqTb2SWQhRBCpFIUsAR1Dzn9ubrGf29rq8LY3h4eP0683+PH4OCQES2KBRaTXgPZJJCFEEKk0h+oqSWNp2xZdc9YLyoKAgPV9oyhAVaky5klkIUQQqSCDpiKMR5zSqhVKzh3Dv78Uw3ymjBBDfwqVSqjWhAFTCE9BrRJIAshhEiFw0CosRtBvnwqjEePBicnOHIEVqRPwfoad4AjaX5WmTpTCCFEKrQB1pGejzplHmZAK2BNmp5VAlkIIcQbxAG2QLSxG2JCrFGPfqVdR7N0WQshhHiDK4CFsRthYixQn0vakUAWQgjxBseRuEjKDPg3zc8ohBBCvMYhINLYjTAxkajPJe1IIAshhHiDA8hgrqR0qM8l7UggCyFENuTh4YGzszNRUfHPFc+fP5/69esDoNPpmDVrFuXKlcPO7jQFC6oFHc6eVft27w5WVmrmLAcHqFIF9u6NP/+iRVC7duJrrlgB1aur+aedndX3c+ZARg8tvn4dNBrVdv3XxInxr4eHq8Us8uSBvHmhc+f42cFu3Eh43Cns7e3RaDTMmKGm1PT396d27do4Ojri6upK7969eZLKeT0lkIUQIpvSarX8+OOPyb42ePBgfvzxR2bNGk54uB2XL8OHH4K/f/w+I0ao6SwfP4Z+/dSyiNoUViicMQMGD4bhwyE0FO7ehblz4eBBePEi7d9bakREqPZHRsKYMfHbfX3h4UO1qEVgoGrr+PHqtUKF4o+JjLTj7NkNmJmZ0aZNGwAePXqEr68vISEhBAQEcPv2bYYPH56q9kggCyFENjV8+HCmT59OREREou1Xrlxh9uzZ+Pn50bChO9bWFtjaqkpx1KhXz6PRQKdOqrK8e/fV1x89grFjVTX80UeqotZooFIlWL4crK3VftHRMGyYCj0XF/j0U3j2TL22Zw8ULKiC3dkZ8ueH33+Pv8brjn1bQUHql4+cOdXCFa1apbSalAVLliylbt26eHh4ANCpUyeaNWuGra0tTk5OfPLJJxw8eDBV15VAFkKIbKpq1arUr1+f6dOnJ9q+a9cuChYsyP/+9z/gzamm1cKSJVCkiArDpA4dUoH5f//3+vOMGgWXL6tVna5ehdu31dSYeqGhKtxv34YFC2DAAFXJpuZYR0c4kOSWb+HCKuR79ID79+O3DxgAmzapcz98qGYGa9781fbqdDqWLNlGt27dUnxP+/bto2wqJ9qWQBZCiGxswoQJ/PTTT4SFhRm2PXjwgPz587/86TkpDeiaPl0Fnb09fP65ug9rbv7qfvfvq3uxFgkeZa5ZUx2bIwfs26fuI//2G8ycqZZbdHCAr75KPC2mpaWqtC0t4f331XUvXUrdsRER8fe08+aFY8cgOBj+/Vct3di5c/y+lSurbvQ8edSXuTn07//q+zpwIJa7dyP46KOPkv18duzYweLFi5mQ8DeD15BAFkKIbKxcuXK0bNmSKVOmGLblyZOHO3fuvPzpBSkF8rBhKuiePoXjx9X94S1bXt0vTx4VyrEJVm385x91bJ48EBcHYWHqPFWqqKB2dIRmzdT2hOdJGOr6JRlTc2xC9vZQtao6l4sL/PwzbN8ev6Zyu3ZQooT6+fFj8PSELl1ePc/ixTG0aVMNe3v7V147fPgwnTp1Ys2aNZQoUSL5hiQhgSyEENnc119/zbx587h9+zYAjRo14tatWxw/fhyw4k1LLmo0UK4c1KqVeNCXXo0a6j7xhg0pnyNvXlUtnz+vgjoiQnVPR6bi8ef/cqy+/aB+MQDV7d23rxoNbm+v7kdv3pz4mGfPYPXqGLp1a/TK+U6ePMkHH3zAwoULadTo1ddTIoEshBDZXLFixWjfvj2zZs0CoHjx4vTv35+OHTuyZ89lXrxQSx2uWAEJCulELl5U92iTu13q6Ajjxqlu3zVrVOUZF6eCT//UlZkZfPIJDBkC9+6pbbdvw7Ztb27/2x575Ijq6o6LgwcPYNAgqF9fDeACqFYN5s9XofvsmeoOr1Ah8TnWrQMnJw0NGlRNtP3cuXM0a9aMn376CR8fnzc3PuH7eKu9hRBCZEljx45N9EzyrFmzGDhwIAMG/Iqj42M8PVUIJcyYadNUBWlnB97eanBU377Jn3/ECPj+e3WMi4v66tsXpk5V95NBfV+sGLz3nhrh3LixCs7UeNOx9vawf7/6/to11aXt4KAqe2tr8POL33fhQvWscsGCUKCA2n/x4sTXW7wYPv7YCo3GNtH2GTNmEBYWRq9evbC3t8fe3j7Vg7pktSchhBCvcQmoikydmRw71HzWJdPkbBLIQgghXkOWXkyZDRBFWnU2S5e1EEKI1zADShm7ESaqFLIeshBCiAxUmzeNtM5+NECdND2jBLIQQog3qAG8+qxt9mYPvJemZ5RAFkIIkSIPDw927nyBupccb/9+KJnKsUzdu6sFG972OP381aYpDqiSpmeUQBZCCPEGBYDYRFvq1En9I0lpcVxqPHkCQ4eCh4d6FKtQIbWYxZEj6XG1WKB4mp5RAlkIIcQbmAEtMOXIiI6Ghg3Ves2bNqkpLwMCoEOH5Kfz/G/MgJak9edhup+uEEIIo9I/FXv06FHKlDmBk1McPXqoWbuSdicHBKjZrhwd1Wxdf/2V/DmTHufhoRapqFBBzZTVvr06f3JmzYIyZeDWrVdfW7pUbV+/Xk32YW6uquSPPopfyxjUjGJNmqhFKEqWhFWr4l/r3l2t9NSihZo0pHp1tR7yq8fGUbLkMVYlPDgNSCALIYRI5MWLF+h0OjQvJ3lesGAB27bt5cqVwly+DJMmJd4/JkbN4OXtraau/OkntXpSarumV62CrVvVOsRnzsCiRa/uM2GC2r53b3ygJ1xScedOaNpUhXBKoqJUoHbqpNq5YoWazvPChfh9VqxQ03w+fKhm/ho9Orlji7FixTr69+/PhYQH/0cSyEIIIQxu3brF7t27iYuL486dO8TExPDw4UNOnTpN3ryjGTHCItE0kwCHD6uFHEaNAisr1XXcsiWv7JeSQYPAzU1VrT4+ao5rPZ1O3Rfevh1274Z8+eJfS7ik4v374Ooa/9qpUyqwc+aMH0S2aZOqyHv0UCs9VaoEbdrA6tXxx7VqBf/7n3q9c+f4tsQfa4+FxSgqVapMmzZtWJ3w4P9IAlkIIQRXr15l79695MqVi9WrV9OuXTuqV6+Oubk5/fr1Y9++ffTu/TdFipgREpL42JAQcHdXizzoFS6sFnhIjYRBql9SUS8iQi3u8OWX8Ys/JCdPHjCsGAlUrKiOXbtW3V8Gtf7xkSPxSzQ6OsLy5RAa+ua2xB8biaPjFzg6OrJ8+XJCEx78H0kgCyFENvf8+XMCAgJwd3fHwcEBe3t7Dh48yKxZszAzM8Pd3Z1JkyaRP38x/PyscXNLfLybG9y8Gb98IcCNG2phhv/KyUlVpz16wMGDKe/XqJGqohOsj/EKd3eoVy9+icaICBW4v/zy5naoYzVERPQjIiKCiIgIIiMj+SU1B6eSBLIQQmRTT548ITIyEhsbG3x8fIiMjGTEiBGMGDGCfv36cf78eWJiYpg9ezZhYWEMGTKELVvcaNMm8XmqV1fV5LRp6n7ynj2wcaMa4ZwW6tdXlWzr1nD0aPL7dO0K+fOrLudz50CrVYPDjh+P36dlS7h8WQ0Ai4lRX8eOqQFpbxJ/bFFiYmKIiYnh2LFjBKTm4FSSQBZCiGwoMjKSffv24e/vz9WrV1m3bh0XL14kPDycCxcu0K9fPy5cuMCLFy/o1KkTjRs3pmjRonh5VadZs+ZotZaGc1lZqQDesgXy5lUDpZYsgVJpOAV2kyZqWUQfHzhxQm1LuKSijY26x1ymjBolrb93fOxY/EhqBwdVRa9Yoap6V1cYOTK+S/t1HBxs2b79c1as2I2bmxuurq6MHDmS6NQcnEqy2pMQQmQjCUdPb968meHDh/P48WOWL19OtWrV8PPz49ChQ0yaNIkjR46wceNGHj58yM2bN/nzzz9xdXWlWrXK7NgRRt68d4HsECFmQBngFGCerlcRQgiRDWi1WkMYA9ja2hITE4O3tzd169YlR44c1KhRA1dXV+bPn88HH3xAv379qFKlCn/++ScFCxYkKiqKIUOGkTfv36jlB7MDa2A16RnGIBWyEEJkCwkr41mzZuHp6UmdOnV4/vw533//PU5OTowcORKAnTt3smjRIj744APatWtnOMeLFy+wsrJKcNbJ6HST0WheM5Iq07MDRgNfpvuVpEIWQohsQKPRcO/ePfr27cuaNWvw9/enbdu2ODs7U6tWLS5fvszRo0eJejlMuVu3bjRr1sxwvE6nSxTGwcHBLFiQF42mCHFxWXVpRjOgCDAiw64mhBAii9HpdCTsAP33339p06YN9vb27Nu3jzlz5qDRaJg0aRI+Pj5UqlSJQYMGUaFCBWxsbGjSpAk5c+Y0nCNhVzeoQWGNGnkTG7sardY20SNPWUdOYB3p3VWtJ13WQgiRxcTFxWH2cpaOkJAQ3F4+ONy3b18ePXrE3LlzcXR0JCAggP/7v/9jxYoVVK5cmb///psyZcrgmnB2jNe4fv06M2fOpF274tSoMQozs6zUdW0P7AcqZtgVpUIWQogsRh/G33zzDR9++CHjxo1jwYIFTJ48mdDQUI4fP45Wq6V06dJ06dKFQYMGAdCwYUNcXV3RarWvnDM2Vi2/qNVquXv3Ll999RUFCxbEwcGBbdvuERu7kZgYq1eOy5xsga1kZBiDBLIQQmRJv/32G6GhoWzbto0nT57w888/kydPHtq2bcu8efO4ceMGAGPHjmXjxo2JjjU3f7WL1sLCwvCai4sL69ev5+DBg3Tr1o3w8HD++CMYS8tdPH9ukcm7r+2B7UCtDL+yBLIQQmRB165do0mTJowZM4YLFy6wcuVKAAYMGMDTp0/5448/DFWvk5NTslVxwjuaOp2Oxo0b06VLF+7du8fEiRPZuXMnxYsXp3r16pw8eZKjR62wsDhEXJw9Ol1mixczwBHVTZ3xYaxvgRBCiCxCH6LOzs60bt2aMmXKsHXrVkqUKMH27du5du0a8+fPp1+/foaqF5KvivUDuY4fP45Go6Fu3bqEh4czc+ZMzp49S+7cuYmKiuL999/H0tKSrVu38uJFGSwsTqLRlEGne81aiCbFDjXxxzEyups6IYs37yKEEMKU6EdQm5m9WlPpQ7Rt27bs27eP6OhodDod3377LX5+fixYsICiRYsCiQd/peTixYt88MEHbNu2jTJlyhATE8MHH3xA9+7dsba2pm7dulSpUoWePXvi7u6Ora0tUAw4hUYzDZgIPMc0Z/TSoCY3GY16tCljRlOn2BoZZS2EEJlHwhC9du0az549o2zZssnue/78eT777DMcHR0xNzfnhx9+oMA7LMG0cOFCDh48SKNGjfj+++/Zvn07O3fuZODAgTRs2JAVK1YY9k04AYkSALQDggBTGoVth3rGeBVQ2shtUaRCFkKITEQfxhMnTsTPz48ff/yR0qVLv1LpxsXFUbZsWXbs2MGjR4/InTs3oEZJJ+2eTm5bQj179sTGxoa7d+8SERHB0qVLGTx4MC4uLpQrVy7RvkmfV1ZhdwrQV8sxQOxbv++0YwFYYipVcUJSIQshhIlL2rU8d+5cNm7ciL+/v2GbPlQT7pu0Wk0ueBPuf/nyZUqUKJHodf05tFott27dol69elSvXp0lS5ZgbW2d4nmTFwjMABajuoszsmK2Q3WbdwO+ADwz8NqpI4EshBAmLGFg3rt3D2dnZzZt2oSfnx9ly5ZFq9Vy+fJlypUrZ5iLGhKH5OPHj8mZM2eK1zh9+jSDBw/G29ubwYMHY2f36mAsfTBfvXqVnDlz4uzs/B/eVRTgB0wBQoFnQHo8K2UG5ADyA6OADqhgNk0SyEIIYeJOnjzJqFGjcHd3p2nTphQpUoR9+/Zx7949KleuzIsXL9i+fTvTpk3D2dkZjUZjqIyHDx+Ok5MTI0aMwMLCgkePHpErVy7DuS9cuEDXrl2ZNGlSormr4dXKPOHPqa+KX0cHHAGmA5tQ3clmQCTvNghMg3qOOA7VLd4SGAZUf/maaZN7yEIIYUKePn36cqSyEhAQwJdffsm3337LlStXGDVqFD///DNDhw417DN9+nQiIiJwcnIyBOaRI0f46quvaN26Nf3790ej0aDT6Zg/fz5ly5alRIkS+Pv78+GHH+Ls7IxOp2P58uVcvXoVGxsbRo4cmajrG0gUzv89jEGF5HvAGlSIXgH+BQ4BB4CLqHu8+qjSvfzSEB+wsYAWKAXUBmoAVYDiZLYneyWQhRDCRDx69IjffvuNgQMH8uTJE0JCQrC1tcXLy4vr16/z/fffM2DAAJo3b87Tp08JDg5m8ODB2NvbM3fuXMM93alTp7Jz504WL15MwYIF0el0hor2vffeo3HjxhQrVoyRI0fi4uJClSpV+O6773j//ffJmTMnq1atolatWtSuXTtRJfzPP/+wbt06hg0bhouLSxq/ezOg5MuvTi+36UM6BNWt/RyIRq1PbIPqjnYjM4ZvciSQhRDCROTKlYvY2FhKly6Ni4sLU6ZMISYmhn379nHkyBH++OMPPD09iYqK4uzZsxQqVIgxY8ZQp04dIL4b2dvbmxEjRqDRaAzdzObm5oSEhPDo0SMaNmxIXFwcXbp0AdSIbb3g4GD27t1LwYIFgfhKeNy4cezfv5+lS5emQxinJGFIZ32Z/1cKIYTIxJJOWfn06VNiY2Np0aIFDRo0wMPDgwoVKtCiRQty5szJ9evX8fb2ZsOGDbi6uiYKY32XcqVKlQz3kPXbfH196devH87Ozvj7+3Pv3j3mzp1ruO7Dhw8ZP348LVu2pGnTpnh4eABw7tw5GjRogKurK7t27Xqn55hF6sigLiGEyGD6EcsJB0kdOHCA8uXLkytXLo4dO0bv3r1ZtmwZ5cuX58SJE6xbt44jR44QFhZG//79+eSTT1J9valTp3L06FEWLFiAo6MjAH///Td9+/bl6NGj3Lt3j2vXrhEbG0vlypUThe6mTZuoUqUK+fPnT9PPQLxKAlkIITJYTEwMlpaWAJw5c4bPPvsMDw8Pzp07x3fffUfDhg0ZN24cx48fx9/fn5CQEGxsbAyjp/WhmnQUdEpTag4ZMoQKFSrQo0cPYmJiMDc3x8zMjFGjRnH27FnOnTvHtGnTaN++PaCqbY1G88ZpNUXakkAWQogMNGLECA4cOMA///zD06dP6datG926daNly5aULVsWKysrtm7diqOjI40aNcLOzo7Hjx/zyy+/ULFiReDNc1CfOXOGU6dOUaVKFcqWLcvQoUPx8vKiY8eOWFmpNYvv3LlD/vz5OXPmDG5ubuTNmzcj3r54Dfn1RwghMoB+qcNp06YREhLC0qVLsbW15ddff8XFxYWKFSsyfPhwPD09GTp0KNbW1qxevZqWLVuydu1aQxgDyVbFeuPGjaNnz56Eh4czePBgDh48iI+PDxs3bmT16tWGNowdO5b79+9ToUIF8ubNm+zyiyJjSYUshBAZbMuWLXTp0oUbN25gZ2fHtGnTePz4MZMmTWLTpk188MEHbNq0iffff99wzJumvQwODubJkyf89ddfDB8+nDVr1jB69GhsbW05evQomzdv5q+//iIkJARra2vmz58v94VNjASyEEKkk4SDt2JiYujatSufffYZtWvXpmvXrjx48AB/f39mzZrFmTNn+PDDD9m4cSN58uQxzLCV8Dwp+e2331i7di2TJ0+mdOnS/PDDD2zdupWVK1cyZMgQcuXKxdy5c9HpdFy+fJmSJdVjRKlZflFkHPmTEEKIdKIP0ZiYGKytrSlSpAhTpkwBYMmSJZw8eZINGzbQsWNHPD09+fLLL6lWrRqTJ0/GycnJ0BWddIEIUGH65MkTRo8ejb+/P7/99huVK1fGysqKBw8e8O233+Lq6kqhQoVYvHgxf//9NxqNxhDGCR+TEqZB/jSEECINxcXFL5Lw9OlTpk6dyqpVqwCYMmUKz58/55tvvgFg/vz5dOrUiRw5cvDll19y6NAhevfubThP0qo4Li7O0G1tZmaGg4MDdnZ23L17l1u3bgHqXvWVK1fYt28fw4YN4+7du+zevZuGDRsmOlfaTH0p0pIEshBCpAGdTpeoC3jnzp3Y2tqi1Wo5f/48J0+eBNS807Nnz+bQoUO8//77tG/fnoMHD6LT6bC3tzcEenLVq5mZGWfOnKFRo0b4+vryww8/8NVXX1GqVCkCAgJ48OAB1tbW+Pr6Ymtry40bN/jmm2947733gMS/LAjTI/eQhRAiDZ0/f56tW7eyYMECZsyYQfXq1Zk0aRIeHh707NkTe3t7ateujUajYd++fa+9N/zixQssLCwwMzNDp9Nx6tQp+vXrx4QJE8iRIwcNGjTg0qVLXLx4kbVr19KxY0caNWqUbGUt3dOmT/6EhBAijaxatYpmzZphbm5O6dKlmT17NjExMXz44YecP3+eWbNmMWvWLMqUKcOQIUNeG8Y3btxg165dREdHExwczIMHDwgLC2PEiBHExsbyxRdfMHXqVDw9PWnRogUFChRgzZo13L59O9F5JIwzD6mQhRDiHSR9DEmr1TJ69Gjq1q3L+++/T1BQEEuWLOHhw4f88MMP/PPPP/zyyy9ERUXxyy+/vHGBhsePHzNs2DBCQ0M5c+YMO3fu5PTp03Ts2BEfHx++++47ihYtSlhYGMHBwRQrVowzZ85Qt27d9H7rIp3Ir01CCPGWdDqdIYxXrlxJUFAQ5ubmaDQa/Pz8AHB3d6dkyZLs27ePtWvXUrNmTebPn8/atWtxcXFJ9n5uwsk5cubMiY2NDceOHeO7776jWLFitGnThiZNmlCkSBGKFi1KQEAA7dq1Y9euXeTKlUvCOJOTClkIIVIpYffvlStX6NGjB4ULF6ZgwYLExsbSvXt3xo8fz8cff8yHH37Ili1b8PPzw9LSkmnTppEnTx7g1eo66eNN/v7+1K5dm5iYGBYuXEhkZCRt27alfPnyhIaG0rp1a4oVK8bZs2f59NNP6du3bwZ/EiI9yHrIQgjxBvrA1IfxvXv32Lt3L+PHj6d+/fo0aNCAcuXK4e7uTocOHRg4cCBbtmxh7969DBw4kFOnThkWk4DEjxwlDPnQ0FD69+/PxYsXqVmzJi1btqRfv34MGDCAo0ePUr58eVxdXfH39+fZs2dYWVkZ5qCWe8WZn/zpCSHEG+hXWbpz5w7vv/8+Bw8e5OjRo8yZM4e6devSvHlzfvnlF+zt7Wnbti1btmyhRYsWnDt3jgIFChASEpJovmmdTsejR48YNGgQoaGhAKxZs4bGjRvTpUsXLly4QIMGDfjzzz8JDQ2lR48eHDx4kDZt2lCxYkUeP35sWBDidY9JicxFKmQhhEiFxYsXs3btWpo0aUKrVq0AGD16NHPnzjXcu502bRr169enZs2a5M+fn5EjR3L48GF+//13cuXKZTiXRqMhV65cjBw5Ent7e7RaLfXr1+fixYs8fPgQgNq1a3Pr1i1mz57NjBkzcHV1ZevWrfz4448ULFjQcC4J4qxD/iSFECIVihUrxqFDh8iXLx8AJUqU4P/+7/+YMmUKO3fupEmTJhw+fJgyZcoAqltaX02XKFEi2XPmz5+fjz76iBkzZpA3b15mz57NxIkTAShcuDCNGjUiLCyMefPmUbp0aYYMGULBggVlZaYsSgZ1CSHESykt4qC/P/vVV19x9uxZNm7cCKg5qqdPn054eDjFixenT58+ifZ/Hf0+u3btYurUqUyePJmqVavSuHFjSpUqxc8//8zTp085ffo0ZcqUMVTYb1poQmReEshCCJECfWgmDMGqVavSu3dvPv3002SPedMyickFqq+vL7du3WL+/PlERETg7OzMvn37qF27tmEfCeKsT7qshRDZWtKaZN26dYwaNQqIvz+r0WgM3cQ//vgjI0aMICgoKNnzJbdog5mZGdeuXTOcS08/IGvs2LE8fPiQn3/+mbx58+Lv729YlUlPwjjrk0AWQmRrSYPO3t4enU5HbGxsou3m5ubodDpq1arF999/j6Oj42vPq9VqE4X9Z599xvLlyxPtY2ZmRlxcHFZWVvTr14/Nmzdz//59mjdvTr58+V75ZUFkbRLIQohsJzIyksePHxt+XrhwIVevXgXU8oVnz57FwuLVh1D0Adm7d2+cnJySPfeRI0cMM3k9efIEgOjoaKpWrYqbmxuQeNUlfRXerFkz1q5da3iuGKQqzm4kkIUQ2YpOp2PDhg2cOnWKQ4cOsXPnTsLCwhgwYAA7d+6kefPmaDQadu7c+cpxCQMyJCTklXP7+/szZ84c/v77b4KDg6lXrx4BAQFYW1tjZWXF/PnzgVcfVdJX4/pHoET2JIEshMgWQkNDmTZtmuHnPn360KlTJ8LCwhg5ciTdunVj7dq1dO3alUqVKhlCUqfTodVqDZODHD58mBo1ahAcHGw419q1a/n999/x9vbG09OTAwcOYGVlRY8ePZgxYwYzZsygW7duWFtbc+fOHcNxOp0OnU5nqMaXLVv2ympNIvuQQBZCZAvR0dF06NCBmJgYSpUqRb58+WjdurVhko9OnTrx888/Y2lpyfz589m8ebPhWP1ArTFjxvD111+zatUqatSoAahHn8LDw9m3bx9BQUG0a9eOsLAwtm3bxqBBgxg9ejQnTpygU6dOBAYG4uDgAJAo5I8cOULdunV59OgRhQoVyuBPRpgKCWQhRJb15MkTzp49C6iJNmJjYxk6dCiXLl1i7dq1PHz4kNWrVxvuJ5uZmfHjjz+yYMECgoKCePDgARqNhvDwcJo1a4abmxubN2/G3d3d0LVsaWmJt7c3xYsXZ/78+RQvXpw6depw/Phx/v77b4oUKcLvv/9Op06duH79OmvWrAHiQ97X1xdfX1/8/PwYMGCAET4lYSpk6kwhRJZ148YNNmzYwKFDh7hy5QotWrSgbNmy7N+/n6ZNm9KyZUv8/f2pVKkSVlZWvHjxgnLlylGsWDFcXFwMqzPlzp0bPz8/w0CuuLi4RI83FSpUiCZNmrBw4UL8/Pzo0KEDAQEB7N+/nyJFilCkSBH69OmDpaVlosFkGzdupFChQkycOFEGcAmpkIUQWYv+viyAs7Mz27ZtY8yYMRQoUID69evTsGFDnJycmDt3Lq1bt6ZQoUJ88803vPfee1y/fh2AuXPncuPGDZ49e2aohBOOqjYzM+P06dM0b96cCRMmsHLlSqpVq0bdunX5+++/CQwMpFOnToSEhPDXX3+h0+kwMzNj7969hIeHG87j4+NDnz59JIwFIBWyECILSTgj1osXL8iXLx/vv/8+xYsXJ2fOnACULFmSevXqsW7dOjZt2sS4ceM4c+YMP/74I87OzgD069cv0fzTSdcr3rBhA+PHj2fy5MmEhIQwY8YMwz3pq1evsnDhQiZPnkzjxo1xd3dHo9Fw9+5dzMzM6NKlS0Z+JCITkakzhRBZQsIwHj16NIGBgTRs2JA+ffqwY8cO/Pz86NixI02aNOHJkycsXLiQK1euMHnyZENYx8bGvvL8ccLzhoSE4ObmxokTJ/Dw8ODff//liy++oEaNGgQGBvLnn38SFBTE1KlT6dOnDw0aNMjYD0FkahLIQogsIzg4mA0bNnDs2DE6dOjA999/T4sWLRg6dChTp04lJCSEDz/8kKtXr1KxYkXKlCmDnZ1dqs49adIkduzYwdy5cylVqhQ3b96kb9++fPvttxQpUoQGDRpQs2ZNfv75Z65du0bRokXT+d2KrEbuIQshMiWdTpdoxiuAESNGsHz5coYNG0aLFi2YOXMmP//8M1euXKFHjx7Y2trSvXt3bGxsqFatGnZ2dslOT/ntt98SGhpquM60adP4999/8fPzo3Tp0mg0Gi5evIiNjQ0VK1Y0BLytrS0vXrwwhLHUO+JtSIUshMjULly4wO3bt6lfvz737t2jY8eOjBo1iiZNmmBpacmUKVNYt24dR44cAeDRo0eGpQyT0s/GtWnTJlq2bAmoLus2bdrQv39/mjRpQnR0NNbW1sTGxlKnTh08PDzYs2cP8+bNMxwjxLuQClkIkWl9/fXXdOvWjWPHjvHBBx/w/PlzunbtypIlSwwV7qhRo3B0dOTEiRPodDpy5cr1SmUNcO3aNcOgrZYtWzJw4ECmTJmCmZkZ5cqV4+TJk2i1WqytrQEV1H///Tf9+/fnxIkTiQJciHchgSyEyJRCQ0N59OgRx44dw9nZmatXr3Lr1i169+6NmZkZv//+O5GRkQBs3ryZypUrGwI36VzSz549Y9KkSezdu5epU6fy559/0qFDBxYtWsTNmzepVKkSd+/eZd68eQBMnTqVbt26Gark/PnzGx6PSnpuIVJL/s8RQpi0lO6qXb16laNHj9K9e3fWrFnDhg0bqFevHjExMYwePZp9+/YZVlvSL52YVFxcHDqdjhw5clC6dGmaNGnCsWPHqFOnDrVr16Zp06aMHDmS5s2b06hRIxYtWkSzZs3Yu3cvM2fONEyDqb+GEP+FPIcshDBp+qr2+fPn2NjYsHfvXsqXL0/t2rUxNzcnKiqKrVu3AnD8+HF27NjBl19+ycaNG8mRI8cr59FL+DgTxD+fXKhQIcPzyD/++CNVqlRh+fLl9O7dm/fee4+HDx/i6emZ7DmE+C/k/yQhhMk5deoUK1euNPw8f/58li5dCsCOHTu4e/cuoO4PX7hwgTVr1jB+/Hi6dOlCkSJFAMiRI8dr7+eamZlx6tQpWrduzaxZs8iRIwc7duxg06ZNbNq0ybDfzJkzGTp0KKGhoeTOndsQxlqtVsJYpCn5v0kIYXLCw8MZOHAgQUFBABw5csQwc9akSZMoXbo0AM2bN2fs2LEEBgZy8+ZNdu/eTYcOHQzneV1g/vPPP/Ts2ZM+ffoQHR3NqFGjCAsLw9fXl7Fjx/Lo0SOmTp1qmPva1dU10fHSRS3Smjz2JIQwCQn/KdJoNAwdOpRr167x559/0qxZMxYuXIi7uzsxMTFYWloC8PTpU2xtbROdR6vVvhKWyW1btmwZtra2uLm50adPHz799FP69+8PwNChQwkICCA6OprVq1cbFpnQPxYlRHqQQBZCGF3Ce7EPHz40LOTg5eXF//73P8zNzenQoQP169c3HDNz5kysra0NIZr0PHoJQ9Tf3x93d3cqVKjA6tWr6dmzJ7Vq1WL27Nl4enry4MEDwsPD8fT05Pbt27i7u6fzOxcingSyEMKoEs4fPWnSJLZt20arVq0YOnQop0+fpmHDhtSsWRMnJyciIyNxdHTE29ubdu3apfoe7vXr1xkwYABPnz7l5s2bLFiwgDx58vDDDz9Qvnx5Bg8ezJkzZxg4cCCffPIJH3/8seHY5KprIdKD3EMWQmQofQ0wZcoUACwsLLhz5w79+/cnPDyc8ePHs3PnTmbPno2XlxdDhgzBysqKxYsXM336dOrUqUOTJk0MYZx04FbSGuPYsWN89tlnNGzYkN27dzNo0CAWL17Mixcv6NWrF7/++itdu3alQ4cOdOnSJVEYg9wrFhlHHnsSIkuKA64At4FnwHPgBWAF2AA5gAJAcTL69/K4uDjMzc1xcnIydCfv2rWLZcuWsXv3bqpUqUJkZCSbNm1i165d+Pr6UrhwYebOnUu/fv0SzROt0WgSVckJu6z1j0lZW1vz4sULLl++DMCgQYPo2bMnW7ZsYcSIERw4cIC7d++SO3duXFxcXjmPEBlFuqyFyPT04XscOAQcAC4CloC+utO9/NK8/ALQAjFAKaA2UAOoSnqG9B9//MGGDRtYuXIlcXFx+Pj4MHr0aGrWrEm3bt2wsbHh119/BdQCDyEhIfj6+vLo0SOcnJzIly9fqq7j6+tLUFAQhQsXZty4cWzdupUtW7bQtm1bGjVqREBAAL169eLLL7+kZcuWhnvM0j0tjEl+BRQiU9KhwrcNYAtUAfoBc4DTQDQQCTx6+fUYePLyv/ptkS/3O/3yuH4vz2P78ryHXl4n7fj4+HDixAnWrl2LmZkZ7733Hr179wbUSk3379/Hz88PgPbt2wNw//59SpQoQb58+ZJ9rli/TavV8uLFC/r168eTJ09YvHgxCxYs4Ntvv6VixYoUL16cTZs2ERYWRunSpfH19aVOnTqJRk1LGAtjkgpZiEwlClgOTANCUd3R6bGYgRmqWzs/MBLoCKRu3eCk9FWnvht41apVjB8/noMHD+Lk5EStWrWoUaMG06dPZ/ny5SxevJh58+ZRuHBhw8pKydH/06UPVH0X9vTp06lVqxZ//PEHly5dMpzr1KlTzJo1Cy8vLwYPHpzoPPIokzAFEshCZAqBwHRgCSosIzPw2vao0O8KDAM8U31kwnuxJ0+epFy5clhaWtKrVy+io6NZtmwZt2/fpnz58uzYsYMqVaowYsQIWrVqRY0aNYDkAzPhtjt37vD999+TL18+Pv/8c1q1asW1a9fo27cvn3/+OQC7du2iUaNGHDp0iDJlyqS4/KIQxiSBLIRJ0wJTgUmo+72xRmyLBeq+tC+qak65ezfh5B0BAQH07NkTBwcHcuXKRZ8+fahXrx6VKlVizJgxdOjQgalTpzJt2jQePHjwVvdxv/rqK44ePYqVlRVRUVEsXryYffv2sXTpUpYtW4aLiwtTpkxh1apVrF27Fg8PD0CqYmGaJJCFMFkBQDsgCNVVbSrsgCLAKqB0oleio6P54osv+OyzzyhZsiRarZbhw4dTrlw5evbsyezZszl9+jT9+/fn3r17DBkyhH379pEnTx42btyIj49Pqltx8eJFPv/8c/z9/bl8+TIrV67k+vXrLFq0iFGjRhEUFERoaChOTk7MmTMHNze3NP0UhEhrMqhLCJOjBSajBlidx7TCGFR7zqPaNxnVXlUVW1tbM3bsWFxcXAgNDcXc3JxDhw4Zuojbtm2Lp6cne/bswdvbmyJFirBo0SKAFMP4dcsv6q9RunRpOnTowJ07d1i6dClTpkxh3rx5zJgxg/Xr1+Pm5mZYr1gIUyWBLIRJuQpURAXdM9J6lHPa0aHaN5mYmHLAVUMXtbOzM59//jnfffcd0dHRdO/enX/++YcnT57g7OyMpaUlt27dAmD9+vV88cUXr71SwuUXAfbu3Ut4eDjNmzfHw8ODJUuWAFCqVCkcHBxYuXIlZ8+eJWfOnFStWhWQx5lE5iCBLITJOAlUAy5gelVxSqIwM7uITleV0NCtNGvWjEmTJuHl5cW5c+f4559/KFeuHLGxsfTr148rV66wZcsWw+QeCZ//TehNyy/eu3cPgE8++YRvv/2WXbt2MWXKFMzNzXF2dubq1auJzidhLDIDCWQhTMIBoC4QQfo8xpR+zM1Bp3uEo6MPn31WGSsrK27dukVISAirV6/G2dmZ0aNH4+joyOjRo/Hx8TEsCKEPyqSB+ablF0uVKoW5uTktWrRgxIgRrFu3josXL/LHH38AYGf3bo9oCWFMMqhLCKM7ADQFnhq7IWnAlmfPNnD6tD2zZ8/m9u3b1KpVi3HjxmFhYZFo9HXS6Sn/6/KLYWFhjBw5kpCQEBYtWvTK+sVCmDqpkIUwqpNAc7JGGAM8JUeOVhQoEIaDgwNDhw7l8OHDhISEAGBpaWkI3qRzUGs0GjQaDREREQB8//33BAUF8emnn+Lp6UlgYKDhHKCWX/z9998BNaBs5cqVeHl5sXXrVgljkSlJhSyE0VxFzR39yNgNSXNarQOtWhVk8eKD2NnZYWVlleK+abX84osXL157HSFMnVTIQhhFLPAhan7prOgJixY9wskpJ1ZWVom6o9Nr+UUJY5HZyfKLQhjFNOA6mW0AV2qZm0Pu3I9Q7/PLRLNipefyi0JkZtJlLUSGC0BNqvHM2A3JADmAf9HP6JVRyy8KkRnJr5ZCZKhYoC3w3NgNySDPUdN/queM02P5RSGyCqmQhchQk4FvyDqjqt9Mp7NDoxlNXNzINF1+UYisRgJZiAwTCJQne3RVJ5WD8+dXUKJE8zRZflGIrEi6rIXIMNNRSyhmPzExzwgMHECnTp3YsWMHv/zyCydPnmTFihUUKFCAkSNH4u3tDah7x/owBiSMRbYhFbIQGSIKcCY7dVW/ypZff53Iv/9eTJPlF4XIaqRCFiJD/AFk90pPQ8eOmrdeflGI7EICWYh0pwOmYswVnF68gI8+Ag8P0Ghgzx5jtCKKnDnnYGlp8VbLLwqRXUggC5HuDgOhxm4EtWvDsmVgzGme4+Juc/36yjcuvyhEdiT3kIVId22AdahKOX15eMDAgbBkCQQHQ7NmsHgx2NjE71OwoArm+vXTvTmv0Grh2jUvihc/lfEXF8LESYUsRLqKA/zJiDDWW7UKtm6FoCA4cwZe3qI1CebmULz4RSBOJvkQIgkJZCHS1RUyesr4QYPAzQ1y5wYfHzh1KkMv/0Y6nQVwReagFiIJ+RshRLo6Tkb/NUt4j9jWFiIjM/Tyb6TRmKHmtxZCJCSBLES6OgSYWCIaXSTqcxFCJCSBLES6OkBG3j9+nehoeP5yTYsXL9T3xhnSqUN9LkKIhCSQhUg3ccBFYzfCoGRJyJEDbt+Gpk3V98HBxmqNGtglhIgnjz0JkW4uAVWRLuvk2KHuI5c0dkOEMBlSIQuRbm4D5sZuhImyAEKM3QghTIoEshDpJjsus/g25PMRIiEJZCHSzXNMZUCX6dGhPh8hhJ4EshDp5gUSyCnRAdHGboQQJkUCWYh04OHhwc6dF0m65OL+/Wq0c2p07w6+vm9/3J49ar5q06YBrI3dCCFMigSyEOnGiqSBXKcOXLr09md61+NS48kTGDpULUxhZweFCqmlGo8cSZ/rKRrA5o17CZGdSCALkW6sjN2AN4qOhoYN4exZ2LQJHj+GgADo0AG2bEnvq+dI7wsIkalIIAuRTo4du0OZMo9xcoIePdTMWEm7kwMC1DKIjo5Qtiz89Vfy50p6nIcHTJ8OFSpArlzQvn38LFxJzZoFZcrArVuvvrZ0qdq+fj2UK6dWY7KzUxXy+PHx+128CE2aqAUrSpZUK0rpde8OAwZAixbg4ADVq0Ng4JuOjQXcUvjkhMieJJCFSCfLl29n2zZLAgPh8mWYNCnx6zExajUmb2+4dw9++gk6d05913RqllmcMEFt37s3PtAdHeHAy5krd+5Us3bZ2aV8nagoFaidOql2rlgB/fvDhQvx+6xYAePGwcOHUKwYjB79pmNjgOKpe6NCZBMSyEKkk4EDB+LuXprcuVVA+fklfv3wYbUS06hRYGWluo5btnx1v5S8bplFnU7dF96+HXbvhnz54l+LiIDatdX39+8nXh3q1CkV2Dlzxg8i27RJVeQ9eoCFBVSqBG3awOrV8ce1agX/+596vXPn+LakfGxu5J8fIRLL2IVahchG3N3dgdrAGQoX1hGSZGKqkBBwd4eEywIXLqzmmk6NpMssJjx/RAT89husXKm6tFOSJw/cuRP/c8WK6tidO6F3b7UtOFgN8HJ0jN8vNhY+/jjltuiXfEz52Lypeo9CZCfyK6oQ6eTmzZtADcCeGzdUNZuQmxvcvAlxCdZYuHEDChT479d2clLVaY8ecPBgyvs1aqSq6KiolPdxd4d69VRQ678iI+GXX97cjuSPdeCXX758m7cjRLYggSxEOpk9eza3bhUkPFzLN9+ogVcJVa+uqslp09T95D17YONGNcI5LdSvD8uXQ+vWcPRo8vt07Qr586su53PnQKtVg8OOH4/fp2VLdQ986VLVzpgYOHZMDUh7k+SPjSEgwDEt3qIQWYoEshDppFOnTnh796No0ad4esZP8qFnZaUCeMsWyJtXDXZasgRKlUq7NjRpAgsXqnvMJ06obfb2aqIRABsbdY+5TBk1Slp/7/jYsfiR1A4OqopesUJV9a6uMHKkemTqTZI/NproaNc3HyxENiPLLwqR7toA65H1f0HVAK2ANcZuiBAmRypkIdLdMGQSDL0cqM9DCJGUVMhCpDsdUAK4auyGmIDiwCWSTikqhJAKWYgMoAFGAPbGboiR2QMjkTAWInlSIQuRIaIAZ+CpsRtiRLZA2Mv/CiGSkgpZiAxhB3Ql+87FYwF0Q8JYiJRJhSxEhglEpyuPRvPM2A0xghzAWcDT2A0RwmRJhSxEBrl0KZZff81rmFYy+7AFxiBhLMTrSSALkc60Wi0zZsygYsWKDBhwk6AgNSNW9mAGFEUNahNCvI50WQuRji5dukSPHj04dOiQYVu5cuacOKHB0jLWiC3LKDmAE0AaTj8mRBYlFbIQ6SBhVZwwjCtWrMiyZf9iafk1aqBXVmaH6qqWMBYiNaRCFiKNJVcVW1hY4Ovry1dffYWlpSWgBSoCF8iaU2qaAWWAU4C5cZsiRCYhgSxEGtFqtfzwww/4+vry/Plzw/aKFSuyaNEivLy8khxxFagGRGRgKzOKI3AMKGbkdgiReUggC5EGUlcVJ+ckUBfISkOv7YH9qB4AIURqSSAL8R+8fVWcnIOAN1ljFi9bYDtQy9gNESLTkUAW4h29e1WcnANAczJ3pWwPbEXCWIh3I4EsxFtKm6o4OaeABsBjMtdALzMgJ7Ab6aYW4t1JIAvxFtK2Kk7OVaAVEIRakMLU2QFFgHXIAC4h/ht5DlmIVHjdc8XHjx9n3LhxaRDGoELtFDAaNamGqS5VqEG1bzSqvRLGQvxXUiEL8QbpXxWnJABoh+lVy/qqeBVQ2shtESLrkApZiBRkXFWcktIkrpaNvXSjBYmrYgljIdKSVMhCJEOn03H16lVKlSpFXJwaYJUxVXFKAoEZwGJUd3FGVsx2gA61nvEXyKpNQqQPqZBFtqYP26Q0Gg358+dn6NChQEZWxSnxBOYA94AfXv5sR/r9FTZ7ef5iwI8vrzsHCWMh0o9UyCJb0ul06HQ6zMxeH2hPnz5lwYIFfPrpp0YK4pTogCPAdGATqjvZDPUc87v8ldagniOOA2KBlsAwoDqmO7BMiKxFAllka2fPnmXixInUqFEDT09PPvjgA+Li4gxBrdPp0GhMPZDigCvAv8Ah1CQjF1GLOujvO+tefmmID9hY1CIXpYDaQA2gClAc6TwTIuMZe5SIEBkmOjoaa2trw88HDhygT58+DBgwABsbGwYOHIibmxtVq1Y1hLLphzGo8Cz58qvTy236kA4BngHPgWjAGrBBDc5yQ8JXCNMhgSyyhVu3bnH27Flq1arFs2fPiIuLIywsjE6dOjFgwAAAYmJi6NOnDydOnHhjV7bpSxjSQojMILP/qyNEqlhYWHDgwAHatWtHvXr1ePbsGcHBwezduxdQjzj17t2b/PnzExAQYOTWCiGyIwlkkS24urpy584dzp8/z/DhwylatCiff/45gYGBzJ49G3Nzc06cOIFGo6Fw4cLGbq4QIhuSQV0iy0o6ICs0NJTVq1fz4MEDWrRoQbVq1Th69Chffvklrq6unDlzhi+++ILu3bsbr9FCiGxLAllkKUuWLCEuLo7u3bsnCmT99xcvXmTevHkUKlSIwYMHAxAcHExERASurq64uLgYs/lCiGxMuqxFlvH8+XPMzc1ZtWoVN2/eRKPRGCb+0AdzqVKlqFmzJkFBQTRr1oyGDRui0Wjw8vLCxcUFrVZrzLcghMjGpEIWWcrp06f5+eefcXR05Lvvvkv0mr5K1mq1BAQEsH79erp3707BggWN1FohhIgngSyyjLFjx7Jt2zaKFSvG5cuXGT58OO3atUs00UdytFot5ubmGdhSIYR4lTyHLDIlfYjqq97Q0FBOnDjBnj17iI6OZs2aNSxdupQaNWrg7u5uCGX9758JB3tl/meOhRBZgfxLJDKduLg4Q0UbEhICQGRkJCdOnECn0+Ho6EjTpk1xcHBgzpw5gArd2NhYNBoNGo2Go0ePMmHCBLRabSaZjUsIkdVJIItMx8zMjJMnT9K8eXNGjx7N5MmTMTc3p0uXLkyZMgUAd3d3rK2t2bp1K/v27QPU5CCgurZHjRpF7969pataCGEypMtamLxnz56RI0cOw89hYWH4+voyfvx4njx5wpAhQ8iXLx8dOnSgV69eODk5ER0dTXh4OL/99hvVqlUD1ICvvn370r17d3bt2iWVsRDCpEiFLExaREQEM2fOJC4ujsuXL3Pw4EGio6Nxdnbm/v37jBo1iu7du/PJJ59QuXJl5s+fT1RUFGfPnmXOnDmGMI6KiuLff//lzz//5NNPP5UwFkKYHBllLUze0KFD2bBhAw4ODvj6+vL+++9Tu3ZtdDodW7ZswdXVFYC1a9fSsmVLrKysDMfGxcUZ7hsLIYQpkwpZmJykk3OUL1+eZ8+e8fHHH/PRRx+h0+no2bMnJUqUQKPRcP78eerUqcM///yTKHgz1xKKQojsTipkYVISPjPs7+9PmTJlKFKkCIcPH6Zbt25s3rwZT09PoqKimDx5MufPnyc4OJiBAwfSq1cvI7deCCHenQSyMAn37t3D2dkZgPPnz9O3b1+KFClCWFgY3bt3p0OHDgwbNozAwEDWrVtHYGAgBQoUwMbGhsePH5MzZ06AN04CIoQQpkoCWRjd119/zY0bN5gxYwY2NjYMHToUHx8fmjdvToUKFciXLx++vr40aNCAWrVq4eHhwe7du1m1ahW1a9cGJIiFEJmfBLIwGn2I3rt3j48//pju3bvTsWNHIiMjCQ4OpnPnznTp0oXw8HCuX7/OnDlziImJ4eDBg1SvXp38+fMb+y0IIUSakZJCZDidTpdoKktnZ2d69erFokWLCAgIwN7engMHDtCoUSOGDRtG06ZN2bt3LytWrMDJyYkPP/yQ/Pnzy8pMQogsRQJZZDiNRoOZmRnnz5/nu+++4/Lly7Rr1w4vLy/mzJlDdHQ0Hh4ebNiwgc2bNzNr1ix69OhB586dDbNtATLLlhAiS5Eua5Ehki7qMGPGDJYvX0779u05e/YsNWrU4OOPP6Z169b07duX5s2bs2jRItavX89HH33Ep59+ajiPPMYkhMiKpEIW6U6/gINGo+HmzZsEBQXx6NEjTpw4QbFixThx4gT58uUjZ86c9O/fnzlz5hgeZdq0aZMhjPWTfAghRFYkgSzSnb5red26dXTp0oXHjx+zbds2ateuzeLFi1m5ciXt2rXj5s2btG7dmkqVKvH48WMAbGxsiIuLA2SZRCFE1iZd1iLdabVa2rRpg4WFBV999RWVK1dm/PjxHDx4kB07dgBw/PhxFixYwIQJE8iXL5+RWyyEEBlPSg6RppL7/c7c3Bxvb29OnDhhmMCjefPmuLi40KFDB3x9fenVqxeVK1eWMBZCZFtSIYt08fz5c2xsbNizZw8VKlQgd+7ctGjRgvLlyxvWLI6IiGDz5s0EBgbSq1cv3NzcjNxqIYQwHglk8Z+dOnWKS5cu0b59ewDmz5+PTqfjk08+wdfXlw4dOlCuXDlu3LhB48aNWbBgAXXq1HnlPFqtVh5lEkJkW9JlLf6z8PBwBg4cSFBQEABHjhyhRIkSAEyaNIly5coRFxdHoUKF6N+/Pz169CA2NjbROXQ6nYSxECJbk0AW70Sn0xm+GjZsyMcff8yQIUPQarVcv36dokWLAhATEwOoEdLPnj3j888/Z/bs2Ykm+ADkcSYhRLYnXdbirSVcyOHhw4c4OTkB4OXlxf/+9z/Mzc3p0KED9evXNxwzc+ZMLCws+Oyzz4zRZCGEMHkSyOKtxMbGGqrbSZMmsW3bNlq1asXQoUM5ffo0DRs2pGbNmjg5OREZGYmjoyPe3t60a9dOniMWQojXkH8hxWvpf1/Tj4y2sLDgzp079O/fn/DwcMaPH8/OnTuZPXs2Xl5eDBkyBCsrKxYvXsz06dOpU6cOTZo0MYSxfpIPIYQQiVm8eReRncXFxWFubo6Tk5NhHuldu3axbNkydu/eTZUqVYiMjGTTpk3s2rULX19fChcuzNy5c+nXr5/hXrL+WKmShRAiedJlLVL0xx9/sGHDBlauXElcXBw+Pj6MHj2amjVr0q1bN2xsbPj1118B+PbbbwkJCcHX15dHjx7h5OQkk3wIIcRbkHJFpMjHx4cTJ06wdu1azMzMeO+99+jduzcAI0aM4P79+/j5+QEYnkG+f/8+JUqUIF++fNI9LYQQb0EqZJGIfnIO/UjqVatWGeaddnJyolatWtSoUYPp06ezfPlyFi9ezLx58yhcuDDR0dFYW1sb+y0IIUSmJBWyMNDfLwY4ffo0MTExtGvXjho1ahgeV1q1ahULFy7k33//pXPnzlSsWJGQkBAArK2tk53LWgghxJtJhSyIiYnB0tISgICAAHr27ImDgwO5cuWiT58+1KtXj0qVKjFmzBg6dOjA1KlTmTZtGg8ePJDpLoUQIo1IhZyNRUdHM3DgQK5duwao7up58+bxySefsH37durXr8/q1au5cOECM2fOZOLEiTx48ICRI0eyaNEiAAljIYRIIxLI2VRMTAzW1taMHTsWFxcXQkNDMTc359ChQ+TKlQuAtm3b4unpyZ49e/D29qZIkSKGIPbx8TFi64UQIuuR55Czmbt37+Li4mLoonZ2dqZ79+7kyZOHyZMn0717d/755x+8vb1xdnbG0tKSW7duAbB+/fpX5qAWQgiRNqRCzkYuX75MzZo1iYyM5ObNmzRt2pSJEyfi5eXFuXPnOHjwIOXLlyc2NpZPP/2Uy5cvs2XLFsPkHvoFILRarTHfhhBCZEkyqCub+eKLLzh79izOzs40atSIsLAw7t69y/bt26lTpw5DhgwhV65cTJgwgXv37lG7dm0GDRpk7GYLIUSWJ4GczYSHh/N///d/ODs78+effxIVFcXZs2eZPXs2t2/fplatWowbNw4LC4tEo68TrvAkhBAi7cm/sNlM7ty5GT58OPv37ycsLAw7Ozvy5cuHg4MDQ4cO5fDhw4bnii0tLQ3PFUsYCyFE+pIKOZE44ApwG3gGPAdeAFaADZADKAAUJzP9LhMXF4dWqzVUuwCff/45Fy9eZOvWrQQEBNCpUyf+/vtv7OzssLKyMmJrhRAie8rGgawP3+PAIeAAcBGwBPTP1upefmlefgFogRigFFAbqAFUxVRD+tKlSwwYMIA//vgDZ2dnw/aIiAi8vLxwdXWlQIECNGvWjD59+gDxKzMJIYTIONkskHXAYWA64I966ssMiHz52tvSAPaocI8FWgDDgPeID3Dj0Gq1/PDDD/j6+vL8+XMaN27M+vXrsbOzM+zj7++Pv78/kyZNInfu3EZsrRBCiGwSyFHAcmAaEIrqjk6PlYjMUN3a+YGRQEfA7rVHpIdLly7Ro0cPDh06ZNhmYWHB8ePHqVChQrLVrwzaEkII48ri/wIHAv0AZ+CLlz9HkT5hzMvzRgFXgSEvr9vv5XXTn1arZcaMGVSsWDFRGFesWJHjx4/j5eUlYSyEECYqi1bIWmAqMAl1vzfWiG2xQN2X9kVVzekz93NKVfGYMWP48ssvEw3oEkIIYXqyYCAHAO2AIFS1airsgCLAKqB0mp016b1ivYoVK7Jo0SK8vLzS7FpCCCHSTxbqp9QCk4EqwHlMK4xBtec8qn2TUe39by5dukSdOnUYNmyYIYwtLCz4+uuvOXr0qISxEEJkIlmkQr4KtML0quKU6KvldUCxtz5aqmIhhMh6skAgnwQaAo9Jv8Fa6cEMyAnsBiqm+ii5VyyEEFlTJg/kA0Bz1HPEmZU9sBWo9dq9pCoWQoisLRMH8gGgKfDU2A1JA7bAdlIKZamKhRAi68ukgXwSqEvmroyTsgf2k7D7WqpiIYTIPjJhIF9FzR39yNgNSQeOwDGgmFTFQgiRzWSyQI5FVZABZK4BXKllhk5XmpkzuzF69FipioUQIhvJZM8hTwOukzXDGCCO588DuHdvhDxXLIQQ2UwmqpADUJNqPDN2Q9Ld06dQpQrY2EhVLIQQ2UUmCWR9V/UF3m2ZxMxFq4X7953JnTsYS0sbYzdHCCFEBsgkXdbTULNwZf0wBjA3BxeXKCwtZxq7KUIIITJIJqiQA4HyZIeu6lflAM4CnsZuiBBCiHSWCSrk6aglFLOjGGCGsRshhBAiA5h4hRwFOJM1ZuN6V7bAPdSCFEIIIbIqE6+Q/wA0xm6EkWmAFcZuhBBCiHRmwoGsA6Zi7OUUDx+GJk0gd27Ilw/atoU7dzKyBVHAFLLLgDYhhMiuTDiQDwOhxm4EDx9Cnz5w/ToEB4ODA/TokdGtuAMcyeiLCiGEyEAmfA+5DbCOjKoMPTxg4EBYskQFb7NmsHgx2CR5DPjECahXD548yZBmvWQGtALWZORFhRBCZCATrZDjAH8yupt21SrYuhWCguDMGVi06NV99u2DsmUztFmoz2MTWXfKUCGEEBbGbkDyrqCaFp2hVx00CNzc1Pc+PnDqVOLXz5yBCRNgw4YMbdZLFqjPpaQxLi6EECKdmWiFfBxjNM3VNf57W1uITLDc8tWr0Lw5/Pgj1KmT4U1DfR7/GuPCQgghMoCJBvIhIPKNe2WU4GBo3BjGjIGPPzZWKyJRn4sQQoisyEQD+QCm8pjP7dvQsKEa8PXpp8ZsiQ71uQghhMiKTDCQ44CLxm6Ewfz5cO0ajB8P9vbxX8ZxERnYJYQQWZMJPvZ0CaiKKXVZmw471H1kGdglhBBZjQlWyLcBc2M3wkRZACHGboQQQoh0YIKBnB2XWXwb8vkIIURWZIKB/BxTGdBlenSoz0cIIURWY4KB/AIJ5JToyOjJUoQQQmSMNwayh4cHO3fufGX7/v37KVkydYOLunfvjq+vbyqPs0K/5OKePVCwYKoukaHGj4cuXdT316+DRgOxsWl3/pTftwawTpNrLFq0iNq1a6fJuYQQQvx371wh16lTh0uXLqXDcTa86xrIkydDkSLqsaSCBaF9+3c6zX9Sv74K6NOnE29v1Upt37Pnv5xdg/p8hBBCZDUm2GWd452OWrwYli6FnTvVlJfHj0OjRmnctFQqUUKtGqX34AEcOqTWU/7v3u3zEUIIYdpSFcjHjh2jTJkyODk50aNHD54/f86ePXsomKBfNSAggPr16+Po6EjZsmX566+/kj1X0uM8PDyYPn06FSpUIFeuXLRvP53nz5Pv/501C8qUgVu3kmsjNG0Knp7qZ1dXtY6xXv364OsLNWuqCtrHRwVl586QMydUq6a6n/UGDwZ3d/ValSqwf39qPimlc2dYuRK0WvWzn5+qkK2s4veJjobPP1eLWbi5qe+jU7g9HBICbdpAvnyPKFLkY2bNmmV4TavVMnnyZDw9PXFwcKBKlSrcvHmT69evo9FoiE3Ql16/fn3mz5+f7DUuXrxIkyZNyJ07NyVLlmTVqlWpf8NCCCH+s1QF8vLly9m2bRuBgYFcvnyZSZMmJXo9JiYGHx8fvL29uXfvHj/99BOdO3dOdZf2qlWr2Lp1K0FBQZw5E8SiRa8m04QJajnEvXvj7686OsKBl7NJvveeqkq/+05Vx/owTGjFClVF374NgYFQowb06AHh4VC6NHz9dfy+1aqp1Z7Cw6FTJ2jbFp6ncoCzm5v6xWH7dvXzkiXQtWvifb75Bg4fVtc4fRqOHoUkHysAcXHqlwcvL7h925pdu/bwww8/sG3bNgC+//57/Pz82Lx5M48fP2bhwoXY2tqmrqEvRUVF0aRJEzp16sS9e/dYsWIF/fv358KFC291HiGEEO8uVYE8cOBA3N3dyZ07N6NHj8bPzy/R64cPHyYyMpJRo0ZhZWVFw4YNadmy5Sv7pWTQoEG4ubmRO3dufHx8OHUql+E1nQ6GDlXhtnt34m7fiAjQj0vq0gV++gm2bYN69cDZGaZOTXydHj1UBZ0rl1q5ydNTLRphYaEC9+TJ+H27dIE8edRrX3yhqte3uWXetasK4osXVTtr1Ej8+vLlMHasame+fDBunPplIaljxyAsTO1rZVWaokWL8cknn7BixQoA5s+fz6RJkyhZsiQajQYvLy/y5MmT+oYCmzZtwsPDgx49emBhYUGlSpVo06YNq1evfqvzCCGEeHepWg/Z3d3d8H3hwoUJCUk8W1RISAju7u6YmZkl2u/27dupaoRrgnUPbW1tCQnJA4QDOiIi4LffVBdwrlwpnUHp3Fl9xcTA+vXq+4oVVVc2gItL/L45crz6c8LlFqdPhwULVHexRgOPH8P9+6l6OwC0bq2CPE+e5FeICgmBwoXjfy5cWG1LKjhYbXd0BLgAOKLVaqnzcg3Imzdv4qnvp39HwcHBHDlyBEd1EQBiY2P52HhLWwkhRLaTqgr55s2bhu9v3LiBm5tbotfd3Ny4efMmcXFxifYrUKDAOzYrL6BWcHBygk2bVHV78GDqjra0VBVvhQpw7tzbX33/fpg2DVatgocPVYWbK5eq1lPL1lZV4b/8knwgu7mpsNW7cUNtS8rdXY0cj4hwICLidyIiInjy5AmbN29++bo7gYGBrxxnZ2cHwNOnTw3bQkNDk22ru7s79erVIyIiwvAVGRnJL7/8kvo3LIQQ4j9JVSDPnj2bW7duER4ezjfffEP7JM8TVa9eHVtbW6ZNm0ZMTAx79uxh48aNdOjQ4R2blYeEqxrVr6+6eFu3Vvdak7NoEfj7w5Mn6r7rli1w/jxUr/72V3/yRHVV58unni+eMEFVyG9r8mR1z9vD49XXOnZU94zDwlTlPWFC/LPNCf3vf+DgAFOnPufZs7JotVrOnTvHsWPHAOjduzdjxozhypUr6HQ6zpw5w4MHD8iXLx8FChRg2bJlaLVaFi5cmGxwA7Rs2ZLLly+zdOlSYmJiiImJ4dixYwQEBLz9mxZCCPFOUhXInTp1wtvbm6JFi+Lp6WmY5EPPysqKjRs3smXLFvLmzUv//v1ZsmQJpUqVesdmOQCJR1o3aQILF6oBTidOqG329vGjn3PmVAFYqJDq3h0xQlWn7zL3RdOm0KyZenypcGGwsVGV6ttyc0v5+r6+ULWqquLLl4fKldW2pMzNVQ/BqVNaihRpSt68eenduzePHj0CYOjQobRr1w5vb29y5sxJr169ePZMzXc9b948vvvuO/LkycP58+epWbNmsm1xcHBg+/btrFixAjc3N1xdXRk5ciTRKQ37FkIIkeZMcPlFvTbAemT9X1C/N7UC1hi7IUIIIdKJCU4MojcMmQRDLwfq8xBCCJFVmXCFrANKAFeN3RATUBy4xLtOKSqEEML0mXCFrAFGoB9tnX3ZAyORMBZCiKzNhCtkgCjAGXj6ph2zMFsg7OV/hRBCZFUmXCED2AFdSeX8JVmQBdANCWMhhMj6TLxCBggEygPPjN0QI8gBnAX+20xcQgghTJ+JV8igwsgXVS1nJ7bAGCSMhRAie8gEFTKAFqgInEeNvs7qzIAywCnA3LhNEUIIkSEyQYUMKpRWATbGbkgGsQZWI2EshBDZRyYJZIDSZI+uaztUV/W7TjsqhBAiM8okXdZ6+q7rC2TNKTWlq1oIIbKrTFQhgwqpdUBOYzckneREvT8JYyGEyG4yWSADFAP+JuvN4GUP7Ea9PyGEENlNJgxkgErAVrLOhBm2qPdT0cjtEEIIYSyZ7B6yEEIIkTVl0gpZCCGEyFokkIUQQggTIIEshBBCmAAJZCGEEMIESCALIYQQJuD/Ab1JyQMYBXJfAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from collections import defaultdict\n",
    "\n",
    "\n",
    "G = nx.Graph()\n",
    "query_node_labels = {}\n",
    "for k, v in query['message']['query_graph']['nodes'].items():\n",
    "    node_ids = v.get('ids', [])\n",
    "    node_categories = v.get('categories', [])\n",
    "    G.add_node(k, categories=', '.join(node_categories))\n",
    "    query_node_labels[k] = '\\n'.join([k] + node_ids + node_categories)\n",
    "    #G.add_node(k, color='red')\n",
    "\n",
    "query_edge_labels = {}\n",
    "q_node_id_keys = set(['object', 'subject'])\n",
    "for q_edge_id,edge_v in query['message']['query_graph']['edges'].items():\n",
    "    for predicate in edge_v['predicates']:\n",
    "        G.add_edge(edge_v['subject'], edge_v['object'], predicate=predicate)\n",
    "    #G.add_edge(edge_v['subject'], edge_v['object'], predicates=', '.join(edge_v['predicates']))\n",
    "    #G.add_edge(edge_v['subject'], edge_v['object'], predicates=[1,2,3])\n",
    "    #G.add_edge(edge_v['subject'], edge_v['object'])\n",
    "    query_edge_labels[(edge_v['subject'], edge_v['object'])] = '\\n'.join(edge_v['predicates'])\n",
    "    \n",
    "query_template_f = './bte_sleeve_query_template.graphml'\n",
    "print(f'Analyze this GraphML file in Cytoscape: {query_template_f}')\n",
    "nx.write_graphml(G, query_template_f)\n",
    "\n",
    "options = {\n",
    "    'node_color': 'yellow',\n",
    "    'node_size': 5000,\n",
    "    'edge_color': 'black',\n",
    "    'width': 3,\n",
    "}\n",
    "query_template_layout = nx.spring_layout(G)\n",
    "nx.draw(G, pos=query_template_layout, **options)\n",
    "nx.draw_networkx_labels(G, query_template_layout, labels=query_node_labels, font_size=12,\n",
    "                        font_color='k', font_family='sans-serif',\n",
    "                        font_weight='normal', alpha=None, bbox=None, horizontalalignment='center',\n",
    "                        verticalalignment='center', ax=None, clip_on=False)\n",
    "nx.draw_networkx_edge_labels(G, pos=query_template_layout,\n",
    "                             edge_labels=query_edge_labels,\n",
    "                             label_pos=0.5, \n",
    "                             font_size=10, font_color='k', font_family='sans-serif',\n",
    "                             font_weight='normal', alpha=None, bbox=None, horizontalalignment='center',\n",
    "                             verticalalignment='center', ax=None, rotate=True, clip_on=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# NOTE: we're including categories specified in the query template\n",
    "# in the next cell, and in a cell further down, we're also including\n",
    "# any categories from the TRAPI results.\n",
    "#\n",
    "# TODO: Some categories are supersets of others. Should we handle\n",
    "# this systematically?\n",
    "\n",
    "curie_categories = set()\n",
    "for v in query[\"message\"][\"query_graph\"][\"nodes\"].values():\n",
    "    if \"categories\" in v:\n",
    "        for category in v[\"categories\"]:\n",
    "            curie_categories.add(category)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "XRfyhW_Is73C",
    "outputId": "76ae9c51-0587-4f1e-91ac-f8d728ce6878"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "curie UMLS:C0903898 has multiple primary names: NATEGLINIDE and nateglinide\n",
      "curie UMLS:C0021641 has multiple primary names: Iletin and Insulin\n",
      "all CURIE prefixes found:\n",
      "{'KEGG.COMPOUND', 'PUBCHEM.COMPOUND', 'CAS', 'HGNC', 'INCHIKEY', 'ENSEMBL', 'MESH', 'GTOPDB', 'UNII', 'UniProtKB', 'DrugCentral', 'DRUGBANK', 'CHEMBL.COMPOUND', 'OMIM', 'UMLS', 'CHEBI', 'NCBIGene', 'HMDB', 'PR'}\n",
      "\n",
      "unified CURIE prefixes found:\n",
      "{'NCBIGene', 'MESH'}\n",
      "\n",
      "failed to unify 886 CURIEs\n"
     ]
    }
   ],
   "source": [
    "# for genes/gene products, chemicals and diseases\n",
    "preferred_prefixes = set([\"NCBIGene\", \"MESH\"])\n",
    "unified_names = set()\n",
    "unified_curies = set()\n",
    "unification_failed_curies = set()\n",
    "unified_prefixes = set()\n",
    "all_prefixes = set()\n",
    "curie_to_name = dict()\n",
    "curie_to_unified_curie = dict()\n",
    "for k, v in trapi_message[\"knowledge_graph\"][\"nodes\"].items():\n",
    "    name = v[\"name\"]\n",
    "    for category in v[\"categories\"]:\n",
    "        curie_categories.add(category)\n",
    "        \n",
    "    for a in v[\"attributes\"]:\n",
    "        if a[\"attribute_type_id\"] == \"biolink:xref\":\n",
    "            curies = a[\"value\"]\n",
    "            \n",
    "            # k should always be one of the curies\n",
    "            if not k in curies:\n",
    "                raise Exception(f\"key {k} not in {curies}\")\n",
    "            \n",
    "            unified_curie = None\n",
    "            intersecting_unified_curies = unified_curies.intersection(set(curies))\n",
    "            if len(intersecting_unified_curies) > 1:\n",
    "                raise Exception(f\"matching multiple: {k} to {list(intersecting_unified_curies)}\")\n",
    "            elif len(intersecting_unified_curies) == 1:\n",
    "                unified_curie = list(intersecting_unified_curies)[0]\n",
    "            else:\n",
    "                # get curie for preferred prefix. usually this is k, but not always.\n",
    "                for curie in curies:\n",
    "                    [prefix, identifier] = curie.split(\":\")\n",
    "                    if prefix in preferred_prefixes:\n",
    "                        unified_curie = curie\n",
    "                        unified_curies.add(unified_curie)\n",
    "                        break\n",
    "                        \n",
    "            if not unified_curie:\n",
    "                if k in curie_to_unified_curie:\n",
    "                    unified_curie = curie_to_unified_curie[k]\n",
    "                else:\n",
    "                    unification_failed_curies.add(k)\n",
    "                    break\n",
    "                    \n",
    "                #raise Exception(f\"failed to find a unified curie for {k} in {curies}\")\n",
    "                \n",
    "            [unified_prefix, unified_identifier] = unified_curie.split(\":\")\n",
    "            unified_prefixes.add(unified_prefix)\n",
    "            \n",
    "            for curie in curies:\n",
    "                [prefix, identifier] = curie.split(\":\")\n",
    "                all_prefixes.add(prefix)\n",
    "                if not curie in curie_to_unified_curie:\n",
    "                    curie_to_unified_curie[curie] = unified_curie\n",
    "                if not curie in curie_to_name:\n",
    "                    curie_to_name[curie] = name\n",
    "                    unified_names.add(name)\n",
    "                elif curie_to_name[curie] != name:\n",
    "                    print(f\"curie {curie} has multiple primary names: {curie_to_name[curie]} and {name}\")\n",
    "                    #raise Exception(f\"curie{curie} has multiple names: {curie_to_name[curie]} and {name}\")\n",
    "\n",
    "\n",
    "#print(f'curie_categories: {curie_categories}')\n",
    "#print(\"\")\n",
    "#print(f\"curie_to_name key count: {len(curie_to_name.keys())}\")\n",
    "#print(f\"curie_to_unified_curie key count: {len(curie_to_unified_curie.keys())}\")\n",
    "#print(\"\")\n",
    "print(\"all CURIE prefixes found:\")\n",
    "print(all_prefixes)\n",
    "print(\"\")\n",
    "print(\"unified CURIE prefixes found:\")\n",
    "print(unified_prefixes)\n",
    "print(\"\")\n",
    "print(f\"failed to unify {len(unification_failed_curies)} CURIEs\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "columns = []\n",
    "q_node_id_keys = set([\"object\", \"subject\"])\n",
    "q_node_ids = []\n",
    "q_edge_ids = []\n",
    "for q_edge_id,edge_v in query[\"message\"][\"query_graph\"][\"edges\"].items():\n",
    "    q_edge_ids.append(q_edge_id)\n",
    "    q_node_id_found = False\n",
    "    for k,v in edge_v.items():\n",
    "        if (k in q_node_id_keys) and (type(v) is str):\n",
    "            if v not in columns:\n",
    "                q_node_ids.append(v)\n",
    "                columns.append(v)\n",
    "            if not q_node_id_found:\n",
    "                q_node_id_found = True\n",
    "                columns.append(q_edge_id)\n",
    "                \n",
    "query_nodes_with_ids = set()\n",
    "for k, v in query['message']['query_graph']['nodes'].items():\n",
    "    node_ids = v.get('ids', [])\n",
    "    if node_ids:\n",
    "        query_nodes_with_ids.add(k)\n",
    "                \n",
    "for q_node_id in q_node_ids:\n",
    "    columns.append(f\"{q_node_id}_curie\")\n",
    "    columns.append(f\"{q_node_id}_unified_curie\")\n",
    "    \n",
    "columns.append(\"trapi_result_curie_combo\")\n",
    "\n",
    "trapi_result_columns = []\n",
    "for i in range(len(q_node_ids)):\n",
    "    trapi_result_columns.append(q_node_ids[i])\n",
    "    if i < len(q_edge_ids):\n",
    "        trapi_result_columns.append(q_edge_ids[i])\n",
    "\n",
    "unified_curie_columns = []\n",
    "for q_node_id in q_node_ids:\n",
    "    unified_curie_columns.append(f\"{q_node_id}_unified_curie\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "id": "Wjqcf5E4Yk_h"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "warning: predicate direction(s) may be switched\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>n0</th>\n",
       "      <th>e0</th>\n",
       "      <th>n1</th>\n",
       "      <th>e1</th>\n",
       "      <th>n2</th>\n",
       "      <th>n0_curie</th>\n",
       "      <th>n0_unified_curie</th>\n",
       "      <th>n1_curie</th>\n",
       "      <th>n1_unified_curie</th>\n",
       "      <th>n2_curie</th>\n",
       "      <th>n2_unified_curie</th>\n",
       "      <th>trapi_result_curie_combo</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>entity_positively_regulates_entity</td>\n",
       "      <td>RAB18</td>\n",
       "      <td>entity_positively_regulated_by_entity</td>\n",
       "      <td>WATER</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>CHEMBL.COMPOUND:CHEMBL1098659</td>\n",
       "      <td>MESH:D014867</td>\n",
       "      <td>(MESH:D014867, NCBIGene:22931, NCBIGene:5872)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>entity_positively_regulates_entity</td>\n",
       "      <td>RAB18</td>\n",
       "      <td>entity_negatively_regulated_by_entity</td>\n",
       "      <td>WATER</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>CHEMBL.COMPOUND:CHEMBL1098659</td>\n",
       "      <td>MESH:D014867</td>\n",
       "      <td>(MESH:D014867, NCBIGene:22931, NCBIGene:5872)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>entity_positively_regulates_entity</td>\n",
       "      <td>RAB18</td>\n",
       "      <td>entity_positively_regulated_by_entity</td>\n",
       "      <td>Isopropyl-beta-D-thiogalactopyranoside</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>PUBCHEM.COMPOUND:656894</td>\n",
       "      <td>MESH:D007544</td>\n",
       "      <td>(MESH:D007544, NCBIGene:22931, NCBIGene:5872)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>entity_positively_regulates_entity</td>\n",
       "      <td>RAB18</td>\n",
       "      <td>entity_positively_regulated_by_entity</td>\n",
       "      <td>Trichostatin A</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>PUBCHEM.COMPOUND:444732</td>\n",
       "      <td>MESH:C012589</td>\n",
       "      <td>(MESH:C012589, NCBIGene:22931, NCBIGene:5872)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>entity_positively_regulates_entity</td>\n",
       "      <td>RAB18</td>\n",
       "      <td>entity_positively_regulated_by_entity</td>\n",
       "      <td>Mannitol</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>NCBIGene:22931</td>\n",
       "      <td>PUBCHEM.COMPOUND:6251</td>\n",
       "      <td>MESH:D008353</td>\n",
       "      <td>(MESH:D008353, NCBIGene:22931, NCBIGene:5872)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6125</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>genetically_interacts_with</td>\n",
       "      <td>TGFBR1</td>\n",
       "      <td>coexists_with</td>\n",
       "      <td>5' Flanking Region</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>UMLS:C0949645</td>\n",
       "      <td>MESH:D024506</td>\n",
       "      <td>(MESH:D024506, NCBIGene:5872, NCBIGene:7046)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6126</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>genetically_interacts_with</td>\n",
       "      <td>TGFBR1</td>\n",
       "      <td>coexists_with</td>\n",
       "      <td>1,3,4,6-Tetra-O-acetyl-2-azido-2-deoxy-beta-D-...</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>PUBCHEM.COMPOUND:133445</td>\n",
       "      <td>MESH:C070423</td>\n",
       "      <td>(MESH:C070423, NCBIGene:5872, NCBIGene:7046)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6127</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>genetically_interacts_with</td>\n",
       "      <td>TGFBR1</td>\n",
       "      <td>increases_response_to</td>\n",
       "      <td>Thioacetamide</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>PUBCHEM.COMPOUND:2723949</td>\n",
       "      <td>MESH:D013853</td>\n",
       "      <td>(MESH:D013853, NCBIGene:5872, NCBIGene:7046)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6128</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>genetically_interacts_with</td>\n",
       "      <td>TGFBR1</td>\n",
       "      <td>affects_response_to</td>\n",
       "      <td>Thioacetamide</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>PUBCHEM.COMPOUND:2723949</td>\n",
       "      <td>MESH:D013853</td>\n",
       "      <td>(MESH:D013853, NCBIGene:5872, NCBIGene:7046)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6129</th>\n",
       "      <td>RAB13</td>\n",
       "      <td>genetically_interacts_with</td>\n",
       "      <td>TGFBR1</td>\n",
       "      <td>coexists_with</td>\n",
       "      <td>Progesterone</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:5872</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>NCBIGene:7046</td>\n",
       "      <td>PUBCHEM.COMPOUND:5994</td>\n",
       "      <td>MESH:D011374</td>\n",
       "      <td>(MESH:D011374, NCBIGene:5872, NCBIGene:7046)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>6130 rows × 12 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         n0                                  e0      n1  \\\n",
       "0     RAB13  entity_positively_regulates_entity   RAB18   \n",
       "1     RAB13  entity_positively_regulates_entity   RAB18   \n",
       "2     RAB13  entity_positively_regulates_entity   RAB18   \n",
       "3     RAB13  entity_positively_regulates_entity   RAB18   \n",
       "4     RAB13  entity_positively_regulates_entity   RAB18   \n",
       "...     ...                                 ...     ...   \n",
       "6125  RAB13          genetically_interacts_with  TGFBR1   \n",
       "6126  RAB13          genetically_interacts_with  TGFBR1   \n",
       "6127  RAB13          genetically_interacts_with  TGFBR1   \n",
       "6128  RAB13          genetically_interacts_with  TGFBR1   \n",
       "6129  RAB13          genetically_interacts_with  TGFBR1   \n",
       "\n",
       "                                         e1  \\\n",
       "0     entity_positively_regulated_by_entity   \n",
       "1     entity_negatively_regulated_by_entity   \n",
       "2     entity_positively_regulated_by_entity   \n",
       "3     entity_positively_regulated_by_entity   \n",
       "4     entity_positively_regulated_by_entity   \n",
       "...                                     ...   \n",
       "6125                          coexists_with   \n",
       "6126                          coexists_with   \n",
       "6127                  increases_response_to   \n",
       "6128                    affects_response_to   \n",
       "6129                          coexists_with   \n",
       "\n",
       "                                                     n2       n0_curie  \\\n",
       "0                                                 WATER  NCBIGene:5872   \n",
       "1                                                 WATER  NCBIGene:5872   \n",
       "2                Isopropyl-beta-D-thiogalactopyranoside  NCBIGene:5872   \n",
       "3                                        Trichostatin A  NCBIGene:5872   \n",
       "4                                              Mannitol  NCBIGene:5872   \n",
       "...                                                 ...            ...   \n",
       "6125                                 5' Flanking Region  NCBIGene:5872   \n",
       "6126  1,3,4,6-Tetra-O-acetyl-2-azido-2-deoxy-beta-D-...  NCBIGene:5872   \n",
       "6127                                      Thioacetamide  NCBIGene:5872   \n",
       "6128                                      Thioacetamide  NCBIGene:5872   \n",
       "6129                                       Progesterone  NCBIGene:5872   \n",
       "\n",
       "     n0_unified_curie        n1_curie n1_unified_curie  \\\n",
       "0       NCBIGene:5872  NCBIGene:22931   NCBIGene:22931   \n",
       "1       NCBIGene:5872  NCBIGene:22931   NCBIGene:22931   \n",
       "2       NCBIGene:5872  NCBIGene:22931   NCBIGene:22931   \n",
       "3       NCBIGene:5872  NCBIGene:22931   NCBIGene:22931   \n",
       "4       NCBIGene:5872  NCBIGene:22931   NCBIGene:22931   \n",
       "...               ...             ...              ...   \n",
       "6125    NCBIGene:5872   NCBIGene:7046    NCBIGene:7046   \n",
       "6126    NCBIGene:5872   NCBIGene:7046    NCBIGene:7046   \n",
       "6127    NCBIGene:5872   NCBIGene:7046    NCBIGene:7046   \n",
       "6128    NCBIGene:5872   NCBIGene:7046    NCBIGene:7046   \n",
       "6129    NCBIGene:5872   NCBIGene:7046    NCBIGene:7046   \n",
       "\n",
       "                           n2_curie n2_unified_curie  \\\n",
       "0     CHEMBL.COMPOUND:CHEMBL1098659     MESH:D014867   \n",
       "1     CHEMBL.COMPOUND:CHEMBL1098659     MESH:D014867   \n",
       "2           PUBCHEM.COMPOUND:656894     MESH:D007544   \n",
       "3           PUBCHEM.COMPOUND:444732     MESH:C012589   \n",
       "4             PUBCHEM.COMPOUND:6251     MESH:D008353   \n",
       "...                             ...              ...   \n",
       "6125                  UMLS:C0949645     MESH:D024506   \n",
       "6126        PUBCHEM.COMPOUND:133445     MESH:C070423   \n",
       "6127       PUBCHEM.COMPOUND:2723949     MESH:D013853   \n",
       "6128       PUBCHEM.COMPOUND:2723949     MESH:D013853   \n",
       "6129          PUBCHEM.COMPOUND:5994     MESH:D011374   \n",
       "\n",
       "                           trapi_result_curie_combo  \n",
       "0     (MESH:D014867, NCBIGene:22931, NCBIGene:5872)  \n",
       "1     (MESH:D014867, NCBIGene:22931, NCBIGene:5872)  \n",
       "2     (MESH:D007544, NCBIGene:22931, NCBIGene:5872)  \n",
       "3     (MESH:C012589, NCBIGene:22931, NCBIGene:5872)  \n",
       "4     (MESH:D008353, NCBIGene:22931, NCBIGene:5872)  \n",
       "...                                             ...  \n",
       "6125   (MESH:D024506, NCBIGene:5872, NCBIGene:7046)  \n",
       "6126   (MESH:C070423, NCBIGene:5872, NCBIGene:7046)  \n",
       "6127   (MESH:D013853, NCBIGene:5872, NCBIGene:7046)  \n",
       "6128   (MESH:D013853, NCBIGene:5872, NCBIGene:7046)  \n",
       "6129   (MESH:D011374, NCBIGene:5872, NCBIGene:7046)  \n",
       "\n",
       "[6130 rows x 12 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trapi_results = trapi_message[\"results\"]\n",
    "\n",
    "result_row_data = []\n",
    "for trapi_result in trapi_results:\n",
    "    curie_to_qnode_ids = dict()\n",
    "    for qnode_id, entries in trapi_result[\"node_bindings\"].items():\n",
    "        for entry in entries:\n",
    "            curie = entry[\"id\"]\n",
    "            if curie not in curie_to_qnode_ids:\n",
    "                curie_to_qnode_ids[curie] = []\n",
    "            curie_to_qnode_ids[curie].append(qnode_id)\n",
    "        \n",
    "    row_data_template = dict()\n",
    "    q_edge_id_to_predicates = dict()\n",
    "    trapi_result_curies = set()\n",
    "    for qedge_id, entries in trapi_result[\"edge_bindings\"].items():\n",
    "        for entry in entries:\n",
    "            curie = entry[\"id\"]\n",
    "            kg_entry = trapi_message[\"knowledge_graph\"][\"edges\"][curie]\n",
    "            subject_curie = kg_entry[\"subject\"]\n",
    "            object_curie = kg_entry[\"object\"]\n",
    "            predicate_curie = kg_entry[\"predicate\"]\n",
    "            [predicate_prefix, predicate_identifier] = predicate_curie.split(\":\")\n",
    "            \n",
    "            if qedge_id not in q_edge_id_to_predicates:\n",
    "                q_edge_id_to_predicates[qedge_id] = set()\n",
    "            q_edge_id_to_predicates[qedge_id].add(predicate_identifier)\n",
    "\n",
    "            for curie in [subject_curie, object_curie]:\n",
    "                for qnode_id in curie_to_qnode_ids[curie]:\n",
    "                    if curie in curie_to_unified_curie:\n",
    "                        unified_curie = curie_to_unified_curie[curie]\n",
    "                    else:\n",
    "                        break\n",
    "\n",
    "                    name = curie_to_name[curie]\n",
    "                    row_data_template[qnode_id] = name\n",
    "\n",
    "                    trapi_result_curies.add(unified_curie)\n",
    "                    row_data_template[qnode_id + \"_curie\"] = curie\n",
    "                    row_data_template[qnode_id + \"_unified_curie\"] = unified_curie\n",
    "                    \n",
    "    if len(trapi_result_curies) != len(q_node_ids):\n",
    "        #print(f'skipping {list(curie_to_qnode_ids.keys())}')\n",
    "        continue\n",
    "        \n",
    "    row_data_template[\"trapi_result_curie_combo\"] = tuple(sorted(trapi_result_curies))\n",
    "    q_edge_ids_processed = set()\n",
    "    row_datas = [row_data_template]\n",
    "    for q_edge_id,predicates in q_edge_id_to_predicates.items():\n",
    "        next_row_datas = []\n",
    "        for row_data in row_datas:\n",
    "            for predicate in predicates:\n",
    "                next_row_data = deepcopy(row_data)\n",
    "                next_row_data[q_edge_id] = predicate\n",
    "                next_row_datas.append(\n",
    "                    next_row_data \n",
    "                )\n",
    "        row_datas = next_row_datas\n",
    "    result_row_data += row_datas\n",
    "        \n",
    "print(\"warning: predicate direction(s) may be switched\")\n",
    "results_df = pd.DataFrame.from_records(result_row_data, columns=columns).drop_duplicates()\n",
    "results_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Unique CURIE count per query node:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "n0: 1\n",
      "n1: 12\n",
      "n2: 1347\n"
     ]
    }
   ],
   "source": [
    "for q_node_id in q_node_ids:\n",
    "    print(f'{q_node_id}: {len(set(results_df[q_node_id]))}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Total results count (excluding any that couldn't be unified):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3158"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(set(results_df[\"trapi_result_curie_combo\"]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "That's too too high for a researcher to efficiently go through manually. Let's try filtering to make that more manageable."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "uCW1x4PxnjYJ",
    "tags": []
   },
   "source": [
    "## Match up BTE TRAPI Results & PFOCR\n",
    "\n",
    "We're going to try using PFOCR to filter and prioritize the results.\n",
    "\n",
    "First we need to get the PFOCR Data. We could get it from the API, but for now, we'll just go ahead and download the entire JSON file we gave to BTE."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "OyJJUHbepKp2",
    "outputId": "7237b5f0-8740-47cb-9e3d-b03ea02bdcab"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "status_code: 200\n"
     ]
    }
   ],
   "source": [
    "pfocr_url = \"https://www.dropbox.com/s/1f14t5zaseocyg6/bte_chemicals_diseases_genes.ndjson?dl=1\"\n",
    "pfocr_request = requests.get(pfocr_url)\n",
    "print(f\"status_code: {pfocr_request.status_code}\")\n",
    "if pfocr_request.status_code != 200:\n",
    "    print(pfocr_request.text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "id": "8fUaenkS0gg1"
   },
   "outputs": [],
   "source": [
    "curies_to_figure_ids = {}\n",
    "figure_id_to_curies = {}\n",
    "figure_id_to_pfocr_result = {}\n",
    "for line in pfocr_request.text.splitlines():\n",
    "    pfocr_result = json.loads(line)\n",
    "    figure_id = pfocr_result[\"_id\"]\n",
    "\n",
    "    curies = set()\n",
    "    for identifier in pfocr_result[\"associatedWith\"][\"mentions\"][\"chemicals\"][\"mesh\"]:\n",
    "        curie = \"MESH:\" + identifier\n",
    "        if curie in curie_to_unified_curie:\n",
    "            unified_curie = curie_to_unified_curie[curie]\n",
    "            curies.add(unified_curie)\n",
    "        else:\n",
    "            #print(f\"{curie} not in list\")\n",
    "            # this curie isn't in the BTE results, but we'll add an item for\n",
    "            # the purpose of the denominator in jaccard/containment cals\n",
    "            curies.add(curie)\n",
    "    for identifier in pfocr_result[\"associatedWith\"][\"mentions\"][\"diseases\"][\"mesh\"]:\n",
    "        curie = \"MESH:\" + identifier\n",
    "        if curie in curie_to_unified_curie:\n",
    "            unified_curie = curie_to_unified_curie[curie]\n",
    "            curies.add(unified_curie)\n",
    "        else:\n",
    "            #print(f\"{curie} not in list\")\n",
    "            # this curie isn't in the BTE results, but we'll add an item for\n",
    "            # the purpose of the denominator in jaccard/containment cals\n",
    "            curies.add(curie)\n",
    "    for identifier in pfocr_result[\"associatedWith\"][\"mentions\"][\"genes\"][\"ncbigene\"]:\n",
    "        curie = \"NCBIGene:\" + identifier\n",
    "        if curie in curie_to_unified_curie:\n",
    "            unified_curie = curie_to_unified_curie[curie]\n",
    "            curies.add(unified_curie)\n",
    "        else:\n",
    "            #print(f\"{curie} not in list\")\n",
    "            # this curie isn't in the BTE results, but we'll add an item for\n",
    "            # the purpose of the denominator in jaccard/containment cals\n",
    "            curies.add(curie)\n",
    "\n",
    "    figure_id_to_pfocr_result[figure_id] = pfocr_result\n",
    "    figure_id_to_curies[figure_id] = curies\n",
    "\n",
    "    curies_key = tuple(sorted(curies))\n",
    "    if curies_key not in curies_to_figure_ids:\n",
    "        curies_to_figure_ids[curies_key] = []\n",
    "    curies_to_figure_ids[curies_key].append(figure_id)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "aZX7IzakKCnH",
    "outputId": "651e9425-f1a7-40ca-c98f-59c7dfdadc14"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>figure_id</th>\n",
       "      <th>figure_url</th>\n",
       "      <th>figure_title</th>\n",
       "      <th>trapi_result_curie_combo</th>\n",
       "      <th>figure_curie_combo</th>\n",
       "      <th>overlap_curie_combo</th>\n",
       "      <th>score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>PMC7884075__elife-58615-fig4.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7...</td>\n",
       "      <td>GWAS hits in the IGF-1 pathway</td>\n",
       "      <td>(MESH:D015365, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(MESH:D009584, NCBIGene:10000, NCBIGene:10890,...</td>\n",
       "      <td>(NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>PMC6406872__cells-08-00192-g003.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6...</td>\n",
       "      <td>Structural hotspots and signaling pathways tha...</td>\n",
       "      <td>(MESH:D015365, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(NCBIGene:1000, NCBIGene:10000, NCBIGene:10050...</td>\n",
       "      <td>(NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>PMC5518130__12864_2017_3907_Fig4_HTML.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5...</td>\n",
       "      <td>Differently expressed protein-coding genes and...</td>\n",
       "      <td>(MESH:D015365, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(MESH:C031143, MESH:D000105, MESH:D000214, MES...</td>\n",
       "      <td>(NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>PMC6472122__MMR-19-05-3564-g04.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6...</td>\n",
       "      <td>Transfer RNA-derived fragment target genes in ...</td>\n",
       "      <td>(MESH:D015365, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(MESH:C031143, MESH:C045651, MESH:D000105, MES...</td>\n",
       "      <td>(NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>PMC5546324__nihms866088f6.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5...</td>\n",
       "      <td>AKT/PKB Signaling: Navigating the Network</td>\n",
       "      <td>(MESH:D015365, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(MESH:D000249, MESH:D005947, NCBIGene:10000, N...</td>\n",
       "      <td>(NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66865</th>\n",
       "      <td>PMC6895093__41598_2019_54408_Fig15_HTML.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6...</td>\n",
       "      <td>Apoptosis adapted from KEGG (ID: hsa04210), th...</td>\n",
       "      <td>(MESH:D006160, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(MESH:C047902, MESH:D005947, MESH:D006695, MES...</td>\n",
       "      <td>(NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66866</th>\n",
       "      <td>PMC3740206__nihms-424399-f0004.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3...</td>\n",
       "      <td>Exocyst Sec10 is Involved in Basolateral Prote...</td>\n",
       "      <td>(MESH:D006160, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(MESH:C117212, MESH:D006160, NCBIGene:10890, N...</td>\n",
       "      <td>(MESH:D006160, NCBIGene:5872)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66867</th>\n",
       "      <td>PMC5663315__nihms889033f1.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5...</td>\n",
       "      <td>Schematic representation of the mTOR signaling...</td>\n",
       "      <td>(MESH:D006160, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(MESH:C030986, MESH:C060980, MESH:D000255, MES...</td>\n",
       "      <td>(MESH:D006160, NCBIGene:3630)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66868</th>\n",
       "      <td>PMC6472122__MMR-19-05-3564-g04.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6...</td>\n",
       "      <td>Transfer RNA-derived fragment target genes in ...</td>\n",
       "      <td>(MESH:D006160, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(MESH:C031143, MESH:C045651, MESH:D000105, MES...</td>\n",
       "      <td>(NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66869</th>\n",
       "      <td>PMC5546324__nihms866088f6.jpg</td>\n",
       "      <td>https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5...</td>\n",
       "      <td>AKT/PKB Signaling: Navigating the Network</td>\n",
       "      <td>(MESH:D006160, NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>(MESH:D000249, MESH:D005947, NCBIGene:10000, N...</td>\n",
       "      <td>(NCBIGene:3630, NCBIGene:5872)</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>66870 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                         figure_id  \\\n",
       "0                 PMC7884075__elife-58615-fig4.jpg   \n",
       "1              PMC6406872__cells-08-00192-g003.jpg   \n",
       "2        PMC5518130__12864_2017_3907_Fig4_HTML.jpg   \n",
       "3               PMC6472122__MMR-19-05-3564-g04.jpg   \n",
       "4                    PMC5546324__nihms866088f6.jpg   \n",
       "...                                            ...   \n",
       "66865  PMC6895093__41598_2019_54408_Fig15_HTML.jpg   \n",
       "66866           PMC3740206__nihms-424399-f0004.jpg   \n",
       "66867                PMC5663315__nihms889033f1.jpg   \n",
       "66868           PMC6472122__MMR-19-05-3564-g04.jpg   \n",
       "66869                PMC5546324__nihms866088f6.jpg   \n",
       "\n",
       "                                              figure_url  \\\n",
       "0      https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7...   \n",
       "1      https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6...   \n",
       "2      https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5...   \n",
       "3      https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6...   \n",
       "4      https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5...   \n",
       "...                                                  ...   \n",
       "66865  https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6...   \n",
       "66866  https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3...   \n",
       "66867  https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5...   \n",
       "66868  https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6...   \n",
       "66869  https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5...   \n",
       "\n",
       "                                            figure_title  \\\n",
       "0                         GWAS hits in the IGF-1 pathway   \n",
       "1      Structural hotspots and signaling pathways tha...   \n",
       "2      Differently expressed protein-coding genes and...   \n",
       "3      Transfer RNA-derived fragment target genes in ...   \n",
       "4              AKT/PKB Signaling: Navigating the Network   \n",
       "...                                                  ...   \n",
       "66865  Apoptosis adapted from KEGG (ID: hsa04210), th...   \n",
       "66866  Exocyst Sec10 is Involved in Basolateral Prote...   \n",
       "66867  Schematic representation of the mTOR signaling...   \n",
       "66868  Transfer RNA-derived fragment target genes in ...   \n",
       "66869          AKT/PKB Signaling: Navigating the Network   \n",
       "\n",
       "                           trapi_result_curie_combo  \\\n",
       "0      (MESH:D015365, NCBIGene:3630, NCBIGene:5872)   \n",
       "1      (MESH:D015365, NCBIGene:3630, NCBIGene:5872)   \n",
       "2      (MESH:D015365, NCBIGene:3630, NCBIGene:5872)   \n",
       "3      (MESH:D015365, NCBIGene:3630, NCBIGene:5872)   \n",
       "4      (MESH:D015365, NCBIGene:3630, NCBIGene:5872)   \n",
       "...                                             ...   \n",
       "66865  (MESH:D006160, NCBIGene:3630, NCBIGene:5872)   \n",
       "66866  (MESH:D006160, NCBIGene:3630, NCBIGene:5872)   \n",
       "66867  (MESH:D006160, NCBIGene:3630, NCBIGene:5872)   \n",
       "66868  (MESH:D006160, NCBIGene:3630, NCBIGene:5872)   \n",
       "66869  (MESH:D006160, NCBIGene:3630, NCBIGene:5872)   \n",
       "\n",
       "                                      figure_curie_combo  \\\n",
       "0      (MESH:D009584, NCBIGene:10000, NCBIGene:10890,...   \n",
       "1      (NCBIGene:1000, NCBIGene:10000, NCBIGene:10050...   \n",
       "2      (MESH:C031143, MESH:D000105, MESH:D000214, MES...   \n",
       "3      (MESH:C031143, MESH:C045651, MESH:D000105, MES...   \n",
       "4      (MESH:D000249, MESH:D005947, NCBIGene:10000, N...   \n",
       "...                                                  ...   \n",
       "66865  (MESH:C047902, MESH:D005947, MESH:D006695, MES...   \n",
       "66866  (MESH:C117212, MESH:D006160, NCBIGene:10890, N...   \n",
       "66867  (MESH:C030986, MESH:C060980, MESH:D000255, MES...   \n",
       "66868  (MESH:C031143, MESH:C045651, MESH:D000105, MES...   \n",
       "66869  (MESH:D000249, MESH:D005947, NCBIGene:10000, N...   \n",
       "\n",
       "                  overlap_curie_combo     score  \n",
       "0      (NCBIGene:3630, NCBIGene:5872)  0.666667  \n",
       "1      (NCBIGene:3630, NCBIGene:5872)  0.666667  \n",
       "2      (NCBIGene:3630, NCBIGene:5872)  0.666667  \n",
       "3      (NCBIGene:3630, NCBIGene:5872)  0.666667  \n",
       "4      (NCBIGene:3630, NCBIGene:5872)  0.666667  \n",
       "...                               ...       ...  \n",
       "66865  (NCBIGene:3630, NCBIGene:5872)  0.666667  \n",
       "66866   (MESH:D006160, NCBIGene:5872)  0.666667  \n",
       "66867   (MESH:D006160, NCBIGene:3630)  0.666667  \n",
       "66868  (NCBIGene:3630, NCBIGene:5872)  0.666667  \n",
       "66869  (NCBIGene:3630, NCBIGene:5872)  0.666667  \n",
       "\n",
       "[66870 rows x 7 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from SetSimilaritySearch import SearchIndex\n",
    "\n",
    "# the reference sets are supposed to be from PFOCR only\n",
    "reference_sets = list()\n",
    "for curies in set(curies_to_figure_ids.keys()):\n",
    "    reference_sets.append(set(curies))\n",
    "# but in order to calculate the scores correctly,\n",
    "# the SetSimilaritySearch library requires that\n",
    "# every curie from the bte results need to be\n",
    "# mentioned at least once in the reference set.\n",
    "# That's the only reason we add them below, as\n",
    "# one large set.\n",
    "\n",
    "unified_curie_columns = [\n",
    "    q_node_id + \"_unified_curie\" for q_node_id in q_node_ids\n",
    "]\n",
    "\n",
    "trapi_result_curie_combos = set()\n",
    "for i, df in results_df[unified_curie_columns].drop_duplicates().iterrows():\n",
    "    trapi_result_curies = []\n",
    "    for unified_curie_column in unified_curie_columns:\n",
    "        trapi_result_curies.append(\n",
    "            df[unified_curie_column]\n",
    "        )\n",
    "    trapi_result_curie_combos.add(tuple(sorted(\n",
    "        trapi_result_curies\n",
    "    )))\n",
    "    \n",
    "reference_set = set()\n",
    "for trapi_result_curie_combo in trapi_result_curie_combos:\n",
    "    reference_set |= set(trapi_result_curie_combo)\n",
    "reference_sets.append(reference_set)\n",
    "\n",
    "matchable_node_min = 2\n",
    "matchable_node_count = len(q_node_ids)\n",
    "index = SearchIndex(reference_sets, similarity_func_name=\"containment\", \n",
    "    similarity_threshold=matchable_node_min/matchable_node_count)\n",
    "\n",
    "trapi_figure_overlap_rows = []\n",
    "for trapi_result_curie_combo in trapi_result_curie_combos:\n",
    "    trapi_curies = set(trapi_result_curie_combo)\n",
    "    results = index.query(trapi_curies)\n",
    "    for result in results:\n",
    "        figure_curie_combos = reference_sets[result[0]]\n",
    "        figure_curie_key = tuple(sorted(figure_curie_combos))\n",
    "        \n",
    "        # needed to not match bte results to themselves\n",
    "        if figure_curie_key in curies_to_figure_ids:\n",
    "            figure_ids = curies_to_figure_ids[figure_curie_key]\n",
    "        else:\n",
    "            continue\n",
    "            \n",
    "        score = result[1]\n",
    "        \n",
    "        common_curies = trapi_curies.intersection(figure_curie_combos)\n",
    "        for figure_id in figure_ids:\n",
    "            pfocr_result = figure_id_to_pfocr_result[figure_id]\n",
    "            trapi_figure_overlap_rows.append({\n",
    "                \"figure_id\": figure_id,\n",
    "                \"figure_url\": pfocr_result[\"associatedWith\"][\"figureUrl\"],\n",
    "                \"figure_title\": pfocr_result[\"associatedWith\"][\"title\"],\n",
    "                \"trapi_result_curie_combo\": trapi_result_curie_combo,\n",
    "                \"figure_curie_combo\": figure_curie_key,\n",
    "                \"overlap_curie_combo\": tuple(sorted(common_curies)),\n",
    "                \"score\": score,\n",
    "            })\n",
    "\n",
    "\n",
    "trapi_figure_overlap_df = pd.DataFrame.from_records(trapi_figure_overlap_rows)\n",
    "trapi_figure_overlap_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### trapi_curie_combo_count_by_figure_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>trapi_curie_combo_count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>figure_id</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>PMC6406872__cells-08-00192-g003.jpg</th>\n",
       "      <td>1470</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PMC6472122__MMR-19-05-3564-g04.jpg</th>\n",
       "      <td>1182</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PMC5518130__12864_2017_3907_Fig4_HTML.jpg</th>\n",
       "      <td>1175</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PMC6770969__genes-10-00728-g006.jpg</th>\n",
       "      <td>1165</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PMC6895093__41598_2019_54408_Fig15_HTML.jpg</th>\n",
       "      <td>1150</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PMC4623426__fncel-09-00426-g0004.jpg</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PMC4623426__fncel-09-00426-g0003.jpg</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PMC4620423__WJSC-7-1150-g001.jpg</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PMC4610971__gr1.jpg</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PMC86511__mb2301061009.jpg</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2296 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                             trapi_curie_combo_count\n",
       "figure_id                                                           \n",
       "PMC6406872__cells-08-00192-g003.jpg                             1470\n",
       "PMC6472122__MMR-19-05-3564-g04.jpg                              1182\n",
       "PMC5518130__12864_2017_3907_Fig4_HTML.jpg                       1175\n",
       "PMC6770969__genes-10-00728-g006.jpg                             1165\n",
       "PMC6895093__41598_2019_54408_Fig15_HTML.jpg                     1150\n",
       "...                                                              ...\n",
       "PMC4623426__fncel-09-00426-g0004.jpg                               1\n",
       "PMC4623426__fncel-09-00426-g0003.jpg                               1\n",
       "PMC4620423__WJSC-7-1150-g001.jpg                                   1\n",
       "PMC4610971__gr1.jpg                                                1\n",
       "PMC86511__mb2301061009.jpg                                         1\n",
       "\n",
       "[2296 rows x 1 columns]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trapi_curie_combo_count_by_figure_df = trapi_figure_overlap_df[\n",
    "    [\"figure_id\", \"trapi_result_curie_combo\"]\n",
    "].rename(columns={\n",
    "    \"trapi_result_curie_combo\": \"trapi_curie_combo_count\"\n",
    "}).groupby(\"figure_id\").count().sort_values(\n",
    "    by=\"trapi_curie_combo_count\", ascending=False)\n",
    "trapi_curie_combo_count_by_figure_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure count by TRAPI CURIE combo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>figure_count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>trapi_result_curie_combo</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>(NCBIGene:4088, NCBIGene:4089, NCBIGene:5872)</th>\n",
       "      <td>1790</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:D002784, NCBIGene:22931, NCBIGene:5872)</th>\n",
       "      <td>192</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:D004317, NCBIGene:22931, NCBIGene:5872)</th>\n",
       "      <td>191</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:D003561, NCBIGene:22931, NCBIGene:5872)</th>\n",
       "      <td>191</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C423915, NCBIGene:22931, NCBIGene:5872)</th>\n",
       "      <td>191</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C085911, NCBIGene:5872, NCBIGene:6789)</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C085911, NCBIGene:5872, NCBIGene:7316)</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C554092, NCBIGene:5147, NCBIGene:5872)</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C506244, NCBIGene:5147, NCBIGene:5872)</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C097613, NCBIGene:5147, NCBIGene:5872)</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2179 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                               figure_count\n",
       "trapi_result_curie_combo                                   \n",
       "(NCBIGene:4088, NCBIGene:4089, NCBIGene:5872)          1790\n",
       "(MESH:D002784, NCBIGene:22931, NCBIGene:5872)           192\n",
       "(MESH:D004317, NCBIGene:22931, NCBIGene:5872)           191\n",
       "(MESH:D003561, NCBIGene:22931, NCBIGene:5872)           191\n",
       "(MESH:C423915, NCBIGene:22931, NCBIGene:5872)           191\n",
       "...                                                     ...\n",
       "(MESH:C085911, NCBIGene:5872, NCBIGene:6789)              1\n",
       "(MESH:C085911, NCBIGene:5872, NCBIGene:7316)              1\n",
       "(MESH:C554092, NCBIGene:5147, NCBIGene:5872)              1\n",
       "(MESH:C506244, NCBIGene:5147, NCBIGene:5872)              1\n",
       "(MESH:C097613, NCBIGene:5147, NCBIGene:5872)              1\n",
       "\n",
       "[2179 rows x 1 columns]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trapi_figure_overlap_df[\n",
    "    [\"trapi_result_curie_combo\", \"figure_id\"]\n",
    "].rename(columns={\n",
    "    \"figure_id\": \"figure_count\"\n",
    "}).groupby(\"trapi_result_curie_combo\").count().sort_values(\n",
    "    by=\"figure_count\", ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Figure containment score by TRAPI CURIE combo\n",
    "\n",
    "This can yield the same order of results, but if one TRAPI result has higher overlap of CURIEs with the figure(s), that TRAPI result could leapfrog over another TRAPI result that overlaps with more figures but at lower containment scores.\n",
    "\n",
    "Note the weakness of containment is it doesn't take into account the number of CURIEs in the figure."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cumulative_trapi_result_curie_combo_figure_score</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>trapi_result_curie_combo</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>(NCBIGene:4088, NCBIGene:4089, NCBIGene:5872)</th>\n",
       "      <td>1195.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:D002784, NCBIGene:22931, NCBIGene:5872)</th>\n",
       "      <td>139.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:D019821, NCBIGene:201475, NCBIGene:5872)</th>\n",
       "      <td>131.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C031143, NCBIGene:22931, NCBIGene:5872)</th>\n",
       "      <td>128.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:D013395, NCBIGene:22931, NCBIGene:5872)</th>\n",
       "      <td>128.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C085911, NCBIGene:5872, NCBIGene:6789)</th>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C085911, NCBIGene:5872, NCBIGene:7316)</th>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C554092, NCBIGene:5147, NCBIGene:5872)</th>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C506244, NCBIGene:5147, NCBIGene:5872)</th>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>(MESH:C097613, NCBIGene:5147, NCBIGene:5872)</th>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2179 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                cumulative_trapi_result_curie_combo_figure_score\n",
       "trapi_result_curie_combo                                                                        \n",
       "(NCBIGene:4088, NCBIGene:4089, NCBIGene:5872)                                        1195.333333\n",
       "(MESH:D002784, NCBIGene:22931, NCBIGene:5872)                                         139.666667\n",
       "(MESH:D019821, NCBIGene:201475, NCBIGene:5872)                                        131.000000\n",
       "(MESH:C031143, NCBIGene:22931, NCBIGene:5872)                                         128.333333\n",
       "(MESH:D013395, NCBIGene:22931, NCBIGene:5872)                                         128.000000\n",
       "...                                                                                          ...\n",
       "(MESH:C085911, NCBIGene:5872, NCBIGene:6789)                                            0.666667\n",
       "(MESH:C085911, NCBIGene:5872, NCBIGene:7316)                                            0.666667\n",
       "(MESH:C554092, NCBIGene:5147, NCBIGene:5872)                                            0.666667\n",
       "(MESH:C506244, NCBIGene:5147, NCBIGene:5872)                                            0.666667\n",
       "(MESH:C097613, NCBIGene:5147, NCBIGene:5872)                                            0.666667\n",
       "\n",
       "[2179 rows x 1 columns]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trapi_result_figure_score_df = trapi_figure_overlap_df[\n",
    "    [\"trapi_result_curie_combo\", \"score\"]\n",
    "].rename(\n",
    "    columns={\"score\": \"cumulative_trapi_result_curie_combo_figure_score\"}\n",
    ").groupby(\"trapi_result_curie_combo\").sum().sort_values(\n",
    "    by=\"cumulative_trapi_result_curie_combo_figure_score\", ascending=False)\n",
    "trapi_result_figure_score_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "results_with_figures_df = trapi_figure_overlap_df.merge(\n",
    "    results_df,\n",
    "    on=\"trapi_result_curie_combo\",\n",
    "    how=\"left\").merge(\n",
    "    trapi_result_figure_score_df.reset_index(),\n",
    "    on=\"trapi_result_curie_combo\",\n",
    "    how=\"left\",\n",
    "    validate=\"many_to_one\"\n",
    ").merge(\n",
    "    trapi_curie_combo_count_by_figure_df.reset_index().rename(columns={\n",
    "        \"trapi_curie_combo_count\": \"trapi_curie_combo_count_by_figure\"\n",
    "    }),\n",
    "    on=\"figure_id\",\n",
    "    how=\"left\",\n",
    "    validate=\"many_to_one\"\n",
    ").sort_values(\n",
    "    by=\"cumulative_trapi_result_curie_combo_figure_score\", ascending=False\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": []
   },
   "source": [
    "## Fisher's Exact Test\n",
    "\n",
    "Let's quantify the match up between figures and TRAPI results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total instance counts:\n",
      "  chemicals: 275456\n",
      "  diseases: 20465\n",
      "  genes: 1369680\n",
      "\n",
      "unique counts:\n",
      "  chemicals: 14482\n",
      "  diseases: 1430\n",
      "  genes: 14253\n",
      "\n",
      "M for Fisher Exact test: 28735\n",
      "TODO: should this number be based on what is in BTE/TRAPI?\n"
     ]
    }
   ],
   "source": [
    "all_genes = set()\n",
    "gene_instance_count = 0\n",
    "all_chemicals = set()\n",
    "chemical_instance_count = 0\n",
    "all_diseases = set()\n",
    "disease_instance_count = 0\n",
    "for figure_id, pfocr_result in figure_id_to_pfocr_result.items():\n",
    "    genes = pfocr_result[\"associatedWith\"][\"mentions\"][\"genes\"][\"ncbigene\"]\n",
    "    for gene in genes:\n",
    "        gene_instance_count += 1\n",
    "        all_genes.add(gene)\n",
    "\n",
    "    chemicals = pfocr_result[\"associatedWith\"][\"mentions\"][\"chemicals\"][\"mesh\"]\n",
    "    for chemical in chemicals:\n",
    "        chemical_instance_count += 1\n",
    "        all_chemicals.add(chemical)\n",
    "\n",
    "    diseases = pfocr_result[\"associatedWith\"][\"mentions\"][\"diseases\"][\"mesh\"]\n",
    "    for disease in diseases:\n",
    "        disease_instance_count += 1\n",
    "        all_diseases.add(disease)\n",
    "\n",
    "print(\"total instance counts:\")\n",
    "print(f'  chemicals: {chemical_instance_count}')\n",
    "print(f'  diseases: {disease_instance_count}')\n",
    "print(f'  genes: {gene_instance_count}')\n",
    "\n",
    "print(\"\")\n",
    "\n",
    "print(\"unique counts:\")\n",
    "print(f'  chemicals: {len(all_chemicals)}')\n",
    "print(f'  diseases: {len(all_diseases)}')\n",
    "print(f'  genes: {len(all_genes)}')\n",
    "\n",
    "print(\"\")\n",
    "\n",
    "curie_category_to_curies = {\n",
    "    \"biolink:ChemicalEntity\": all_chemicals,\n",
    "    \"biolink:SmallMolecule\": all_chemicals,\n",
    "    \"biolink:Disease\": all_diseases,\n",
    "    \"biolink:Gene\": all_genes,\n",
    "}\n",
    "\n",
    "\n",
    "all_figure_curies = set()\n",
    "for curie_category in set(curie_category_to_curies.keys()).intersection(curie_categories):\n",
    "    all_figure_curies |= curie_category_to_curies[curie_category]\n",
    "all_figure_curie_count = len(all_figure_curies)\n",
    "\n",
    "print(f'M for Fisher Exact test: {all_figure_curie_count}')\n",
    "print('TODO: should this number be based on what is in BTE/TRAPI?')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "See [this contingency table](https://docs.google.com/spreadsheets/d/1d-48R5N0auuKXSLJ7oZ2E_ldO26Z5KiEH3AQx5kDrC4/edit#gid=0):\n",
    "\n",
    "|      |  |\n",
    "| ----------- | ----------- |\n",
    "| trapi_curies_in_pwy      | q_node_id_count - trapi_curies_in_pwy       |\n",
    "| curies_in_pwy - trapi_curies_in_pwy   | 28735 - curies_in_pwy - q_node_id_count + trapi_curies_in_pwy        |\n",
    "\n",
    "and the [scipy documentation](https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.fisher_exact.html):\n",
    "\n",
    "|      |  |\n",
    "| ----------- | ----------- |\n",
    "| x      | n - x       |\n",
    "| N - x   | M - (n + N) + x        |\n",
    "\n",
    "and the [fisher documentation](https://pypi.org/project/fisher/)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# trapi_curies_in_pwy\n",
    "# x\n",
    "results_with_figures_df[\"yes_trapi_yes_pwy\"] = (\n",
    "    results_with_figures_df[\"overlap_curie_combo\"].map(len)\n",
    ")\n",
    "\n",
    "# q_node_id_count - trapi_curies_in_pwy\n",
    "# n - x\n",
    "results_with_figures_df[\"yes_trapi_no_pwy\"] = (\n",
    "    len(q_node_ids) - results_with_figures_df[\"yes_trapi_yes_pwy\"]\n",
    ")\n",
    "\n",
    "# curies_in_pwy - trapi_curies_in_pwy\n",
    "# N - x\n",
    "results_with_figures_df[\"no_trapi_yes_pwy\"] = (\n",
    "    results_with_figures_df[\"figure_curie_combo\"].map(len) - \n",
    "    results_with_figures_df[\"yes_trapi_yes_pwy\"]\n",
    ")\n",
    "\n",
    "# 28735 - curies_in_pwy - q_node_id_count + trapi_curies_in_pwy\n",
    "# M - (n + N) + x\n",
    "results_with_figures_df[\"no_trapi_no_pwy\"] = (\n",
    "    all_figure_curie_count -\n",
    "    results_with_figures_df[\"figure_curie_combo\"].map(len) -\n",
    "    len(q_node_ids) +\n",
    "    results_with_figures_df[\"yes_trapi_yes_pwy\"]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "127522    1.572988e-08\n",
       "127438    1.639452e-08\n",
       "127503    1.707763e-08\n",
       "127338    1.777945e-08\n",
       "127361    1.850024e-08\n",
       "              ...     \n",
       "53293     3.258204e-04\n",
       "37846     4.786686e-04\n",
       "37847     4.786686e-04\n",
       "37848     4.786686e-04\n",
       "37849     4.786686e-04\n",
       "Name: p_value, Length: 135113, dtype: float64"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# see https://stackoverflow.com/a/58661068/5354298\n",
    "\n",
    "_, _, twosided = pvalue_npy(\n",
    "    results_with_figures_df[\"yes_trapi_yes_pwy\"].to_numpy('uint'),\n",
    "    results_with_figures_df[\"yes_trapi_no_pwy\"].to_numpy('uint'),\n",
    "    results_with_figures_df[\"no_trapi_yes_pwy\"].to_numpy('uint'),\n",
    "    results_with_figures_df[\"no_trapi_no_pwy\"].to_numpy('uint'),\n",
    ")\n",
    "# TODO: do we need the odds ratio?\n",
    "#odds = (\n",
    "#    results_with_figures_df[\"yes_trapi_yes_pwy\"].to_numpy('uint') *\n",
    "#    results_with_figures_df[\"no_trapi_no_pwy\"].to_numpy('uint')\n",
    "#) / (\n",
    "#    results_with_figures_df[\"yes_trapi_no_pwy\"].to_numpy('uint') *\n",
    "#    results_with_figures_df[\"no_trapi_yes_pwy\"].to_numpy('uint')\n",
    "#)\n",
    "\n",
    "results_with_figures_df[\"p_value\"] = pd.Series(twosided, index=results_with_figures_df.index)\n",
    "results_with_figures_df[\"p_value\"].sort_values()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([8.8841e+04, 3.8305e+04, 1.5470e+03, 4.3590e+03, 1.4740e+03,\n",
       "        0.0000e+00, 5.8300e+02, 0.0000e+00, 0.0000e+00, 4.0000e+00]),\n",
       " array([1.57298812e-08, 4.78810163e-05, 9.57463027e-05, 1.43611589e-04,\n",
       "        1.91476876e-04, 2.39342162e-04, 2.87207448e-04, 3.35072735e-04,\n",
       "        3.82938021e-04, 4.30803308e-04, 4.78668594e-04]),\n",
       " <BarContainer object of 10 artists>)"
      ]
     },
     "execution_count": 21,
     "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": [
    "plt.hist(\n",
    "    results_with_figures_df[\"p_value\"]\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Analysis of results with vs. without BTE Sleeve"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "With the current query graph, requiring co-occurrence of TRAPI result CURIE(s) for particular node(s) in at least one PFOCR figure cuts down on the number of TRAPI result CURIE combos.\n",
    "\n",
    "Note: for the number of TRAPI result CURIE combos returned, we're only including a combo when every CURIE in it can be unified. If any of the CURIEs in a combo cannot be unified, we drop that combo."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Without BTE Sleeve"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3158 TRAPI results\n",
      "77719 figures\n"
     ]
    }
   ],
   "source": [
    "print(f'{len(set(results_df[\"trapi_result_curie_combo\"]))} TRAPI results')\n",
    "print(f'{len(figure_id_to_curies.keys())} figures')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### With BTE Sleeve\n",
    "Below are the totals we get when we require that the CURIEs for specified query nodes must be found in at least one figure, e.g., when the CURIE for `n0` and the CURIE for `n1` must be found in the same figure at least once."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "n0 & n1\n",
      "  2055 TRAPI results\n",
      "  196 figures\n",
      "n0 & n2\n",
      "  268 TRAPI results\n",
      "  105 figures\n",
      "n1 & n2\n",
      "  306 TRAPI results\n",
      "  2144 figures\n",
      "n0 & n1 & n2\n",
      "  41 TRAPI results\n",
      "  48 figures\n"
     ]
    }
   ],
   "source": [
    "for i in range(2, len(q_node_ids) + 1):\n",
    "    for q_node_id_combination in combinations(q_node_ids, i):\n",
    "        print(f'{\" & \".join(q_node_id_combination)}')\n",
    "        s = pd.Series([True] * len(results_with_figures_df), index=results_with_figures_df.index)\n",
    "        for q_node_id in q_node_id_combination:\n",
    "            s = s & results_with_figures_df.apply(\n",
    "                lambda r: r[f'{q_node_id}_unified_curie'] in r[\"overlap_curie_combo\"], axis=1\n",
    "            )\n",
    "        trapi_result_count = len(set(results_with_figures_df[s == True][\"trapi_result_curie_combo\"]))\n",
    "        print(f'  {trapi_result_count} TRAPI results')\n",
    "        figure_count = len(set(results_with_figures_df[s == True][\"figure_id\"]))\n",
    "        print(f'  {figure_count} figures')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": []
   },
   "source": [
    "## View TRAPI results and figures\n",
    "\n",
    "Compare these results with the `results_df` table from earlier in this notebook and also with the [ARAX UI](https://arax.ncats.io/) (copy the query as JSON into the input field). You can change the definition of `cooccurrence_df` below to update which nodes must be found in at least one figure. You can also change `outer_limit` and `inner_limit` to control how many results are shown. Click a figure to view its paper.\n",
    "\n",
    "Questions:\n",
    "1. Is it useful to view figures by TRAPI result or TRAPI results by figure?\n",
    "2. Does requiring co-occurrence for specific nodes like `n0` and `n2` help? Is it true that we should generally require co-occurrence by node(s) for which the user specified `ids`?\n",
    "3. Does ranking by p-value help?\n",
    "4. For ranking a given TRAPI result, should we take into account the number of figures that match? We're currently consider a TRAPI result to be better if it has just one figure with a slightly lower p-value than if it had 50 figures with marginally higher p-values. The cumulative containment does take into account the number of figures, but it does not take into account differences in pathway size like Fisher's Exact Test does.\n",
    "5. Is it possible we'd want to uprank a TRAPI result when it matches for some nodes but downrank it when it has matches for other nodes? Example: maybe we DON'T want the TRAPI results that indicate what is already well-known for a certain hop, so we downrank when there are a large number of figures matching that hop."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "n0\n"
     ]
    }
   ],
   "source": [
    "## any two nodes\n",
    "#cooccurrence_df = results_with_figures_df.sort_values(\n",
    "#    \"p_value\"\n",
    "#)\n",
    "\n",
    "## first and any other query node\n",
    "#q_node_id_combination = (q_node_ids[0],)\n",
    "\n",
    "# first and second query nodes\n",
    "#q_node_id_combination = (q_node_ids[0], q_node_ids[1])\n",
    "\n",
    "## first and last query nodes\n",
    "#q_node_id_combination = (q_node_ids[0], q_node_ids[-1])\n",
    "\n",
    "## second and last query nodes\n",
    "#q_node_id_combination = (q_node_ids[1], q_node_ids[-1])\n",
    "\n",
    "# every query node with 'ids' specified\n",
    "q_node_id_combination = query_nodes_with_ids\n",
    "\n",
    "print(f'{\" & \".join(q_node_id_combination)}')\n",
    "s = pd.Series([True] * len(results_with_figures_df), index=results_with_figures_df.index)\n",
    "for q_node_id in q_node_id_combination:\n",
    "    s = s & results_with_figures_df.apply(\n",
    "        lambda r: r[f'{q_node_id}_unified_curie'] in r[\"overlap_curie_combo\"], axis=1\n",
    "    )\n",
    "cooccurrence_df = results_with_figures_df[s == True].sort_values(\n",
    "    \"p_value\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### By TRAPI result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "<div>\n",
       "    <span style=\"font-weight: bold;\">\n",
       "        TRAPI result NCBIGene:5872, NCBIGene:22931, MESH:D002784\n",
       "    </span> \n",
       "    <span>\n",
       "        (cumulative containment: 139.67)\n",
       "    </span>\n",
       "    <div style=\"font-size: small;\">\n",
       "        \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB18</span>\n",
       " - entity_negatively_regulated_by_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">Cholesterol</span>\n",
       "\n",
       "    </div>\n",
       "</div>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div style=\"display: flex; flex-flow: row wrap; text-align: center;\">\n",
       "    \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4747172/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4747172/bin/oncotarget-06-41535-g001.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                Mechanisms for autophagy modulation by...<br>\n",
       "                p-value: 1.6e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2636993/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2636993/bin/zh00010958160009.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                RST inhibits the first step of...<br>\n",
       "                p-value: 1.6e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5837599/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5837599/bin/kfx074f3.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                Typical structure and content of exosomes<br>\n",
       "                p-value: 1.7e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4804636/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4804636/bin/nihms-715982-f0002.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                Contrasting Lifestyles Within the Host Cell<br>\n",
       "                p-value: 1.8e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6691912/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6691912/bin/TSTA_A_1629835_F0001_OC.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                EV production procedure.Exosomes, 50–150...<br>\n",
       "                p-value: 1.9e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "    </div>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "...\n",
      "\n",
      "plus 187 more figures\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<div>\n",
       "    <span style=\"font-weight: bold;\">\n",
       "        TRAPI result NCBIGene:5872, NCBIGene:201475, MESH:D019821\n",
       "    </span> \n",
       "    <span>\n",
       "        (cumulative containment: 131.00)\n",
       "    </span>\n",
       "    <div style=\"font-size: small;\">\n",
       "        \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_negatively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB12</span>\n",
       " - entity_positively_regulated_by_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">Simvastatin</span>\n",
       "<br>\n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB12</span>\n",
       " - entity_positively_regulated_by_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">Simvastatin</span>\n",
       "\n",
       "    </div>\n",
       "</div>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div style=\"display: flex; flex-flow: row wrap; text-align: center;\">\n",
       "    \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7680335/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7680335/bin/10555_2020_9902_Fig2_HTML.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                Effect of prenylation inhibition on...<br>\n",
       "                p-value: 2.0e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6404178/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6404178/bin/JAH3-7-e008543-g001.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                The HMG‐CoA (3‐hydroxy‐3‐methylglutaryl...<br>\n",
       "                p-value: 2.2e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4246236/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4246236/bin/viruses-06-04505-g001.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                Cholesterol metabolism, lipid rafts and...<br>\n",
       "                p-value: 2.2e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5902023/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5902023/bin/molecules-22-00886-g001.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                Mammalian isoprenoid biosynthetic pathway...<br>\n",
       "                p-value: 2.3e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
       "            <figure style=\"margin: 5px !important;\">\n",
       "                <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6018057/\">\n",
       "                  <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6018057/bin/nihms974743f1.jpg\" style=\"height: 100px\">\n",
       "              </a>\n",
       "            </figure>\n",
       "            <div style=\"font-size: x-small;\">\n",
       "                The mevalonate pathway and statin mechanism<br>\n",
       "                p-value: 2.3e-08, \n",
       "                containment: 1.0\n",
       "            </div>\n",
       "        </div>\n",
       "        \n",
       "    </div>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "...\n",
      "\n",
      "plus 185 more figures\n",
      "\n",
      "...\n",
      "\n",
      "plus 2161 more TRAPI results, each with figures\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from IPython.display import Image\n",
    "from IPython.core.display import HTML \n",
    "\n",
    "\n",
    "outer_limit = 2\n",
    "inner_limit = 5\n",
    "\n",
    "count = 0\n",
    "for trapi_result_curie_combo, one_trapi_result_curie_combo_df in cooccurrence_df.groupby(\n",
    "    \"trapi_result_curie_combo\", sort=False\n",
    "):\n",
    "    trapi_curie_names = []\n",
    "    for curie in one_trapi_result_curie_combo_df[\"trapi_result_curie_combo\"].tolist()[0]:\n",
    "        name = curie_to_name[curie]\n",
    "        trapi_curie_names.append(name)\n",
    "\n",
    "    cumulative_trapi_result_curie_combo_figure_score = list(set(\n",
    "        one_trapi_result_curie_combo_df[\"cumulative_trapi_result_curie_combo_figure_score\"]\n",
    "    ))[0]\n",
    "    \n",
    "    ordered_trapi_result_curie_combo = []\n",
    "    for unified_curie_column in unified_curie_columns:\n",
    "        ordered_trapi_result_curie_combo.append(list(set(\n",
    "            one_trapi_result_curie_combo_df[unified_curie_column]\n",
    "        ))[0])\n",
    "\n",
    "    overlap_names = []\n",
    "    for curie in one_trapi_result_curie_combo_df[\"overlap_curie_combo\"].tolist()[0]:\n",
    "        name = curie_to_name[curie]\n",
    "        overlap_names.append(name)\n",
    "\n",
    "    trapi_result_rows = []\n",
    "    for i,row in one_trapi_result_curie_combo_df[trapi_result_columns].drop_duplicates().iterrows():\n",
    "        trapi_result_chunks = []\n",
    "        for trapi_result_column in trapi_result_columns:\n",
    "            row_value = row[trapi_result_column]\n",
    "            if row_value in overlap_names:\n",
    "                trapi_result_chunk = f'''\n",
    "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">{row_value}</span>\n",
    "'''\n",
    "            else:\n",
    "                trapi_result_chunk = f'{row_value}'\n",
    "            trapi_result_chunks.append(trapi_result_chunk)\n",
    "        trapi_result_rows.append(\n",
    "            f'{\" - \".join(trapi_result_chunks)}'\n",
    "        )\n",
    "    trapi_result_row_count = len(trapi_result_rows)\n",
    "    \n",
    "\n",
    "    display(HTML(data=f'''\n",
    "<div>\n",
    "    <span style=\"font-weight: bold;\">\n",
    "        TRAPI result {\", \".join(ordered_trapi_result_curie_combo)}\n",
    "    </span> \n",
    "    <span>\n",
    "        (cumulative containment: {cumulative_trapi_result_curie_combo_figure_score:.2f})\n",
    "    </span>\n",
    "    <div style=\"font-size: small;\">\n",
    "        {\"<br>\".join(trapi_result_rows)}\n",
    "    </div>\n",
    "</div>\n",
    "'''))\n",
    "\n",
    "    figures = []\n",
    "    row_height='100px'\n",
    "    for figure_id, one_figure_id_df in one_trapi_result_curie_combo_df.groupby(\"figure_id\", sort=False):\n",
    "        [pmc,filename] = figure_id.split(\"__\")\n",
    "        paper_url = f\"https://www.ncbi.nlm.nih.gov/pmc/articles/{pmc}/\"\n",
    "        figure_url = f\"https://www.ncbi.nlm.nih.gov/pmc/articles/{pmc}/bin/{filename}\"\n",
    "        figure_title = list(set(one_figure_id_df[\"figure_title\"]))[0]\n",
    "        figure_title_limited = textwrap.shorten(figure_title, width=45, placeholder=\"...\")\n",
    "\n",
    "        p_value = list(set(\n",
    "            one_figure_id_df[\"p_value\"]\n",
    "        ))[0]\n",
    "        score = list(set(one_figure_id_df[\"score\"]))[0]\n",
    "\n",
    "        # TODO: why does the commented out block below not display correctly\n",
    "        # on GitHub? The figcaption appears to the right of the image.\n",
    "        # The div solution below is a kludge.\n",
    "        figures.append(f'''\n",
    "        <div style=\"margin: 5px !important; padding: 5px !important; border: 1px solid slategray;>\n",
    "            <figure style=\"margin: 5px !important;\">\n",
    "                <a target=\"_blank\" href=\"{paper_url}\">\n",
    "                  <img src=\"{figure_url}\" style=\"height: {row_height}\">\n",
    "              </a>\n",
    "            </figure>\n",
    "            <div style=\"font-size: x-small;\">\n",
    "                {figure_title_limited}<br>\n",
    "                p-value: {p_value:.1e}, \n",
    "                containment: {score:.2}\n",
    "            </div>\n",
    "        </div>\n",
    "        ''')\n",
    "        \n",
    "#        figures.append(f'''\n",
    "#            <figure style=\"margin: 5px !important;\">\n",
    "#                <a target=\"_blank\" href=\"{paper_url}\">\n",
    "#                  <img src=\"{figure_url}\" style=\"height: {row_height}\">\n",
    "#                </a>\n",
    "#                <figcaption style=\"font-size: small;\">\n",
    "#                      {figure_title_limited}<br>\n",
    "#                      p-value: {p_value:.1e}, \n",
    "#                      containment: {score:.2}\n",
    "#                </figcaption>\n",
    "#            </figure>\n",
    "#        ''')\n",
    "\n",
    "    figure_count = len(figures)\n",
    "    display(HTML(data=f'''\n",
    "    <div style=\"display: flex; flex-flow: row wrap; text-align: center;\">\n",
    "    {''.join(figures[:inner_limit])}\n",
    "    </div>\n",
    "'''))\n",
    "\n",
    "    if figure_count > inner_limit:\n",
    "        print(f\"...\")\n",
    "        print(\"\")\n",
    "        print(f\"plus {figure_count - inner_limit} more figures\")\n",
    "        print(\"\")\n",
    "\n",
    "\n",
    "    count += 1\n",
    "    if count >= outer_limit:\n",
    "        print(\"...\")\n",
    "        print(\"\")\n",
    "        remaining = len(set(cooccurrence_df[\"trapi_result_curie_combo\"])) - outer_limit\n",
    "        print(f'plus {remaining} more TRAPI results, each with figures')\n",
    "        print(\"\")\n",
    "        break"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### By figure"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "<div style=\"border: 1px solid slategray; padding: 6px;\">\n",
       "    <figure style=\"margin: 5px !important;\">\n",
       "        <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4747172/\">\n",
       "          <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4747172/bin/oncotarget-06-41535-g001.jpg\" style=\"max-height: 500px; max-width: 100%;\">\n",
       "        </a>\n",
       "    </figure>\n",
       "      Mechanisms for autophagy modulation by isoprenoid biosynthetic pathway inhibitors in multiple myeloma cells\n",
       "</div>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<div>\n",
       "    <span style=\"font-weight: bold; font-size: small;\">\n",
       "        TRAPI result NCBIGene:5872, NCBIGene:22931, MESH:D002784\n",
       "    </span> \n",
       "    (p-value: 1.6e-08 containment: 1.0)\n",
       "    <br>\n",
       "    \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB18</span>\n",
       " - entity_negatively_regulated_by_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">Cholesterol</span>\n",
       "</span>\n",
       "</div>\n",
       "<br>\n",
       "<div>\n",
       "    <span style=\"font-weight: bold; font-size: small;\">\n",
       "        TRAPI result NCBIGene:5872, NCBIGene:201475, MESH:C511404\n",
       "    </span> \n",
       "    (p-value: 1.9e-05 containment: 0.67)\n",
       "    <br>\n",
       "    \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB12</span>\n",
       " - associated_with_resistance_to - Zibotentan</span><br>\n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_negatively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB12</span>\n",
       " - associated_with_resistance_to - Zibotentan</span>\n",
       "</div>\n",
       "<br>\n",
       "<div>\n",
       "    <span style=\"font-weight: bold; font-size: small;\">\n",
       "        TRAPI result NCBIGene:5872, NCBIGene:201475, MESH:C522973\n",
       "    </span> \n",
       "    (p-value: 1.9e-05 containment: 0.67)\n",
       "    <br>\n",
       "    \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_negatively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB12</span>\n",
       " - associated_with_sensitivity_to - PI-103</span><br>\n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_negatively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB12</span>\n",
       " - associated_with_resistance_to - PI-103</span><br>\n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB12</span>\n",
       " - associated_with_sensitivity_to - PI-103</span><br>\n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB12</span>\n",
       " - associated_with_resistance_to - PI-103</span>\n",
       "</div>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "...\n",
      "\n",
      "plus 280 more TRAPI results\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<div style=\"border: 1px solid slategray; padding: 6px;\">\n",
       "    <figure style=\"margin: 5px !important;\">\n",
       "        <a target=\"_blank\" href=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2636993/\">\n",
       "          <img src=\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2636993/bin/zh00010958160009.jpg\" style=\"max-height: 500px; max-width: 100%;\">\n",
       "        </a>\n",
       "    </figure>\n",
       "      RST inhibits the first step of cholesterol/isoprenoid biosynthesis by blocking 3-hydroxy-3-methyl-glutaryl-coenzyme A (HMG-CoA) reductase, thus reducing the formation of mevalonate from HMG-CoA\n",
       "</div>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "\n",
       "<div>\n",
       "    <span style=\"font-weight: bold; font-size: small;\">\n",
       "        TRAPI result NCBIGene:5872, NCBIGene:22931, MESH:D002784\n",
       "    </span> \n",
       "    (p-value: 1.6e-08 containment: 1.0)\n",
       "    <br>\n",
       "    \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB18</span>\n",
       " - entity_negatively_regulated_by_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">Cholesterol</span>\n",
       "</span>\n",
       "</div>\n",
       "<br>\n",
       "<div>\n",
       "    <span style=\"font-weight: bold; font-size: small;\">\n",
       "        TRAPI result NCBIGene:5872, NCBIGene:22931, MESH:C521047\n",
       "    </span> \n",
       "    (p-value: 2.0e-05 containment: 0.67)\n",
       "    <br>\n",
       "    \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB18</span>\n",
       " - associated_with_resistance_to - Amuvatinib</span>\n",
       "</div>\n",
       "<br>\n",
       "<div>\n",
       "    <span style=\"font-weight: bold; font-size: small;\">\n",
       "        TRAPI result NCBIGene:5872, NCBIGene:22931, MESH:C561310\n",
       "    </span> \n",
       "    (p-value: 2.0e-05 containment: 0.67)\n",
       "    <br>\n",
       "    \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB18</span>\n",
       " - associated_with_sensitivity_to - 2-Cyclohexyl-N-(1-isopropylpiperidin-4-yl)-6-methoxy-7-(3-(pyrrolidin-1-yl)propoxy)quinazolin-4-amine</span><br>\n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB13</span>\n",
       " - entity_positively_regulates_entity - \n",
       "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">RAB18</span>\n",
       " - associated_with_resistance_to - 2-Cyclohexyl-N-(1-isopropylpiperidin-4-yl)-6-methoxy-7-(3-(pyrrolidin-1-yl)propoxy)quinazolin-4-amine</span>\n",
       "</div>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "...\n",
      "\n",
      "plus 281 more TRAPI results\n",
      "\n",
      "...\n",
      "\n",
      "\n",
      "plus 198 more figures, each with TRAPI results\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from IPython.display import Image\n",
    "from IPython.core.display import HTML \n",
    "\n",
    "\n",
    "outer_limit = 2\n",
    "inner_limit = 3\n",
    "\n",
    "count = 0\n",
    "for figure_id, one_figure_id_df in cooccurrence_df.groupby(\"figure_id\", sort=False):\n",
    "    [pmc,filename] = figure_id.split(\"__\")\n",
    "    paper_url = f\"https://www.ncbi.nlm.nih.gov/pmc/articles/{pmc}/\"\n",
    "    figure_url = f\"https://www.ncbi.nlm.nih.gov/pmc/articles/{pmc}/bin/{filename}\"\n",
    "    figure_title = list(set(one_figure_id_df[\"figure_title\"]))[0]\n",
    "\n",
    "    display(HTML(data=f'''\n",
    "<div style=\"border: 1px solid slategray; padding: 6px;\">\n",
    "    <figure style=\"margin: 5px !important;\">\n",
    "        <a target=\"_blank\" href=\"{paper_url}\">\n",
    "          <img src=\"{figure_url}\" style=\"max-height: 500px; max-width: 100%;\">\n",
    "        </a>\n",
    "    </figure>\n",
    "      {figure_title}\n",
    "</div>\n",
    "'''))\n",
    "\n",
    "    trapi_results_row_sets = []\n",
    "    \n",
    "    for trapi_result_curie_combo, one_trapi_result_curie_combo_df in one_figure_id_df[(\n",
    "        trapi_result_columns +\n",
    "        unified_curie_columns +\n",
    "        [\"trapi_result_curie_combo\", \"overlap_curie_combo\", \"p_value\", \"score\"]\n",
    "    )].drop_duplicates().groupby(\"trapi_result_curie_combo\", sort=False):\n",
    "        overlap_names = []\n",
    "        for curie in one_trapi_result_curie_combo_df[\"overlap_curie_combo\"].tolist()[0]:\n",
    "            name = curie_to_name[curie]\n",
    "            overlap_names.append(name)\n",
    "\n",
    "        trapi_results_rows = []\n",
    "        for i,row in one_trapi_result_curie_combo_df.iterrows():\n",
    "            trapi_result_chunks = []\n",
    "            for trapi_result_column in trapi_result_columns:\n",
    "                row_value = row[trapi_result_column]\n",
    "                if row_value in overlap_names:\n",
    "                    trapi_result_chunk = f'''\n",
    "<span style=\"border: 1px orange solid; padding: 1px; border-radius: 4px;\">{row_value}</span>\n",
    "'''\n",
    "                else:\n",
    "                    trapi_result_chunk = f'{row_value}'\n",
    "                trapi_result_chunks.append(trapi_result_chunk)\n",
    "                \n",
    "            trapi_results_rows.append(\n",
    "                f'{\" - \".join(trapi_result_chunks)}</span>'\n",
    "            )\n",
    "        \n",
    "        ordered_trapi_result_curie_combo = []\n",
    "        for unified_curie_column in unified_curie_columns:\n",
    "            ordered_trapi_result_curie_combo.append(list(set(\n",
    "                one_trapi_result_curie_combo_df[unified_curie_column]\n",
    "            ))[0])\n",
    "        \n",
    "        p_value = list(set(one_trapi_result_curie_combo_df[\"p_value\"]))[0]\n",
    "        score = list(set(one_trapi_result_curie_combo_df[\"score\"]))[0]\n",
    "        trapi_results_row_sets.append(f'''\n",
    "<div>\n",
    "    <span style=\"font-weight: bold; font-size: small;\">\n",
    "        TRAPI result {\", \".join(ordered_trapi_result_curie_combo)}\n",
    "    </span> \n",
    "    (p-value: {p_value:.1e} containment: {score:.2})\n",
    "    <br>\n",
    "    {\"<br>\".join(trapi_results_rows)}\n",
    "</div>\n",
    "''')\n",
    "\n",
    "    trapi_results_row_set_count = len(trapi_results_row_sets)\n",
    "    display(HTML(data=\"<br>\".join(trapi_results_row_sets[:inner_limit])))\n",
    "    if trapi_results_row_set_count > inner_limit:\n",
    "        print(f\"...\")\n",
    "        print(\"\")\n",
    "        print(f\"plus {trapi_results_row_set_count - inner_limit} more TRAPI results\")\n",
    "    print(\"\")\n",
    "\n",
    "    count += 1\n",
    "    if count >= outer_limit:\n",
    "        print(\"...\")\n",
    "        print(\"\")\n",
    "        print(f'''\n",
    "plus {len(set(cooccurrence_df[\"figure_id\"])) - outer_limit} more figures, each with TRAPI results\n",
    "''')\n",
    "        print(\"\")\n",
    "        break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "eBphe1RvE1He"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2O1Ocl1QE1Ji"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "XaQvH-tQ1Zt0"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "bte_clustering.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}