model_intepret.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "f5f1e7f4",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import dgl\n",
    "import time\n",
    "import torch\n",
    "import seaborn\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "import argparse\n",
    "from warnings import simplefilter\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.model_selection import KFold, StratifiedKFold\n",
    "from torch.utils.data import TensorDataset\n",
    "# from load_data import load_data, load_graph, remove_graph\n",
    "#     get_data_loaders,, topk_filtering\n",
    "from load_data import topk_filtering\n",
    "from model import Model\n",
    "from utils import get_metrics, get_metrics_auc, set_seed, \\\n",
    "    plot_result_auc, plot_result_aupr, checkpoint"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "e543593a",
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_data(dataset, k):\n",
    "    drug_drug = pd.read_csv('./dataset/{}/drug_drug.csv'.format(dataset), header=None)\n",
    "    drug_drug_link = topk_filtering(drug_drug.values, k)\n",
    "    disease_disease = pd.read_csv('./dataset/{}/disease_disease.csv'.format(dataset), header=None)\n",
    "    disease_disease_link = topk_filtering(disease_disease.values, k)\n",
    "    drug_disease = pd.read_csv('./dataset/{}/drug_disease.csv'.format(dataset), header=None)\n",
    "    drug_disease_link = np.array(np.where(drug_disease == 1)).T\n",
    "    disease_drug_link = np.array(np.where(drug_disease.T == 1)).T\n",
    "    links = {'drug-drug': drug_drug_link, 'drug-disease': drug_disease_link,\n",
    "             'disease-disease': disease_disease_link}\n",
    "    graph_data = {('drug', 'drug-drug', 'drug'): (torch.tensor(drug_drug_link[:, 0]),\n",
    "                                                  torch.tensor(drug_drug_link[:, 1])),\n",
    "                  ('drug', 'drug-disease', 'disease'): (torch.tensor(drug_disease_link[:, 0]),\n",
    "                                                        torch.tensor(drug_disease_link[:, 1])),\n",
    "                  ('disease', 'disease-drug', 'drug'): (torch.tensor(disease_drug_link[:, 0]),\n",
    "                                                        torch.tensor(disease_drug_link[:, 1])),\n",
    "                  ('disease', 'disease-disease', 'disease'): (torch.tensor(disease_disease_link[:, 0]),\n",
    "                                                              torch.tensor(disease_disease_link[:, 1]))}\n",
    "    g = dgl.heterograph(graph_data)\n",
    "    drug_feature = np.hstack((drug_drug.values, np.zeros(drug_disease.shape)))\n",
    "    dis_feature = np.hstack((np.zeros(drug_disease.T.shape), disease_disease.values))\n",
    "    g.nodes['drug'].data['h'] = torch.from_numpy(drug_feature).to(torch.float32)\n",
    "    g.nodes['disease'].data['h'] = torch.from_numpy(dis_feature).to(torch.float32)\n",
    "    if '{}_temp_{}k'.format(dataset, k) in os.listdir():\n",
    "        print('Load data and label(It takes time)...')\n",
    "        data = np.load('{}_temp_{}k/data.npy'.format(dataset, k))\n",
    "        label = np.load('{}_temp_{}k/label.npy'.format(dataset, k))\n",
    "    print('Data prepared !')\n",
    "    return g, data, label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "4601eadf",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Some settings to load the pre-trained model with accurate path\n",
    "seed = 0\n",
    "batch_size = 1024\n",
    "k = 5\n",
    "nfold = 5\n",
    "aggregate_type = 'BiTrans'\n",
    "hidden_feats = 128\n",
    "learning_rate = 0.001\n",
    "epoch = 2000\n",
    "topk = 3\n",
    "num_layer = 1\n",
    "dropout = 0.4\n",
    "batch_norm = False\n",
    "path = 'result/C-dataset'\n",
    "# Define the query dataset, drug, and disease\n",
    "dataset = 'C-dataset'\n",
    "drug_id = 241\n",
    "disease_id = 280"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "8e5c28eb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Load data and label(It takes time)...\n",
      "Data prepared !\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No handles with labels found to put in legend.\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Top5 Attentions: [0.03501033 0.03501033 0.03318125 0.03022725 0.03003906]\n",
      "Top5 Meta-path Instances:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[241,  15, 280, 280],\n",
       "       [241,  15, 280, 280],\n",
       "       [241, 143,  25, 280],\n",
       "       [241, 281,  25, 280],\n",
       "       [241,  33,  25, 280]])"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load data\n",
    "g, data, label = load_data(dataset, k)\n",
    "data = torch.tensor(data)\n",
    "label = torch.tensor(label)\n",
    "feature = {'drug': g.nodes['drug'].data['h'],\n",
    "           'disease': g.nodes['disease'].data['h']}\n",
    "# Load pre-trained MilGNet model\n",
    "model_d = []\n",
    "attns = []\n",
    "for idx in range(nfold):\n",
    "    model = Model(g.etypes,\n",
    "            {'drug': feature['drug'].shape[1], \n",
    "             'disease': feature['disease'].shape[1]},\n",
    "                      hidden_feats=hidden_feats,\n",
    "                      num_emb_layers=num_layer,\n",
    "                      agg_type=aggregate_type,\n",
    "                      dropout=dropout,\n",
    "                      bn=batch_norm,\n",
    "                      k=topk)\n",
    "    model.load_state_dict(\n",
    "        torch.load('result/C-dataset/'\n",
    "                   f'k{k}_topk{topk}_nl{num_layer}_ep{epoch}_'\n",
    "                   f'lr{learning_rate}_dp{dropout}_bs{batch_size}_'\n",
    "                   f'hf{hidden_feats}_{seed}/model_{idx+1}.pkl'))\n",
    "    model_d.append(model)\n",
    "# Define the data to be used for generating the instance attentions\n",
    "pred_data = data[(drug_id)*g.num_nodes('disease')+\n",
    "                 disease_id].unsqueeze(dim=0)\n",
    "attns = np.zeros(k*k+2*k+1)\n",
    "for model in model_d:\n",
    "    model.eval()\n",
    "    pred, attn = model(g, feature, pred_data)\n",
    "    attns += attn.detach().numpy().squeeze() / nfold\n",
    "# Visualize the attention coefficient distribution   \n",
    "seaborn.set(style='white')\n",
    "seaborn.histplot(attn.detach().numpy()[0], palette='Set2')\n",
    "plt.xlabel('Attention Coefficients')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "# Inferences for the topK meta-path instances\n",
    "topk_idx = np.argsort(attns)[-5:][::-1]\n",
    "print('Top5 Attentions: {}'.format(attns[topk_idx]))\n",
    "print('Top5 Meta-path Instances:')\n",
    "metapath = pred_data.numpy().squeeze()\n",
    "attncoef = attn.detach().numpy()[0]\n",
    "pred_data.numpy().squeeze()[topk_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "d7e3b298",
   "metadata": {
    "scrolled": true
   },
   "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>Drug1</th>\n",
       "      <th>Drug2</th>\n",
       "      <th>Disease1</th>\n",
       "      <th>Disease2</th>\n",
       "      <th>Attn</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00158</td>\n",
       "      <td>D608456</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.036804</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00158</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.035363</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00158</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.035363</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00642</td>\n",
       "      <td>D114500</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.035153</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00158</td>\n",
       "      <td>D175100</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.034215</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00650</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.033997</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00158</td>\n",
       "      <td>D114500</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.033320</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00384</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.032033</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00384</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.032033</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00188</td>\n",
       "      <td>D114500</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.030808</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00384</td>\n",
       "      <td>D133239</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.030302</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00642</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.029778</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00642</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.029778</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00188</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.029326</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00188</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.029326</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00384</td>\n",
       "      <td>D114500</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.029125</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00158</td>\n",
       "      <td>D133239</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.028900</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00188</td>\n",
       "      <td>D608456</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.027864</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00642</td>\n",
       "      <td>D175100</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.027738</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00188</td>\n",
       "      <td>D133239</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.027272</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00642</td>\n",
       "      <td>D133239</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.026812</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D175100</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.025025</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D175100</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.025025</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D114500</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.025004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D114500</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.025004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00188</td>\n",
       "      <td>D175100</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.024801</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00384</td>\n",
       "      <td>D175100</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.024644</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D133239</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.021990</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D133239</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.021990</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.021840</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.021840</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.021840</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D276300</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.021840</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D608456</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.021424</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00563</td>\n",
       "      <td>D608456</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.021424</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>DB00563</td>\n",
       "      <td>DB00384</td>\n",
       "      <td>D608456</td>\n",
       "      <td>D276300</td>\n",
       "      <td>0.020999</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      Drug1    Drug2 Disease1 Disease2      Attn\n",
       "16  DB00563  DB00158  D608456  D276300  0.036804\n",
       "1   DB00563  DB00158  D276300  D276300  0.035363\n",
       "15  DB00563  DB00158  D276300  D276300  0.035363\n",
       "32  DB00563  DB00642  D114500  D276300  0.035153\n",
       "14  DB00563  DB00158  D175100  D276300  0.034215\n",
       "6   DB00563  DB00650  D276300  D276300  0.033997\n",
       "12  DB00563  DB00158  D114500  D276300  0.033320\n",
       "3   DB00563  DB00384  D276300  D276300  0.032033\n",
       "25  DB00563  DB00384  D276300  D276300  0.032033\n",
       "17  DB00563  DB00188  D114500  D276300  0.030808\n",
       "23  DB00563  DB00384  D133239  D276300  0.030302\n",
       "5   DB00563  DB00642  D276300  D276300  0.029778\n",
       "35  DB00563  DB00642  D276300  D276300  0.029778\n",
       "20  DB00563  DB00188  D276300  D276300  0.029326\n",
       "2   DB00563  DB00188  D276300  D276300  0.029326\n",
       "22  DB00563  DB00384  D114500  D276300  0.029125\n",
       "13  DB00563  DB00158  D133239  D276300  0.028900\n",
       "21  DB00563  DB00188  D608456  D276300  0.027864\n",
       "34  DB00563  DB00642  D175100  D276300  0.027738\n",
       "18  DB00563  DB00188  D133239  D276300  0.027272\n",
       "33  DB00563  DB00642  D133239  D276300  0.026812\n",
       "9   DB00563  DB00563  D175100  D276300  0.025025\n",
       "29  DB00563  DB00563  D175100  D276300  0.025025\n",
       "27  DB00563  DB00563  D114500  D276300  0.025004\n",
       "7   DB00563  DB00563  D114500  D276300  0.025004\n",
       "19  DB00563  DB00188  D175100  D276300  0.024801\n",
       "24  DB00563  DB00384  D175100  D276300  0.024644\n",
       "8   DB00563  DB00563  D133239  D276300  0.021990\n",
       "28  DB00563  DB00563  D133239  D276300  0.021990\n",
       "10  DB00563  DB00563  D276300  D276300  0.021840\n",
       "30  DB00563  DB00563  D276300  D276300  0.021840\n",
       "4   DB00563  DB00563  D276300  D276300  0.021840\n",
       "0   DB00563  DB00563  D276300  D276300  0.021840\n",
       "11  DB00563  DB00563  D608456  D276300  0.021424\n",
       "31  DB00563  DB00563  D608456  D276300  0.021424\n",
       "26  DB00563  DB00384  D608456  D276300  0.020999"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Ensemble all generated meta-path instances\n",
    "# and their attention attention coefficients\n",
    "dr = pd.read_csv(f'dataset/{dataset}/drug.csv')\n",
    "di = pd.read_csv(f'dataset/{dataset}/disease.csv')\n",
    "mpac = []\n",
    "for i in range(len(metapath)):\n",
    "    new = []\n",
    "    for j in range(2):\n",
    "        new.append(dr.loc[metapath[i, j]].values[1])\n",
    "    for j in range(2, 4):\n",
    "        new.append(di.loc[metapath[i, j]].values[1])\n",
    "    new.append(attncoef[i][0])\n",
    "    mpac.append(new)\n",
    "df_mp = pd.DataFrame(np.array(mpac),\n",
    "                     columns=['Drug1', 'Drug2', \n",
    "                              'Disease1', 'Disease2', 'Attn'])\n",
    "df_mp['Attn'] = df_mp['Attn'].astype(float)\n",
    "df_mp = df_mp.sort_values(by='Attn', ascending=False)\n",
    "df_mp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c52f68df",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}