Link_Prediction_on_a_Heterogeneous_Graph.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3UzfrQpio80w"
      },
      "source": [
        "The project is based on [Hetionet](https://het.io/about/)\n",
        "which is a heterogeneous biomedical knowledge graph. It encodes relationships uncovered by millions of studies conducted over the last half-century into a single resource.\n",
        "\n",
        "\n",
        "the json data can be found [here](https://github.com/hetio/hetionet/tree/master/hetnet/json)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KTstuH15zUXP"
      },
      "source": [
        "![image.png](data:image/png;base64,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)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "L-1hWSln0Rwm"
      },
      "source": [
        "I'm only using Disease, Gene and Compound meta nodes and the relationships between them.\n",
        "\n",
        "The target is to predict new edges between Diseases and Genes (Disease-Gene associations)\n",
        "\n",
        "and between those two node types (Disease, Gene) I'm only using the association edge type to make sure there is no information leak from other edge types"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Nq49La3S-iGp"
      },
      "source": [
        "I used two different approaches for encoding the graph\n",
        "\n",
        "*   building a normal homogeneous GNN and transform it with to_hetero function (basically makes a layer for every edge_type) like this [tutorial](https://pytorch-geometric.readthedocs.io/en/latest/notes/heterogeneous.html) \n",
        "*   Used RGCN module which is based on this [paper](https://arxiv.org/abs/1703.06103)\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "623P39bHjqud"
      },
      "source": [
        "[colab link](https://colab.research.google.com/drive/1fhF8h7xJoYaKr4ZPLrToN0e5sDhiiXKX?usp=sharing)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ZbfhsbNzXMv-"
      },
      "outputs": [],
      "source": [
        "\"\"\"\n",
        "those versions were guaranteed to work on colab on cpu(which is not slow in this case) \n",
        "colab might crash if I tried to use them on gpu (but locally on my machine latest versions work fine)\n",
        "\n",
        "\"\"\"\n",
        "!pip install torch==1.10.0+cu111  -f https://download.pytorch.org/whl/torch_stable.html\n",
        "!pip install torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric -f https://data.pyg.org/whl/torch-1.10.0+cpu.html"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7L9RRFc4dmMX",
        "outputId": "1d3c93bb-5091-4475-a55f-eadde555770e"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1AkInoTO1SNlJnKEyYi7SpiBxaS2ecc4q\n",
            "To: /content/torchHetnetData.zip\n",
            "\r  0% 0.00/3.18M [00:00<?, ?B/s]\r100% 3.18M/3.18M [00:00<00:00, 225MB/s]\n"
          ]
        }
      ],
      "source": [
        "#this just loads the the zip file containing torch tensors I processed for the model\n",
        "!gdown  1AkInoTO1SNlJnKEyYi7SpiBxaS2ecc4q "
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "SH0B_2V9d4cH"
      },
      "outputs": [],
      "source": [
        "!unzip  torchHetnetData.zip"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "JsPbTyMzVRQ3"
      },
      "outputs": [],
      "source": [
        "from torch_geometric.utils import (\n",
        "    add_self_loops,\n",
        "    negative_sampling,\n",
        "    remove_self_loops ,\n",
        ")\n",
        "import torch.nn.functional as F\n",
        "from sklearn.metrics import accuracy_score,roc_auc_score,average_precision_score\n",
        "from sklearn.metrics import precision_score\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")\n",
        "import time\n",
        "\n",
        "\n",
        "import numpy as np\n",
        "import scipy.sparse as sp\n",
        "from torch import nn\n",
        "import pandas as pd\n",
        "# import networkx as nx\n",
        "from typing import List , Dict,Tuple\n",
        "from torch_geometric.nn import SAGEConv, to_hetero\n",
        "import torch\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import torch_geometric.nn as  gnn\n",
        "from torch_geometric.data  import Data\n",
        "import pandas as pd\n",
        "from sklearn.preprocessing import LabelEncoder\n",
        "\n",
        "if torch.cuda.is_available():  \n",
        "  dev = \"cuda:0\" \n",
        "  print(\"gpu up\")\n",
        "else:  \n",
        "  dev = \"cpu\"  \n",
        "device = torch.device(dev)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Kvt1Ti26VRQ5"
      },
      "outputs": [],
      "source": [
        "\"\"\"\n",
        "this cell transforms the json graph into torch tensors compatible with HeteroData Object\n",
        "\"\"\"\n",
        "# !wget \"https://media.githubusercontent.com/media/hetio/hetionet/master/hetnet/json/hetionet-v1.0.json.bz2\"\n",
        "# !bzip2 -dk /content/hetionet-v1.0.json.bz2\n",
        "\n",
        "# import json\n",
        "# with open(\"/content/hetionet-v1.0.json\") as f: \n",
        "#   hetnet =  json.load(f)\n",
        "\n",
        "# nodes = pd.DataFrame(hetnet[\"nodes\"])\n",
        "\n",
        "\n",
        "# nodeTypeEncoder = {}\n",
        "# for t in nodes.kind.unique():\n",
        "#   tNodes = nodes[nodes.kind == t].reset_index()\n",
        "#   enc = LabelEncoder()\n",
        "#   enc.fit(tNodes.identifier)\n",
        "#   nodeTypeEncoder[t] = enc\n",
        "  \n",
        "# graphDict = {}\n",
        "\n",
        "# for tup in meta_graph[\"metaedge_tuples\"]:\n",
        "#   # edgeKey = f'{meta_graph[\"kind_to_abbrev\"][tup[0]]}_{meta_graph[\"kind_to_abbrev\"][tup[2]]}_{meta_graph[\"kind_to_abbrev\"][tup[1]]}'\n",
        "#   edgeKey = f'{tup[0]}_{tup[2]}_{tup[1]}'\n",
        "#   revEdgeKey = f'{tup[1]}_{tup[2]}_{tup[0]}'\n",
        "#   graphDict[edgeKey] = []\n",
        "#   if tup[-1] == \"both\":\n",
        "#      graphDict[revEdgeKey] = []\n",
        "\n",
        "\n",
        "# for edge in hetnet[\"edges\"]:\n",
        "#   edgeKey = f'{edge[\"source_id\"][0]}_{edge[\"kind\"]}_{edge[\"target_id\"][0]}'\n",
        "#   # src = nodeTypeEncoder[edge[\"source_id\"][0]].transform([edge[\"source_id\"][1]])[0]\n",
        "#   src = list(nodeTypeEncoder[edge[\"source_id\"][0]].classes_).index(edge[\"source_id\"][1])\n",
        "#   revEdgeKey = f'{edge[\"target_id\"][0]}_{edge[\"kind\"]}_{edge[\"source_id\"][0]}'\n",
        "#   # print(edge)\n",
        "#   trg = list(nodeTypeEncoder[edge[\"target_id\"][0]].classes_).index(edge[\"target_id\"][1])\n",
        "#   graphDict[edgeKey].append([src,trg])\n",
        "#   if edge[\"direction\"] == 'both':\n",
        "#     graphDict[revEdgeKey].append([trg,src])\n",
        "\n",
        "\n",
        "# import numpy as np\n",
        "# for k,v  in graphDict.items():\n",
        "#   np.save(\"./hetnetData/\"+k+\".npy\",np.array(v))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sJg88T7oVRQ8",
        "outputId": "205e315d-6d15-4b60-d34f-869f9c7a2076"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Gene 20945\n",
            "Compound 1552\n",
            "Disease 137\n"
          ]
        }
      ],
      "source": [
        "from torch_geometric.data import HeteroData\n",
        "import numpy as np\n",
        "import os \n",
        "import torch\n",
        "\n",
        "\n",
        "data = HeteroData()\n",
        "path = \"./torchHetnetData\"\n",
        "edge_types = []\n",
        "for f in os.listdir(path):\n",
        "  # print(f)\n",
        "  name = f[:-3].split(\"_\") \n",
        "  if len(name) == 1:\n",
        "    data[name[0]].x = torch.load(os.path.join(path,f))\n",
        "    print(name[0] , len(data[name[0]].x))\n",
        "    continue\n",
        "  if name[0] not in [\"Gene\",\"Compound\",\"Disease\"]: continue\n",
        "  if name[2] not in [\"Gene\",\"Compound\",\"Disease\"]: continue\n",
        "  # I removed any relations ships between disease and genes that are not the target edge\n",
        "  #  \n",
        "  if name[0] == \"Gene\" and name[2] == \"Disease\":\n",
        "      if name[1] != \"associates\": continue \n",
        "  if name[0] == \"Disease\" and name[2] == \"Gene\":\n",
        "      if name[1] != \"associates\": continue \n",
        "  e =  torch.load(os.path.join(path,f))\n",
        "  if len(e[0]) <= 1000 : continue # disregard any edge type with less than 1000 instances \n",
        "  edge_types.append(name[1])\n",
        "  data[name[0],name[1],name[2]].edge_index = e\n",
        "\n",
        "from torch_geometric.transforms import RandomLinkSplit\n",
        "\n",
        "\n",
        "transform = RandomLinkSplit(is_undirected=False,add_negative_train_samples=True,disjoint_train_ratio=0.35,edge_types=[(\"Disease\",\"associates\",\"Gene\")],rev_edge_types=[(\"Gene\",\"associates\",\"Disease\")])\n",
        "\n",
        "train,valid,test = transform(data) \n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iiZ3acReVRQ_",
        "outputId": "5dfe11ee-b6c5-4000-e59f-ae8f6adf7d9d"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "12623.0\n",
            "12623\n"
          ]
        }
      ],
      "source": [
        "print(  \n",
        "  len(train[\"Disease\",\"associates\",\"Gene\"].edge_index[0])+ # train visible message passing edges \n",
        "  len(train[\"Disease\",\"associates\",\"Gene\"].edge_label) /2  + # train positive supervision edges \n",
        "  len(valid[\"Disease\",\"associates\",\"Gene\"].edge_label) /2 +  # validation positive evaluation edges\n",
        "  len(test[\"Disease\",\"associates\",\"Gene\"].edge_label) /2  )  # test positive evaluation edges\n",
        "print(len(data[\"Disease\",\"associates\",\"Gene\"].edge_index[0]) )  # all existing target edges in the graph"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ochEOSEaVRRB",
        "outputId": "950a1562-46b6-493b-db1f-3872d4dbf344"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "[('Compound', 'upregulates', 'Gene'),\n",
              " ('Gene', 'associates', 'Disease'),\n",
              " ('Gene', 'downregulates', 'Compound'),\n",
              " ('Disease', 'resembles', 'Disease'),\n",
              " ('Compound', 'resembles', 'Compound'),\n",
              " ('Gene', 'regulates', 'Gene'),\n",
              " ('Compound', 'downregulates', 'Gene'),\n",
              " ('Gene', 'binds', 'Compound'),\n",
              " ('Gene', 'interacts', 'Gene'),\n",
              " ('Compound', 'binds', 'Gene'),\n",
              " ('Disease', 'associates', 'Gene'),\n",
              " ('Gene', 'covaries', 'Gene'),\n",
              " ('Gene', 'upregulates', 'Compound')]"
            ]
          },
          "execution_count": 12,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "data.edge_types"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nZeyFg4IVRRF",
        "outputId": "bdd1f3bf-c0ae-4920-e8e8-a6919059d9da"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Proxy(getattr_1)\n"
          ]
        }
      ],
      "source": [
        "\n",
        "\"\"\"\n",
        "the GAE-like model follows the one here on this tutorial \n",
        "\"\"\"\n",
        "class Encoder(nn.Module):\n",
        "  \"\"\"\n",
        "  basic gnn with two layers to get representations for every node  \n",
        "  \"\"\"\n",
        "  def __init__(self,node_dim=64):\n",
        "    super(Encoder,self).__init__()   \n",
        "    self.gconv1  =  gnn.SAGEConv((-1,-1),  node_dim)\n",
        "    self.gconv2  = gnn.SAGEConv((-1,-1), node_dim)\n",
        "\n",
        "  def forward(self,x,edge_index):\n",
        "    h = self.gconv1(x,edge_index).relu()\n",
        "    h = self.gconv2(h,edge_index)\n",
        "    return h\n",
        "\n",
        "\n",
        "EPS = 1e-15\n",
        "\n",
        "class EdgeDecoder(torch.nn.Module):\n",
        "    def __init__(self, hidden_channels):\n",
        "        super().__init__()\n",
        "        self.lin1 = nn.Linear(2 * hidden_channels, hidden_channels)\n",
        "        self.lin2 = nn.Linear(hidden_channels, 1)\n",
        "\n",
        "    def forward(self, z_dict, edge_label_index):\n",
        "        z = torch.cat([z_dict['Disease'][edge_label_index[0]], z_dict['Gene'][edge_label_index[1]]], dim=-1)\n",
        "\n",
        "        z = self.lin1(z).relu()\n",
        "        z = self.lin2(z).sigmoid()\n",
        "        return z.view(-1)\n",
        "\n",
        "\n",
        "class MyGAE(nn.Module):\n",
        "\n",
        "  def __init__(self,encoder,decoder,data:HeteroData,node_types:List[Tuple],node_dim:int):\n",
        "    super(MyGAE,self).__init__()\n",
        "    self.node_types = {}\n",
        "    for n in node_types:\n",
        "      self.node_types[n[0]]  = nn.Embedding(n[1],node_dim) # this part is redundant(we can use embeddings from torch geometric directly)\n",
        "      self.node_types[n[0]].to(device)\n",
        "    self.hetero_encoder =to_hetero(encoder,data.metadata())\n",
        "    self.decoder = decoder\n",
        "\n",
        "  def encode(self,nodes, edges):\n",
        "    xVectors= {}\n",
        "    for i in self.node_types.keys():\n",
        "      xVectors[i] = self.node_types[i](nodes[i])\n",
        "    return self.hetero_encoder(xVectors,edges)\n",
        "    \n",
        "  def decode(self,h,edge_label_index):\n",
        "    \n",
        "    return self.decoder(h,edge_label_index)\n",
        "\n",
        "  def recon_loss(self, z, pos_edge_index, neg_edge_index):\n",
        "        r\"\"\"Given latent variables :obj:`z`, computes the binary cross\n",
        "        entropy loss for positive edges :obj:`pos_edge_index` and negative\n",
        "        sampled edges.\n",
        "        \"\"\"\n",
        "\n",
        "        pos_loss = -torch.log(\n",
        "            self.decode(z, pos_edge_index) + EPS).mean()\n",
        "\n",
        "        # Do not include self-loops in negative samples\n",
        "        # pos_edge_index, _ = remove_self_loops(pos_edge_index)\n",
        "        # pos_edge_index, _ = add_self_loops(pos_edge_index)\n",
        "        neg_edge_index, _ = remove_self_loops(neg_edge_index)\n",
        "        \n",
        "        neg_loss = -torch.log(1 -\n",
        "                              self.decode(z, neg_edge_index) +\n",
        "                              EPS).mean()\n",
        "\n",
        "        return pos_loss + neg_loss\n",
        "\n",
        "  def test(self,z, edge_index, y ):\n",
        "    r\"\"\"Given latent variables :obj:`z`, edges\n",
        "    computes area under the ROC curve (AUC) , average precision (AP)\n",
        "    and (Acc) accuracy scores  .\n",
        "    Args:\n",
        "        z (Tensor): The latent space :math:`\\mathbf{Z}`.\n",
        "        edge_index (LongTensor):   edges\n",
        "        y  (Tensor):labels\n",
        "    \"\"\"\n",
        "    pred =  self.decode(z, edge_index,)\n",
        "\n",
        "    y, pred = y.detach().cpu().numpy(), pred.detach().cpu().numpy()\n",
        "\n",
        "    return roc_auc_score(y, pred), precision_score(y, pred>=0.5),accuracy_score(y,pred>0.5)\n",
        "node_types = [(\"Compound\",1552),(\"Gene\",20945),(\"Disease\",137)]\n",
        "\n",
        "myGAE = MyGAE(Encoder(),EdgeDecoder(64),data,node_types,node_dim=64)\n",
        "myGAE.to(device)\n",
        "train = train.to(device)\n",
        "with torch.no_grad():\n",
        "    myGAE.encode(train.x_dict, train.edge_index_dict)\n",
        "\n",
        "\n",
        "optimizer = torch.optim.Adam(myGAE.parameters(), lr=0.01, weight_decay=0.0005)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "id": "HfyG3ThUVRRH",
        "outputId": "c29fea17-70e7-4d98-a721-da88ef147fb6"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "validationMetrics = []\n",
        "for i in range(40):\n",
        "    myGAE.train()\n",
        "    optimizer.zero_grad()\n",
        "    # pred = myGAE(train.x_dict, train.edge_index_dict,\n",
        "    #              train['Gene', \"associates\",'Disease'].edge_label_index)\n",
        "    target = train['Disease',\"associates\", 'Gene'].edge_label\n",
        "    label_edge_index = train['Disease',\"associates\", 'Gene'].edge_label_index\n",
        "    h = myGAE.encode(train.x_dict,train.edge_index_dict)\n",
        "    loss  = myGAE.recon_loss(h,label_edge_index[:,target==1],label_edge_index[:,target==0])\n",
        "    loss.backward()\n",
        "    optimizer.step()\n",
        "    auc, p ,acc= myGAE.test(h, train['Disease',\"associates\", 'Gene'].edge_label_index,train['Disease',\"associates\", 'Gene'].edge_label)\n",
        "    # print(i,\"train\", np.array([auc, p , acc]).round(3),end=' ')\n",
        "    \n",
        "    with torch.no_grad():\n",
        "      myGAE.eval()\n",
        "      test.to(device)\n",
        "      z = myGAE.encode(test.x_dict,test.edge_index_dict)\n",
        "      auc, p ,acc= myGAE.test(z, test['Disease',\"associates\", 'Gene'].edge_label_index,test['Disease',\"associates\", 'Gene'].edge_label)\n",
        "      # auc, p = myGAE.test(z, valid.edge_label_index[:,valid.edge_label==1],valid.edge_label_index[:,valid.edge_label==0])\n",
        "      # print(\"valid\",np.array([auc, p , acc]).round(3))\n",
        "      validationMetrics.append([auc, p , acc])\n",
        "\n",
        "plt.plot(np.arange(len(validationMetrics)),np.array(validationMetrics)[:,0],label='auc')\n",
        "plt.plot(np.arange(len(validationMetrics)),np.array(validationMetrics)[:,1],label='precision')\n",
        "plt.plot(np.arange(len(validationMetrics)),np.array(validationMetrics)[:,2],label='acc')\n",
        "plt.legend()\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "KqMoQOFjVRRJ"
      },
      "outputs": [],
      "source": [
        "from torch_geometric.data import Data\n",
        "from typing import Tuple,List\n",
        "\n",
        "\"\"\"\n",
        "RGCN model requires a homogeneous \"typed\" graph\n",
        "with a  tensor for edge types   \n",
        "this function is the same as applying .to_homogeneous on a HeteroData object\n",
        "but it's useful if I need to transform each node type a certain way \n",
        "\n",
        "\"\"\"\n",
        "\n",
        "def  convert_hetero_object_to_RGCN(heteroData:HeteroData,edge_types:List[Tuple[str]]):\n",
        "    dataRGCN = Data()\n",
        "    datax  =  torch.arange(len(heteroData[\"Disease\"].x) +len(heteroData[\"Gene\"].x) + len(heteroData[\"Compound\"].x))\n",
        "    data_edge_index = torch.tensor([]).reshape(2,0)\n",
        "    data_edge_types = torch.tensor([])\n",
        "    # edge_types = []   \n",
        "    for _, (key,v) in enumerate(heteroData.edge_index_dict.copy().items()):\n",
        "        ourV = torch.tensor(v.cpu().numpy())\n",
        "        if key[0] == \"Compound\":\n",
        "            ourV[0,:] = ourV[0]+len(heteroData[\"Disease\"].x) + len(heteroData[\"Gene\"].x) \n",
        "        if key[0] == \"Gene\":\n",
        "            ourV[0,:] = ourV[0]+len(heteroData[\"Disease\"].x)\n",
        "        \n",
        "        if key[1] == \"Compound\":\n",
        "            ourV[1,:] = ourV[1]+len(heteroData[\"Disease\"].x) + len(heteroData[\"Gene\"].x)\n",
        "        if key[1] == \"Gene\":\n",
        "            ourV[1,:] = ourV[1]+len(heteroData[\"Disease\"].x)\n",
        "        data_edge_index = torch.cat([data_edge_index,ourV],dim = -1)\n",
        "        data_edge_types = torch.cat([data_edge_types,torch.zeros(len(ourV[0])) + edge_types.index(key)])\n",
        "    dataRGCN.edge_index = data_edge_index\n",
        "    dataRGCN.x = datax\n",
        "    edge_label_index =torch.tensor( heteroData[\"Disease\",\"associates\",\"Gene\"].edge_label_index.cpu().numpy())\n",
        "    edge_label_index[1,:] += len(heteroData[\"Disease\"].x)  \n",
        "    edge_label = torch.tensor(heteroData[\"Disease\",\"associates\",\"Gene\"].edge_label.cpu().numpy())\n",
        "    target_label_type = torch.zeros(len(edge_label)) +data.edge_types.index((\"Disease\",\"associates\",\"Gene\"))\n",
        "    dataRGCN.edge_label_index = edge_label_index\n",
        "    dataRGCN.edge_label = edge_label\n",
        "    dataRGCN.edge_type = data_edge_types\n",
        "    dataRGCN.label_edge_type = target_label_type\n",
        "    return  dataRGCN\n",
        "\n",
        "\n",
        "\n",
        "rgcn_valid = convert_hetero_object_to_RGCN(valid,data.edge_types)\n",
        "rgcn_train = convert_hetero_object_to_RGCN(train,data.edge_types)\n",
        "rgcn_test = convert_hetero_object_to_RGCN(test,data.edge_types)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ijueiBXQiOkV",
        "outputId": "34644570-1759-46f6-eca7-0b2fe6595a34"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "Data(edge_index=[2, 811786], x=[22634], edge_label_index=[2, 6184], edge_label=[6184], edge_type=[811786], label_edge_type=[6184])"
            ]
          },
          "execution_count": 19,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "rgcn_train"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "3hFrW21HVRRL"
      },
      "outputs": [],
      "source": [
        "from torch import nn\n",
        "class RGCNEncoder(nn.Module):\n",
        "\n",
        "  def __init__(self,num_nodes,num_relations,node_dim=32):\n",
        "    super(RGCNEncoder,self).__init__()   \n",
        "    self.gconv1 = gnn.RGCNConv(node_dim, node_dim, num_relations,num_blocks=16)\n",
        "    self.gconv2 = gnn.RGCNConv(node_dim, node_dim, num_relations,num_blocks=16)\n",
        "  def forward(self,x,edge_index:torch.Tensor,edge_type:torch.Tensor):\n",
        "    h = self.gconv1(x,edge_index,edge_type).relu()\n",
        "    # print(h.shape)\n",
        "    # return h\n",
        "    h = self.gconv2(h,edge_index,edge_type)\n",
        "    return h\n",
        "\n",
        "\n",
        "EPS = 1e-15\n",
        "\n",
        "class RGCNLinkPredictor(torch.nn.Module):\n",
        "    def __init__(self, hidden_channels):\n",
        "        super().__init__()\n",
        "        self.lin1 = nn.Linear(2 * hidden_channels, hidden_channels)\n",
        "        self.lin2 = nn.Linear(hidden_channels, 1)\n",
        "\n",
        "    def forward(self, z, edge_label_index):\n",
        "        z = torch.cat([z[edge_label_index[0]], z[edge_label_index[1]]], dim=-1)\n",
        "\n",
        "        z = self.lin1(z).relu()\n",
        "        z = self.lin2(z).sigmoid()\n",
        "        return z.view(-1)\n",
        "\n",
        "\n",
        "class RGCNModel(nn.Module):\n",
        "\n",
        "  def __init__(self,encoder:RGCNEncoder,decoder:RGCNLinkPredictor,node_dim:int):\n",
        "    super(RGCNModel,self).__init__()\n",
        "    \n",
        "    self.nodeVectors = nn.Embedding(len(rgcn_test.x),node_dim)\n",
        "    self.nodeVectors.to(device)\n",
        "    self.encoder =encoder\n",
        "    self.decoder = decoder\n",
        "\n",
        "  def encode(self,x, edge_index,edge_types):\n",
        "    x = self.nodeVectors(x)\n",
        "    return self.encoder(x,edge_index,edge_types)\n",
        "    \n",
        "  def decode(self,h,edge_label_index):\n",
        "    \n",
        "    return self.decoder(h,edge_label_index)\n",
        "\n",
        "  def recon_loss(self, z, pos_edge_index, neg_edge_index):\n",
        "        r\"\"\"Given latent variables :obj:`z`, computes the binary cross\n",
        "        entropy loss for positive edges :obj:`pos_edge_index` and negative\n",
        "        sampled edges.\n",
        "        \"\"\"\n",
        "\n",
        "        pos_loss = -torch.log(\n",
        "            self.decode(z, pos_edge_index,) + EPS).mean()\n",
        "\n",
        "        # Do not include self-loops in negative samples\n",
        "        # pos_edge_index, _ = remove_self_loops(pos_edge_index)\n",
        "        # pos_edge_index, _ = add_self_loops(pos_edge_index)\n",
        "        neg_edge_index, _ = remove_self_loops(neg_edge_index)\n",
        "        \n",
        "        neg_loss = -torch.log(1 -\n",
        "                              self.decode(z, neg_edge_index) +\n",
        "                              EPS).mean()\n",
        "\n",
        "        return pos_loss + neg_loss\n",
        "\n",
        "  def test(self,z, edge_index, y ):\n",
        "    r\"\"\"Given latent variables :obj:`z`, edges\n",
        "    computes area under the ROC curve (AUC) , average precision (AP)\n",
        "    and (Acc) accuracy scores  .\n",
        "    Args:\n",
        "        z (Tensor): The latent space :math:`\\mathbf{Z}`.\n",
        "        edge_index (LongTensor):   edges\n",
        "        y  (Tensor):labels\n",
        "    \"\"\"\n",
        "    pred =  self.decode(z, edge_index)\n",
        "\n",
        "    y, pred = y.detach().cpu().numpy(), pred.detach().cpu().numpy()\n",
        "\n",
        "    return roc_auc_score(y, pred), precision_score(y, pred>=0.5),accuracy_score(y,pred>0.5)\n",
        "\n",
        "rgcn = RGCNModel(RGCNEncoder(len(rgcn_train.x),len(data.edge_types),node_dim=64),RGCNLinkPredictor(64),64).to(\"cpu\")\n",
        "rgcn.to(device)\n",
        "optimizer = torch.optim.Adam(rgcn.parameters(), lr=0.001, weight_decay=0.0005)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "id": "__7ClP2QVRRP",
        "outputId": "901a6af2-9ec9-49dd-b1b6-a7dee88652dc"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "validationMetrics = []\n",
        "for i in range(50):\n",
        "    rgcn.train()\n",
        "    optimizer.zero_grad()\n",
        "    # pred = rgcn(train.x_dict, train.edge_index_dict,\n",
        "    #              train['Gene', \"associates\",'Disease'].edge_label_index)\n",
        "    target = rgcn_train.edge_label\n",
        "    label_edge_index = rgcn_train.edge_label_index\n",
        "    rgcn_train.to(device)\n",
        "    h = rgcn.encode(rgcn_train.x.long(),rgcn_train.edge_index.long(),rgcn_train.edge_type.long())\n",
        "    loss  = rgcn.recon_loss(h,label_edge_index[:,target==1],label_edge_index[:,target==0])\n",
        "    loss.backward()\n",
        "    optimizer.step()\n",
        "\n",
        "    rgcn.eval()\n",
        "    auc, p ,acc= rgcn.test(h, rgcn_train.edge_label_index,rgcn_train.edge_label)\n",
        "    # print(i, \"train\" ,np.array([auc, p , acc,loss.item()]).round(3),end=\" \")\n",
        "\n",
        "    with torch.no_grad():\n",
        "        rgcn_test.to(device)\n",
        "        z = rgcn.encode(rgcn_test.x.long(),rgcn_test.edge_index.long(),rgcn_test.edge_type.long())\n",
        "        auc, p ,acc= rgcn.test(z, rgcn_test.edge_label_index,rgcn_test.edge_label)\n",
        "        # auc, p = rgcn.test(z, valid.edge_label_index[:,valid.edge_label==1],valid.edge_label_index[:,valid.edge_label==0])\n",
        "        validationMetrics.append([auc, p , acc])\n",
        "        # print(\"valid\",np.array([auc, p , acc]).round(3))\n",
        "\n",
        "plt.plot(np.arange(len(validationMetrics)),np.array(validationMetrics)[:,0],label='auc')\n",
        "plt.plot(np.arange(len(validationMetrics)),np.array(validationMetrics)[:,1],label='precision')\n",
        "plt.plot(np.arange(len(validationMetrics)),np.array(validationMetrics)[:,2],label='acc')\n",
        "plt.legend()\n",
        "plt.show()"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "collapsed_sections": [],
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3.10.4 ('gnns')",
      "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": "3.8.13"
    },
    "orig_nbformat": 4,
    "vscode": {
      "interpreter": {
        "hash": "65071604a904ffcbac85598bd4b7cd132c32dd5e96ad1769b5378a4247c1d236"
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}