peptide_mhc_presentation_classifiers.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4LQ21TT7AbQc"
      },
      "source": [
        "# MHCI - Peptides Presentation\n",
        "Author: **Gabriel Schwab**\n",
        "\n",
        "### Problem\n",
        "Peptides are short chains of amino acids that can be presented on the surface of cells by MHC proteins.\n",
        "MHC class I (MHC1) proteins are expressed on the surface of most cells and present peptides to T cells,\n",
        "allowing the immune system to monitor the cell for signs of infection or abnormal activity. MHC1\n",
        "proteins are generated by genetic variations in the MHC1 genes, known as alleles, and each individual has\n",
        "a unique set of MHC1 proteins.\n",
        "\n",
        "In this notebook, we present our method to build classifiers that predicts if a given peptide will be presented by a\n",
        "specific MHC1 protein, identified by its allele name.\n",
        "\n",
        "### Data\n",
        "\n",
        "The dataset to use is extracted from the training and evaluation data of NetMHCPan4.1, a framework developed in Reynisson et al., 2020, available [here](https://services.healthtech.dtu.dk/services/NetMHCpan-4.1/).\n",
        "\n",
        "### Outline\n",
        "This notebook is divided in four sections.\n",
        "\n",
        "**1. Dataset Analysis**\n",
        "\n",
        "In this section, we examine the dataset's integrity, explore its statistics, and identify potential issues.\n",
        "\n",
        "**2. Method Explanation**\n",
        "\n",
        "Here, we introduce three different approaches to address the problem: two classical machine learning methods and one deep learning method.\n",
        "\n",
        "**3. Experiment Description**\n",
        "\n",
        "This section describes a series of experiments to evaluate the performance of our classifiers, highlighting their advantages and drawbacks.\n",
        "\n",
        "**4. Result Analysis**\n",
        "\n",
        "Finally, we summarize and provide a comprehensive analysis of the results."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-E6d95u7F9AX"
      },
      "source": [
        "## 1. Dataset Analysis"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-2xZRkMujYgd",
        "outputId": "258a8ea7-f486-4ecd-cbd5-e20503c6dd1e"
      },
      "outputs": [],
      "source": [
        "# If you are working from Google Colab, uncomment this\n",
        "# from google.colab import drive\n",
        "# drive.mount('/content/drive')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "Jf98yfS37ezt"
      },
      "outputs": [],
      "source": [
        "# Change `ROOT_DIR` variable to point on where your dataset folder is located\n",
        "ROOT_DIR = ''"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "2Zbq6E71jT8N"
      },
      "outputs": [],
      "source": [
        "import os\n",
        "import time\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "import xgboost as xgb\n",
        "from sklearn.ensemble import RandomForestClassifier\n",
        "from sklearn.metrics import RocCurveDisplay, classification_report, accuracy_score, roc_curve, auc\n",
        "from tqdm import tqdm\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "import torch.nn.functional as F\n",
        "from sklearn.model_selection import train_test_split\n",
        "from torch.utils.data import DataLoader, Dataset"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YsAB8jIqqvJy"
      },
      "source": [
        "Let's first check the shape and the content of the whole train dataset."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KQDk9AVclS5n",
        "outputId": "10de005c-5965-42d8-fac5-a9da86f74592"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(3679405, 3)\n",
            "<class 'pandas.core.frame.DataFrame'>\n",
            "Index: 3679405 entries, 0 to 735688\n",
            "Data columns (total 3 columns):\n",
            " #   Column   Dtype \n",
            "---  ------   ----- \n",
            " 0   peptide  object\n",
            " 1   allele   object\n",
            " 2   hit      int64 \n",
            "dtypes: int64(1), object(2)\n",
            "memory usage: 112.3+ MB\n",
            "None\n"
          ]
        }
      ],
      "source": [
        "train_data = pd.concat([pd.read_csv(os.path.join(ROOT_DIR, f'netmhcpan41_data/fold_{i}.csv')) for i in range(5)])\n",
        "test_data = pd.read_csv(os.path.join(ROOT_DIR, 'netmhcpan41_data/test.csv'))\n",
        "print(train_data.shape)\n",
        "print(train_data.info())"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TdSheR04owJ7"
      },
      "source": [
        "Now let's check the integrity of the dataset."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9o_ehRAdpb4k",
        "outputId": "149b3e4a-39cf-4eb6-8ef1-ff3f73609460"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "peptide    0\n",
            "allele     0\n",
            "hit        0\n",
            "dtype: int64\n",
            "peptide    0\n",
            "allele     0\n",
            "hit        0\n",
            "dtype: int64\n"
          ]
        }
      ],
      "source": [
        "print(train_data.isnull().sum())\n",
        "print(train_data.isna().sum())"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "a5cbbwTbp7ML"
      },
      "source": [
        "As we can see, the train dataset is clean, there is non missing values or NaN.\n",
        "Now we will explore some of its statistics, to better understand the data."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 484
        },
        "id": "yLpdC09BrRGc",
        "outputId": "ce6d4bc7-4962-482b-d605-804e6fd46eb6"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Number of unique peptides: 3584329\n",
            "Number of unique alleles: 130\n"
          ]
        },
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABKUAAAGwCAYAAAB4ntbEAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABpyklEQVR4nO3deVhV5fr/8c8GZFAER8AZcwRn0RTLMRWNPFFmahY4ZKcCU7EsGxzrUJZTR9M8J6XJNCut1JxQ9Kg4oZhDeuykYio4wk5SUFi/P/qxv24BBYS9Ud6v61pXrfXca6372WzYj/de61kmwzAMAQAAAAAAADbkYO8EAAAAAAAAUPpQlAIAAAAAAIDNUZQCAAAAAACAzVGUAgAAAAAAgM1RlAIAAAAAAIDNUZQCAAAAAACAzVGUAgAAAAAAgM1RlAIAAAAAAIDNUZQCAAAAAACAzVGUAuxg8ODB8vX1tXcayEV0dLRMJpOOHz9us3MeP35cJpNJ0dHRxX6u3Prn6+urRx55pNjPLUmxsbEymUyKjY21yfkAAHeGMUvpVZDxCe+T/2PLcZXEOBJ3P4pSwA1MJlO+lpL2hzD7D3T24uLiIm9vb3Xp0kX/+Mc/dO7cuUIf+9ChQ5o4caJNizS3smjRIs2cOTPf8bf6oMx+3b755ptbHuOjjz4q0Af9jT8LJycnVapUSQEBARo5cqQOHTqU7+PcTkHzsqWSnBsA3AsYs+R0L4xZbnxtvLy81LFjRy1btqz4klTB87SFiRMnymQy6fz58/ZOJVfF+V5jHFmyc0PRMxmGYdg7CaCk+OKLL6zWP/vsM61bt06ff/651fYePXrI29u70Oe5du2asrKy5OLiUuhj3Cg2NlZdu3bVSy+9pLZt2yozM1Pnzp3Ttm3b9OOPP8rT01Nff/21unXrVuBjf/PNN+rXr582btyoLl26FEm+d+KRRx7RgQMH8j0I8PX1VdOmTbVixYocbdmv29KlS/XEE09IkjIzM3Xt2jW5uLjIZDJJkpo2baoqVarke2BvMpnUo0cPhYaGyjAMpaamat++fVq6dKnS0tL03nvvKTIy0hJvGIbS09NVpkwZOTo65uschckrr/7d6jUqrLxyy8rKUkZGhpydneXgwPciAFBYjFlyuhfGLBUrVtSYMWMkSadPn9bHH3+s3377TXPnztXzzz9v0zwLMj4ZPHiwYmNji6xIM3HiRE2aNEnnzp1TlSpViuSYRelW77U7HVcxjmQcWdo42TsBoCR5+umnrda3b9+udevW5dh+sz///FNly5bN93nKlClTqPxup2PHjpbiSrZ9+/apZ8+e6tu3rw4dOqRq1aoVy7nvFY6OjgX6QM9Lw4YNc7xv3n33XfXp00djxoxR48aN9fDDD0v6a/Dh6up6x+e8lbS0NJUrV67I+ldYDg4Oxd5XACgNGLPcm2rUqGH1MwwNDVX9+vU1Y8aMYitK5cUW4xPkjnFk7hhH3psoLwIF1KVLFzVt2lTx8fHq1KmTypYtq9dff12S9P333ys4OFjVq1eXi4uL6tWrpylTpigzM9PqGDffd599L/gHH3yg+fPnq169enJxcVHbtm21a9euO8q3RYsWmjlzplJSUjR79mzL9hMnTujFF19Uo0aN5ObmpsqVK6tfv35W33BFR0erX79+kqSuXbvmuBUgv/09evSo+vbtKx8fH7m6uqpmzZoaMGCAUlNTreK++OILBQQEyM3NTZUqVdKAAQN08uRJS3uXLl20cuVKnThxwpJLUc9fcPO98r6+vjp48KA2bdpkOWdhv32tXLmyFi9eLCcnJ73zzjuW7bnNBZCUlKQhQ4aoZs2acnFxUbVq1fToo4/mK6/sPmzatEkvvviivLy8VLNmzVz7d6O1a9eqZcuWcnV1lb+/v7777jur9uxL6e/kNctrLoClS5dafvZVqlTR008/rVOnTlnFDB48WO7u7jp16pRCQkLk7u6uqlWr6uWXX87xngMAMGa5F8YsPj4+8vPz07FjxyzbTp06paFDh8rb21suLi5q0qSJFixYYLVf9uftkiVL9Prrr8vHx0flypXT3/72t3znmddcRcuXL1fTpk3l6uqqpk2b5nl7YVZWlmbOnKkmTZrI1dVV3t7e+vvf/65Lly4V+HXIy+HDh/XEE0+oUqVKcnV1VZs2bfTDDz9YxWSPU7Zu3arIyEhVrVpV5cqV02OPPZbjdtGsrCxNnDhR1atXV9myZdW1a1cdOnRIvr6+Gjx4sOV4t3qvZduyZYvuv/9+ubq66r777tNnn312R31lHMk48l7FlVJAIVy4cEG9e/fWgAED9PTTT1sui4+Ojpa7u7siIyPl7u6uDRs2aPz48TKbzXr//fdve9xFixbpjz/+0N///neZTCZNnTpVjz/+uH777bc7+qbyiSee0LBhw7R27VrLh9iuXbu0bds2DRgwQDVr1tTx48c1d+5cdenSRYcOHVLZsmXVqVMnvfTSS/rwww/1+uuvy8/PT5Is/81PfzMyMhQUFKT09HSNGDFCPj4+OnXqlFasWKGUlBR5enpKkt555x299dZbevLJJ/Xss8/q3Llz+uc//6lOnTpp7969qlChgt544w2lpqbq999/14wZMyRJ7u7ut+3/tWvXcp2T4OYBZm5mzpypESNGyN3dXW+88YYk3dFtELVr11bnzp21ceNGmc1meXh45BrXt29fHTx4UCNGjJCvr6/Onj2rdevWKTExUb6+vvnK68UXX1TVqlU1fvx4paWl3TKvo0ePqn///nr++ecVFhamhQsXql+/flq9erV69OhRoD4W9DWLjo7WkCFD1LZtW0VFRSk5OVmzZs3S1q1bLT/7bJmZmQoKClK7du30wQcfaP369Zo2bZrq1aunF154oUB5AkBpwJjl7hqz3OzatWs6efKkKleuLElKTk5W+/btZTKZFBERoapVq+qnn37SsGHDZDabNWrUKKv933nnHZlMJr366qs6e/asZs6cqe7duyshIUFubm4FznPt2rXq27ev/P39FRUVpQsXLliKHzf7+9//bvmMf+mll3Ts2DHNnj1be/fu1datW+/4KryDBw/qgQceUI0aNfTaa6+pXLly+vrrrxUSEqJvv/1Wjz32mFX8iBEjVLFiRU2YMEHHjx/XzJkzFRERoSVLllhixo0bp6lTp6pPnz4KCgrSvn37FBQUpKtXr1pibvdek6Rff/3V8l4OCwvTggULNHjwYAUEBKhJkyaF7jPjyJwYR94DDAB5Cg8PN27+NencubMhyZg3b16O+D///DPHtr///e9G2bJljatXr1q2hYWFGXXq1LGsHzt2zJBkVK5c2bh48aJl+/fff29IMn788cdb5rlx40ZDkrF06dI8Y1q0aGFUrFjxlrnGxcUZkozPPvvMsm3p0qWGJGPjxo054vPT37179942t+PHjxuOjo7GO++8Y7V9//79hpOTk9X24OBgq9fudurUqWNIuuVyY24LFy40JBnHjh2zbGvSpInRuXPnfJ9TkhEeHp5n+8iRIw1Jxr59+wzD+L+f/8KFCw3DMIxLly4Zkoz333//lufJK6/sPjz44IPG9evXc227sX/Zr9G3335r2ZaammpUq1bNaNWqlWXbhAkTcvw+5HXMvHLLfq9mv58yMjIMLy8vo2nTpsaVK1cscStWrDAkGePHj7dsCwsLMyQZkydPtjpmq1atjICAgBznAoDShDHLvTFm6dmzp3Hu3Dnj3Llzxr59+4wBAwYYkowRI0YYhmEYw4YNM6pVq2acP3/eat8BAwYYnp6eln5mv841atQwzGazJe7rr782JBmzZs26bZ43j08MwzBatmxpVKtWzUhJSbFsW7t2rSHJ6hj/+c9/DEnGl19+aXXM1atX57r9ZtljjnPnzuUZ89BDDxnNmjWzer9mZWUZHTp0MBo0aGDZlj1O6d69u5GVlWXZPnr0aMPR0dHSl6SkJMPJyckICQmxOs/EiRMNSUZYWJhl263ea9njqs2bN1u2nT171nBxcTHGjBlzy34bBuPIW+XGOPLexO17QCG4uLhoyJAhOba7ublZ/v+PP/7Q+fPn1bFjR/355586fPjwbY/bv39/VaxY0bLesWNHSdJvv/12xzm7u7vrjz/+yDXXa9eu6cKFC6pfv74qVKigPXv25OuY+elv9reKa9as0Z9//pnrcb777jtlZWXpySef1Pnz5y2Lj4+PGjRooI0bNxa4vzdq166d1q1bl2P54IMP7ui4hZX9DeSNP48bubm5ydnZWbGxsXd0ifvw4cPzfd9/9erVrb5R9PDwUGhoqPbu3aukpKRC53A7u3fv1tmzZ/Xiiy9azREQHBysxo0ba+XKlTn2uXlOjY4dOxbJ7wgA3IsYs+Q8Rkkes6xdu1ZVq1ZV1apV1aJFCy1dulTPPPOM3nvvPRmGoW+//VZ9+vSRYRhW5w8KClJqamqO1yM0NFTly5e3rD/xxBOqVq2aVq1aVeDczpw5o4SEBIWFhVleK+mvyfT9/f2tYpcuXSpPT0/16NHDKs+AgAC5u7vf8et08eJFbdiwQU8++aTl53n+/HlduHBBQUFBOnr0aI7bt5577jmr28c6duyozMxMnThxQpIUExOj69ev68UXX7Tab8SIEQXOz9/f3/I7IUlVq1ZVo0aNiuz3Q2IcKTGOvFeU6tv3Nm/erPfff1/x8fE6c+aMli1bppCQkAIdwzAMTZs2TfPnz9eJEydUpUoVvfjii5ZLDXFvqlGjhpydnXNsP3jwoN58801t2LBBZrPZqi0/t4rVrl3baj17sFcU995fvnzZalBy5coVRUVFaeHChTp16pSMGx7EmZ9cpfz1t27duoqMjNT06dP15ZdfqmPHjvrb3/6mp59+2jKgOXr0qAzDUIMGDXI9z51e3l2lShV17949x3YnJ/v8Cbx8+bIkWf08buTi4qL33ntPY8aMkbe3t9q3b69HHnlEoaGh8vHxyfd56tatm+/Y+vXr57jPv2HDhpL+mqugIOctiOyBYKNGjXK0NW7cWFu2bLHa5urqqqpVq1ptq1ixYpHOTwEA9xLGLH+5W8Ys7dq109tvvy2TyaSyZcvKz8/PcvvR2bNnlZKSovnz52v+/Pm57n/27Fmr9ZvzNJlMql+/fqGekpf9mZ1b3xs1amRVEDt69KhSU1Pl5eWVrzwL6tdff5VhGHrrrbf01ltv5XmOGjVqWNZv957N7l/9+vWt4ipVqmRVgM2Pm8+Vfb6i+v2QGEdKjCPvFaW6KJWWlqYWLVpo6NChevzxxwt1jJEjR2rt2rX64IMP1KxZM128eFEXL14s4kxR0tz4bVu2lJQUde7cWR4eHpo8ebLq1asnV1dX7dmzR6+++qqysrJue9y8vo24cfBVGNeuXdN///tfNW3a1LJtxIgRWrhwoUaNGqXAwEB5enrKZDJpwIAB+cq1IP2dNm2aBg8erO+//15r167VSy+9pKioKG3fvl01a9ZUVlaWTCaTfvrpp1xfg8LMwVCSHThwQI6Ojrf8sB81apT69Omj5cuXa82aNXrrrbcUFRWlDRs2qFWrVvk6T27v0zuR2+SUkmw6OaQ9n/gCAHcjxix315glry/SJFnyfPrppxUWFpZrTPPmze/o/EUlKytLXl5e+vLLL3Ntv7kwUJjjS9LLL7+soKCgXGNuLi4V13s2N8V5LsaRhcc4smQq1UWp3r17q3fv3nm2p6en64033tBXX32llJQUNW3aVO+9955l9v9ffvlFc+fO1YEDByzV2YJUlHFviY2N1YULF/Tdd9+pU6dOlu03Pi3FXr755htduXLF6kP7m2++UVhYmKZNm2bZdvXqVaWkpFjtm9cHSEH726xZMzVr1kxvvvmmtm3bpgceeEDz5s3T22+/rXr16skwDNWtW9fyrUpe8sqnOBXlORMTE7Vp0yYFBgbm+Q1Xtnr16mnMmDEaM2aMjh49qpYtW2ratGn64osvijyv7G8cbzzmf//7X0myPIUn+1vClJQUq0kjs7+lulF+c6tTp44k6ciRI+rWrZtV25EjRyztAICiw5jlL3fbmKVq1aoqX768MjMz8yxc3ezo0aNW64Zh6Ndff7UqXhX0M/vmY0p/fWbfqF69elq/fr0eeOCBIi9uSNJ9990n6a8r0/L7WtxOdv9+/fVXq3/TXbhwIceVNPYYj0qMI2/GOPLewJxStxAREaG4uDgtXrxYP//8s/r166devXpZ/hD/+OOPuu+++7RixQrVrVtXvr6+evbZZ7lSqpTKrrzf+A1IRkaGPvroI3ulJEnat2+fRo0apYoVKyo8PNyy3dHRMce3Nf/85z9zfFtRrlw5Scox8Mtvf81ms65fv261rVmzZnJwcFB6erok6fHHH5ejo6MmTZqUIyfDMHThwgWrfPJ7qX5RKVeuXI7+F8bFixc1cOBAZWZm3vIW3z///NPqKS/SXwOL8uXLW16zosxLkk6fPm31SGez2azPPvtMLVu2tFxyXa9ePUl/3fqcLS0tTZ9++mmO4+U3tzZt2sjLy0vz5s2z6ttPP/2kX375RcHBwYXtEgAgD4xZ7s4xi6Ojo/r27atvv/1WBw4cyNF+7ty5HNs+++wzq7mHvvnmG505c8bqi/n85lmtWjW1bNlSn376qVX8unXrdOjQIavYJ598UpmZmZoyZUqO41y/fv2Oxy9eXl7q0qWLPv74Y505cyZHe26vxe089NBDcnJy0ty5c622z549O0dsXu+14sQ4MifGkfeGUn2l1K0kJiZq4cKFSkxMVPXq1SX9dXno6tWrtXDhQv3jH//Qb7/9phMnTmjp0qX67LPPlJmZqdGjR+uJJ57Qhg0b7NwD2FqHDh1UsWJFhYWF6aWXXpLJZNLnn39eLJcE5+U///mPrl69qszMTF24cEFbt27VDz/8IE9PTy1btszqfu5HHnlEn3/+uTw9PeXv76+4uDitX7/e8sjhbC1btpSjo6Pee+89paamysXFRd26dct3fzds2KCIiAj169dPDRs21PXr1/X5559bBlbSXx9Sb7/9tsaNG6fjx48rJCRE5cuX17Fjx7Rs2TI999xzevnllyVJAQEBWrJkiSIjI9W2bVu5u7urT58+xfq6BgQEaO7cuXr77bdVv359eXl55fg25mb//e9/9cUXX8gwDJnNZu3bt09Lly7V5cuXNX36dPXq1euW+z700EN68skn5e/vLycnJy1btkzJyckaMGDAHeWVl4YNG2rYsGHatWuXvL29tWDBAiUnJ2vhwoWWmJ49e6p27doaNmyYXnnlFTk6OmrBggWqWrWqEhMTC/WalSlTRu+9956GDBmizp07a+DAgZZH+fr6+mr06NGF6g8AIG+MWe7eMcu7776rjRs3ql27dho+fLj8/f118eJF7dmzR+vXr8/x5XilSpX04IMPasiQIUpOTtbMmTNVv359DR8+3BJTkDyjoqIUHBysBx98UEOHDtXFixf1z3/+U02aNLHMdSRJnTt31t///ndFRUUpISFBPXv2VJkyZXT06FEtXbpUs2bN0hNPPHHb/k6fPl1ly5a12ubg4KDXX39dc+bM0YMPPqhmzZpp+PDhuu+++5ScnKy4uDj9/vvv2rdvX0FeWnl7e2vkyJGaNm2a/va3v6lXr17at2+ffvrpJ1WpUsXq6p283mt5zaFVUIwjGUeWKsX+fL+7hCRj2bJllvXsx0iWK1fOanFycjKefPJJwzAMY/jw4YYk48iRI5b94uPjDUnG4cOHbd0FFIO8Hq/cpEmTXOO3bt1qtG/f3nBzczOqV69ujB071lizZk2OR8bm9Xjl3B7dKsmYMGHCLfPMfjxq9lKmTBmjatWqRqdOnYx33nnHOHv2bI59Ll26ZAwZMsSoUqWK4e7ubgQFBRmHDx826tSpY/XIW8MwjH/961/GfffdZzg6Olr1JT/9/e2334yhQ4ca9erVM1xdXY1KlSoZXbt2NdavX58jp2+//dZ48MEHLb9vjRs3NsLDw61+xy5fvmw89dRTRoUKFXI8fjg3derUMYKDg2/5ut346OfcHkublJRkBAcHG+XLlzck5fqI2hvd+LNwcHAwKlSoYLRq1coYOXKkcfDgwRzxNz/K9/z580Z4eLjRuHFjo1y5coanp6fRrl074+uvv7baL6+8svuwa9euHOfK61G+wcHBxpo1a4zmzZsbLi4uRuPGjXN9JHZ8fLzRrl07w9nZ2ahdu7Yxffr0Ar1mNz/KN9uSJUuMVq1aGS4uLkalSpWMQYMGGb///rtVTFhYmFGuXLkcOeX1iGEAKE0Ys/zlXh2z3Cg5OdkIDw83atWqZZQpU8bw8fExHnroIWP+/Pk5XuevvvrKGDdunOHl5WW4ubkZwcHBxokTJ6yOl1eeN49Pbuy7n5+f4eLiYvj7+xvfffddjvdJtvnz5xsBAQGGm5ubUb58eaNZs2bG2LFjjdOnT9+yj9mf7bktjo6Olrj//e9/RmhoqOHj42OUKVPGqFGjhvHII48Y33zzjSUmr3FRbmOS69evG2+99Zbh4+NjuLm5Gd26dTN++eUXo3Llysbzzz9vtX9e77W8fo6dO3e+7RjSMBhH3io3xpH3JpNh2PArkRLMZDJZPX1vyZIlGjRokA4ePJhjQjR3d3f5+PhowoQJ+sc//qFr165Z2q5cuaKyZctq7dq16tGjhy27AAAAAACKjY1V165dtXTp0nxdkYS8paSkqGLFinr77bd5wjpQDLh9Lw+tWrVSZmamzp49q44dO+Ya88ADD+j69ev63//+Z7lHNntCNyZVAwAAAIC7x5UrV3JMzD5z5kxJsjzsCkDRKtVFqcuXL+vXX3+1rB87dkwJCQmqVKmSGjZsqEGDBik0NFTTpk1Tq1atdO7cOcXExKh58+YKDg5W9+7d1bp1aw0dOlQzZ85UVlaWwsPD1aNHj9s+jQMAAAAAUHIsWbJE0dHRevjhh+Xu7q4tW7boq6++Us+ePfXAAw/YOz3gnlSqi1K7d+9W165dLeuRkZGSpLCwMEVHR2vhwoV6++23NWbMGJ06dUpVqlRR+/bt9cgjj0j6a5K9H3/8USNGjFCnTp1Urlw59e7d2+pxtQAAAACAkq958+ZycnLS1KlTZTabLZOfv/322/ZODbhnMacUAAAAAAAAbM7B3gkAAAAAAACg9KEoBQAAAAAAAJsrdXNKZWVl6fTp0ypfvrxMJpO90wEAAMXMMAz98ccfql69uhwc+D7OHhh/AQBQuuR3/FXqilKnT59WrVq17J0GAACwsZMnT6pmzZr2TqNUYvwFAEDpdLvxV6krSpUvX17SXy+Mh4eHnbMBAADFzWw2q1atWpYxAGyP8RcAAKVLfsdfpa4olX3JuIeHB4MiAABKEW4bsx/GXwAAlE63G38xsQIAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAAAAAABszq5Fqblz56p58+by8PCQh4eHAgMD9dNPP+UZHx0dLZPJZLW4urraMGMAAAAAAAAUBSd7nrxmzZp699131aBBAxmGoU8//VSPPvqo9u7dqyZNmuS6j4eHh44cOWJZN5lMtkoXAAAAAAAARcSuRak+ffpYrb/zzjuaO3eutm/fnmdRymQyycfHxxbpAQAAAAAAoJiUmDmlMjMztXjxYqWlpSkwMDDPuMuXL6tOnTqqVauWHn30UR08eNCGWQIAAAAAAKAo2PVKKUnav3+/AgMDdfXqVbm7u2vZsmXy9/fPNbZRo0ZasGCBmjdvrtTUVH3wwQfq0KGDDh48qJo1a+a6T3p6utLT0y3rZrO5WPoBAAAAAACA/LP7lVKNGjVSQkKCduzYoRdeeEFhYWE6dOhQrrGBgYEKDQ1Vy5Yt1blzZ3333XeqWrWqPv744zyPHxUVJU9PT8tSq1at4uoKAAAAAAAA8slkGIZh7yRu1L17d9WrV++WhaYb9evXT05OTvrqq69ybc/tSqlatWopNTVVHh4eRZLz3cb3tZX2TgF2dPzdYHunAAA2ZTab5enpWao/++2tKH4G9h6/8PkJAED+5fez3+5XSt0sKyvLqoh0K5mZmdq/f7+qVauWZ4yLi4s8PDysFgAAAAAAANiXXeeUGjdunHr37q3atWvrjz/+0KJFixQbG6s1a9ZIkkJDQ1WjRg1FRUVJkiZPnqz27durfv36SklJ0fvvv68TJ07o2WeftWc3AAAAAAAAUEB2LUqdPXtWoaGhOnPmjDw9PdW8eXOtWbNGPXr0kCQlJibKweH/Lua6dOmShg8frqSkJFWsWFEBAQHatm1bnhOjAwAAAAAAoGSya1Hqk08+uWV7bGys1fqMGTM0Y8aMYswIAAAAAAAAtlDi5pQCAAAAAADAvY+iFAAAAAAAAGyOohQAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAIAdnDp1Sk8//bQqV64sNzc3NWvWTLt377a0G4ah8ePHq1q1anJzc1P37t119OhRq2NcvHhRgwYNkoeHhypUqKBhw4bp8uXLVjE///yzOnbsKFdXV9WqVUtTp07NkcvSpUvVuHFjubq6qlmzZlq1apVVe35yAQAAKCiKUgAAADZ26dIlPfDAAypTpox++uknHTp0SNOmTVPFihUtMVOnTtWHH36oefPmaceOHSpXrpyCgoJ09epVS8ygQYN08OBBrVu3TitWrNDmzZv13HPPWdrNZrN69uypOnXqKD4+Xu+//74mTpyo+fPnW2K2bdumgQMHatiwYdq7d69CQkIUEhKiAwcOFCgXAACAgjIZhmHYOwlbMpvN8vT0VGpqqjw8POydjl34vrbS3inAjo6/G2zvFADApkriZ/9rr72mrVu36j//+U+u7YZhqHr16hozZoxefvllSVJqaqq8vb0VHR2tAQMG6JdffpG/v7927dqlNm3aSJJWr16thx9+WL///ruqV6+uuXPn6o033lBSUpKcnZ0t516+fLkOHz4sSerfv7/S0tK0YsUKy/nbt2+vli1bat68efnK5XaK4mdg7/ELn58AAORffj/7uVIKAADAxn744Qe1adNG/fr1k5eXl1q1aqV//etflvZjx44pKSlJ3bt3t2zz9PRUu3btFBcXJ0mKi4tThQoVLAUpSerevbscHBy0Y8cOS0ynTp0sBSlJCgoK0pEjR3Tp0iVLzI3nyY7JPk9+cgEAACgMilIAAAA29ttvv2nu3Llq0KCB1qxZoxdeeEEvvfSSPv30U0lSUlKSJMnb29tqP29vb0tbUlKSvLy8rNqdnJxUqVIlq5jcjnHjOfKKubH9drncLD09XWaz2WoBAAC4mZO9EwAAAChtsrKy1KZNG/3jH/+QJLVq1UoHDhzQvHnzFBYWZufs7lxUVJQmTZpk7zQAAEAJx5VSAAAANlatWjX5+/tbbfPz81NiYqIkycfHR5KUnJxsFZOcnGxp8/Hx0dmzZ63ar1+/rosXL1rF5HaMG8+RV8yN7bfL5Wbjxo1TamqqZTl58mSucQAAoHSjKAUAAGBjDzzwgI4cOWK17b///a/q1KkjSapbt658fHwUExNjaTebzdqxY4cCAwMlSYGBgUpJSVF8fLwlZsOGDcrKylK7du0sMZs3b9a1a9csMevWrVOjRo0sT/oLDAy0Ok92TPZ58pPLzVxcXOTh4WG1AAAA3IyiFAAAgI2NHj1a27dv1z/+8Q/9+uuvWrRokebPn6/w8HBJkslk0qhRo/T222/rhx9+0P79+xUaGqrq1asrJCRE0l9XVvXq1UvDhw/Xzp07tXXrVkVERGjAgAGqXr26JOmpp56Ss7Ozhg0bpoMHD2rJkiWaNWuWIiMjLbmMHDlSq1ev1rRp03T48GFNnDhRu3fvVkRERL5zAQAAKAzmlAIAALCxtm3batmyZRo3bpwmT56sunXraubMmRo0aJAlZuzYsUpLS9Nzzz2nlJQUPfjgg1q9erVcXV0tMV9++aUiIiL00EMPycHBQX379tWHH35oaff09NTatWsVHh6ugIAAValSRePHj9dzzz1nienQoYMWLVqkN998U6+//roaNGig5cuXq2nTpgXKBQAAoKBMhmEY9k7Clsxmszw9PZWamlpqLyX3fW2lvVOAHR1/N9jeKQCATfHZb39F8TOw9/iFz08AAPIvv5/93L4HAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJtzsncCAAAAAG7N97WVdj3/8XeD7Xp+AMC9iSulAAAAAAAAYHMUpQAAAAAAAGBzFKUAAAAAAABgcxSlAAAAAAAAYHMUpQAAAAAAAGBzdi1KzZ07V82bN5eHh4c8PDwUGBion3766Zb7LF26VI0bN5arq6uaNWumVatW2ShbAAAAAAAAFBUne568Zs2aevfdd9WgQQMZhqFPP/1Ujz76qPbu3asmTZrkiN+2bZsGDhyoqKgoPfLII1q0aJFCQkK0Z88eNW3a1A49AAAAAFDcfF9badfzH3832K7nB4B7lV2vlOrTp48efvhhNWjQQA0bNtQ777wjd3d3bd++Pdf4WbNmqVevXnrllVfk5+enKVOmqHXr1po9e7aNMwcAAAAAAMCdKDFzSmVmZmrx4sVKS0tTYGBgrjFxcXHq3r271bagoCDFxcXZIkUAAAAAAAAUEbvevidJ+/fvV2BgoK5evSp3d3ctW7ZM/v7+ucYmJSXJ29vbapu3t7eSkpLyPH56errS09Mt62azuWgSBwAAAAAAQKHZ/UqpRo0aKSEhQTt27NALL7ygsLAwHTp0qMiOHxUVJU9PT8tSq1atIjs2AAAAAAAACsfuRSlnZ2fVr19fAQEBioqKUosWLTRr1qxcY318fJScnGy1LTk5WT4+Pnkef9y4cUpNTbUsJ0+eLNL8AQAAAAAAUHB2v33vZllZWVa3290oMDBQMTExGjVqlGXbunXr8pyDSpJcXFzk4uJS1GkCAAAAgE3w9EEA9yq7FqXGjRun3r17q3bt2vrjjz+0aNEixcbGas2aNZKk0NBQ1ahRQ1FRUZKkkSNHqnPnzpo2bZqCg4O1ePFi7d69W/Pnz7dnNwAAAAAAAFBAdi1KnT17VqGhoTpz5ow8PT3VvHlzrVmzRj169JAkJSYmysHh/+4w7NChgxYtWqQ333xTr7/+uho0aKDly5eradOm9uoCAAAAAAAACsGuRalPPvnklu2xsbE5tvXr10/9+vUrpowAAAAAAABgC3af6BwAAAAAAAClD0UpAAAAAAAA2BxFKQAAAAAAANgcRSkAAAAAAADYHEUpAAAAAAAA2BxFKQAAAAAAANgcRSkAAAAAAADYHEUpAAAAAAAA2BxFKQAAAAAAANgcRSkAAAAAAADYHEUpAAAAAAAA2BxFKQAAAAAAANgcRSkAAAAAAADYHEUpAAAAG5s4caJMJpPV0rhxY0v71atXFR4ersqVK8vd3V19+/ZVcnKy1TESExMVHByssmXLysvLS6+88oquX79uFRMbG6vWrVvLxcVF9evXV3R0dI5c5syZI19fX7m6uqpdu3bauXOnVXt+cgEAACgMilIAAAB20KRJE505c8aybNmyxdI2evRo/fjjj1q6dKk2bdqk06dP6/HHH7e0Z2ZmKjg4WBkZGdq2bZs+/fRTRUdHa/z48ZaYY8eOKTg4WF27dlVCQoJGjRqlZ599VmvWrLHELFmyRJGRkZowYYL27NmjFi1aKCgoSGfPns13LgAAAIVFUQoAAMAOnJyc5OPjY1mqVKkiSUpNTdUnn3yi6dOnq1u3bgoICNDChQu1bds2bd++XZK0du1aHTp0SF988YVatmyp3r17a8qUKZozZ44yMjIkSfPmzVPdunU1bdo0+fn5KSIiQk888YRmzJhhyWH69OkaPny4hgwZIn9/f82bN09ly5bVggUL8p0LAABAYVGUAgAAsIOjR4+qevXquu+++zRo0CAlJiZKkuLj43Xt2jV1797dEtu4cWPVrl1bcXFxkqS4uDg1a9ZM3t7elpigoCCZzWYdPHjQEnPjMbJjso+RkZGh+Ph4qxgHBwd1797dEpOfXHKTnp4us9lstQAAANyMohQAAICNtWvXTtHR0Vq9erXmzp2rY8eOqWPHjvrjjz+UlJQkZ2dnVahQwWofb29vJSUlSZKSkpKsClLZ7dltt4oxm826cuWKzp8/r8zMzFxjbjzG7XLJTVRUlDw9PS1LrVq18vfCAACAUsXJ3gkAAACUNr1797b8f/PmzdWuXTvVqVNHX3/9tdzc3OyYWdEYN26cIiMjLetms5nCFAAAyIErpQAAAOysQoUKatiwoX799Vf5+PgoIyNDKSkpVjHJycny8fGRJPn4+OR4Al72+u1iPDw85ObmpipVqsjR0THXmBuPcbtccuPi4iIPDw+rBQAA4GYUpQAAAOzs8uXL+t///qdq1aopICBAZcqUUUxMjKX9yJEjSkxMVGBgoCQpMDBQ+/fvt3pK3rp16+Th4SF/f39LzI3HyI7JPoazs7MCAgKsYrKyshQTE2OJyU8uAAAAhcXtewAAADb28ssvq0+fPqpTp45Onz6tCRMmyNHRUQMHDpSnp6eGDRumyMhIVapUSR4eHhoxYoQCAwPVvn17SVLPnj3l7++vZ555RlOnTlVSUpLefPNNhYeHy8XFRZL0/PPPa/bs2Ro7dqyGDh2qDRs26Ouvv9bKlSsteURGRiosLExt2rTR/fffr5kzZyotLU1DhgyRpHzlAgAAUFgUpQAAAGzs999/18CBA3XhwgVVrVpVDz74oLZv366qVatKkmbMmCEHBwf17dtX6enpCgoK0kcffWTZ39HRUStWrNALL7ygwMBAlStXTmFhYZo8ebIlpm7dulq5cqVGjx6tWbNmqWbNmvr3v/+toKAgS0z//v117tw5jR8/XklJSWrZsqVWr15tNfn57XIBAAAoLIpSAAAANrZ48eJbtru6umrOnDmaM2dOnjF16tTRqlWrbnmcLl26aO/evbeMiYiIUERExB3lAgAAUBjMKQUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJtjTikAAAAAQInl+9rK2wcVk+PvBtvt3EBpwJVSAAAAAAAAsDmKUgAAAAAAALA5ilIAAAAAAACwOYpSAAAAAAAAsDmKUgAAAAAAALA5ilIAAAAAAACwOSd7JwAAAAAAAHLyfW2lXc9//N1gu54f9z6ulAIAAAAAAIDN2bUoFRUVpbZt26p8+fLy8vJSSEiIjhw5cst9oqOjZTKZrBZXV1cbZQwAAAAAAICiYNei1KZNmxQeHq7t27dr3bp1unbtmnr27Km0tLRb7ufh4aEzZ85YlhMnTtgoYwAAAAAAABQFu84ptXr1aqv16OhoeXl5KT4+Xp06dcpzP5PJJB8fn+JODwAAAAAAAMWkRM0plZqaKkmqVKnSLeMuX76sOnXqqFatWnr00Ud18OBBW6QHAAAAAACAIlJiilJZWVkaNWqUHnjgATVt2jTPuEaNGmnBggX6/vvv9cUXXygrK0sdOnTQ77//nmt8enq6zGaz1QIAAAAAAAD7suvtezcKDw/XgQMHtGXLllvGBQYGKjAw0LLeoUMH+fn56eOPP9aUKVNyxEdFRWnSpElFni8AAAAAAAAKr0RcKRUREaEVK1Zo48aNqlmzZoH2LVOmjFq1aqVff/011/Zx48YpNTXVspw8ebIoUgYAAAAAAMAdsOuVUoZhaMSIEVq2bJliY2NVt27dAh8jMzNT+/fv18MPP5xru4uLi1xcXO40VQAAAAAAABQhuxalwsPDtWjRIn3//fcqX768kpKSJEmenp5yc3OTJIWGhqpGjRqKioqSJE2ePFnt27dX/fr1lZKSovfff18nTpzQs88+a7d+AAAAAAAAoGDsWpSaO3euJKlLly5W2xcuXKjBgwdLkhITE+Xg8H93GV66dEnDhw9XUlKSKlasqICAAG3btk3+/v62ShsAAAAAAAB3yO63791ObGys1fqMGTM0Y8aMYsoIAAAAAAAAtlAiJjoHAAAAAABA6UJRCgAAAAAAADZHUQoAAAAAAAA2R1EKAAAAAAAANkdRCgAAAAAAADZHUQoAAAAAAAA2R1EKAAAAAAAANkdRCgAAAAAAADZHUQoAAAAAAAA2R1EKAAAAAAAANudk7wQAAAAAAABu5vvaSrue//i7wXY9f2nAlVIAAAAAAACwOYpSAAAAAAAAsDmKUgAAAAAAALA5ilIAAAAAAACwOYpSAAAAAAAAsDmKUgAAAAAAALA5ilIAAAAAAACwOYpSAAAAdvbuu+/KZDJp1KhRlm1Xr15VeHi4KleuLHd3d/Xt21fJyclW+yUmJio4OFhly5aVl5eXXnnlFV2/ft0qJjY2Vq1bt5aLi4vq16+v6OjoHOefM2eOfH195erqqnbt2mnnzp1W7fnJBQAAoKAoSgEAANjRrl279PHHH6t58+ZW20ePHq0ff/xRS5cu1aZNm3T69Gk9/vjjlvbMzEwFBwcrIyND27Zt06effqro6GiNHz/eEnPs2DEFBwera9euSkhI0KhRo/Tss89qzZo1lpglS5YoMjJSEyZM0J49e9SiRQsFBQXp7Nmz+c4FAACgMChKAQAA2Mnly5c1aNAg/etf/1LFihUt21NTU/XJJ59o+vTp6tatmwICArRw4UJt27ZN27dvlyStXbtWhw4d0hdffKGWLVuqd+/emjJliubMmaOMjAxJ0rx581S3bl1NmzZNfn5+ioiI0BNPPKEZM2ZYzjV9+nQNHz5cQ4YMkb+/v+bNm6eyZctqwYIF+c4FAACgMChKAQAA2El4eLiCg4PVvXt3q+3x8fG6du2a1fbGjRurdu3aiouLkyTFxcWpWbNm8vb2tsQEBQXJbDbr4MGDlpibjx0UFGQ5RkZGhuLj461iHBwc1L17d0tMfnK5WXp6usxms9UCAABwMyd7JwAAAFAaLV68WHv27NGuXbtytCUlJcnZ2VkVKlSw2u7t7a2kpCRLzI0Fqez27LZbxZjNZl25ckWXLl1SZmZmrjGHDx/Ody43i4qK0qRJk27RewAAAK6UAgAAsLmTJ09q5MiR+vLLL+Xq6mrvdIrcuHHjlJqaallOnjxp75QAAEAJRFEKAADAxuLj43X27Fm1bt1aTk5OcnJy0qZNm/Thhx/KyclJ3t7eysjIUEpKitV+ycnJ8vHxkST5+PjkeAJe9vrtYjw8POTm5qYqVarI0dEx15gbj3G7XG7m4uIiDw8PqwUAAOBmFKUAAABs7KGHHtL+/fuVkJBgWdq0aaNBgwZZ/r9MmTKKiYmx7HPkyBElJiYqMDBQkhQYGKj9+/dbPSVv3bp18vDwkL+/vyXmxmNkx2Qfw9nZWQEBAVYxWVlZiomJscQEBATcNhcAAIDCYE4pAAAAGytfvryaNm1qta1cuXKqXLmyZfuwYcMUGRmpSpUqycPDQyNGjFBgYKDat28vSerZs6f8/f31zDPPaOrUqUpKStKbb76p8PBwubi4SJKef/55zZ49W2PHjtXQoUO1YcMGff3111q5cqXlvJGRkQoLC1ObNm10//33a+bMmUpLS9OQIUMkSZ6enrfNBQAAoDAoSgEAAJRAM2bMkIODg/r27av09HQFBQXpo48+srQ7OjpqxYoVeuGFFxQYGKhy5copLCxMkydPtsTUrVtXK1eu1OjRozVr1izVrFlT//73vxUUFGSJ6d+/v86dO6fx48crKSlJLVu21OrVq60mP79dLgAAAIVBUQoAAKAEiI2NtVp3dXXVnDlzNGfOnDz3qVOnjlatWnXL43bp0kV79+69ZUxERIQiIiLybM9PLgAAAAXFnFIAAAAAAACwOYpSAAAAAAAAsDmKUgAAAAAAALA5ilIAAAAAAACwOYpSAAAAAAAAsDmKUgAAAAAAALA5ilIAAAAAAACwObsWpaKiotS2bVuVL19eXl5eCgkJ0ZEjR26739KlS9W4cWO5urqqWbNmWrVqlQ2yBQAAAAAAQFGxa1Fq06ZNCg8P1/bt27Vu3Tpdu3ZNPXv2VFpaWp77bNu2TQMHDtSwYcO0d+9ehYSEKCQkRAcOHLBh5gAAAAAAALgThSpK3Xfffbpw4UKO7SkpKbrvvvvyfZzVq1dr8ODBatKkiVq0aKHo6GglJiYqPj4+z31mzZqlXr166ZVXXpGfn5+mTJmi1q1ba/bs2YXpCgAAQL4V1RgIAAAAhSxKHT9+XJmZmTm2p6en69SpU4VOJjU1VZJUqVKlPGPi4uLUvXt3q21BQUGKi4vLNT49PV1ms9lqAQAAKIziGgMBAACURk4FCf7hhx8s/79mzRp5enpa1jMzMxUTEyNfX99CJZKVlaVRo0bpgQceUNOmTfOMS0pKkre3t9U2b29vJSUl5RofFRWlSZMmFSonAAAAqXjHQAAAAKVVgYpSISEhkiSTyaSwsDCrtjJlysjX11fTpk0rVCLh4eE6cOCAtmzZUqj98zJu3DhFRkZa1s1ms2rVqlWk5wAAAPe24hwDAQAAlFYFKkplZWVJkurWratdu3apSpUqRZJERESEVqxYoc2bN6tmzZq3jPXx8VFycrLVtuTkZPn4+OQa7+LiIhcXlyLJEwAAlE7FNQYCAAAozQo1p9SxY8eKZDBmGIYiIiK0bNkybdiwQXXr1r3tPoGBgYqJibHatm7dOgUGBt5xPgAAALdSVGMgAAAAFPBKqRvFxMQoJiZGZ8+etXx7mG3BggX5OkZ4eLgWLVqk77//XuXLl7fMC+Xp6Sk3NzdJUmhoqGrUqKGoqChJ0siRI9W5c2dNmzZNwcHBWrx4sXbv3q358+cXtisAAAD5VhRjIAAAABTySqlJkyapZ8+eiomJ0fnz53Xp0iWrJb/mzp2r1NRUdenSRdWqVbMsS5YsscQkJibqzJkzlvUOHTpo0aJFmj9/vlq0aKFvvvlGy5cvv+Xk6AAAAEWhqMZAAAAAKOSVUvPmzVN0dLSeeeaZOzq5YRi3jYmNjc2xrV+/furXr98dnRsAAKCgimoMBAAAgEJeKZWRkaEOHToUdS4AAAAlGmMgAACAolOootSzzz6rRYsWFXUuAAAAJRpjIAAAgKJTqNv3rl69qvnz52v9+vVq3ry5ypQpY9U+ffr0IkkOAACgJGEMBAAAUHQKVZT6+eef1bJlS0nSgQMHrNpMJtMdJwUAAFASMQYCAAAoOoUqSm3cuLGo8wAAACjxGAMBAAAUnULNKQUAAAAAAADciUJdKdW1a9dbXqK+YcOGQicEAABQUjEGAgAAKDqFKkplz6WQ7dq1a0pISNCBAwcUFhZWFHkBAACUOIyBAAAAik6hilIzZszIdfvEiRN1+fLlO0oIAACgpGIMBAAAUHSKdE6pp59+WgsWLCjKQwIAAJR4jIEAAAAKrkiLUnFxcXJ1dS3KQwIAAJR4jIEAAAAKrlC37z3++ONW64Zh6MyZM9q9e7feeuutIkkMAACgpGEMBAAAUHQKVZTy9PS0WndwcFCjRo00efJk9ezZs0gSAwAAKGkYAwEAABSdQhWlFi5cWNR5AAAAlHiMgQAAAIpOoYpS2eLj4/XLL79Ikpo0aaJWrVoVSVIAAAAlGWMgAACAO1eootTZs2c1YMAAxcbGqkKFCpKklJQUde3aVYsXL1bVqlWLMkcAAIASgTEQAABA0SnU0/dGjBihP/74QwcPHtTFixd18eJFHThwQGazWS+99FJR5wgAAFAiMAYCAAAoOoW6Umr16tVav369/Pz8LNv8/f01Z84cJvkEAAD3LMZAAAAARadQV0plZWWpTJkyObaXKVNGWVlZd5wUAABAScQYCAAAoOgUqijVrVs3jRw5UqdPn7ZsO3XqlEaPHq2HHnqoyJIDAAAoSRgDAQAAFJ1CFaVmz54ts9ksX19f1atXT/Xq1VPdunVlNpv1z3/+s6hzBAAAKBEYAwEAABSdQs0pVatWLe3Zs0fr16/X4cOHJUl+fn7q3r17kSYHAABQkjAGAgAAKDoFulJqw4YN8vf3l9lslslkUo8ePTRixAiNGDFCbdu2VZMmTfSf//ynuHIFAACwi6IeA82dO1fNmzeXh4eHPDw8FBgYqJ9++snSfvXqVYWHh6ty5cpyd3dX3759lZycbHWMxMREBQcHq2zZsvLy8tIrr7yi69evW8XExsaqdevWcnFxUf369RUdHZ0jlzlz5sjX11eurq5q166ddu7cadWen1wAAAAKo0BFqZkzZ2r48OHy8PDI0ebp6am///3vmj59epElBwAAUBIU9RioZs2aevfddxUfH6/du3erW7duevTRR3Xw4EFJ0ujRo/Xjjz9q6dKl2rRpk06fPq3HH3/csn9mZqaCg4OVkZGhbdu26dNPP1V0dLTGjx9viTl27JiCg4PVtWtXJSQkaNSoUXr22We1Zs0aS8ySJUsUGRmpCRMmaM+ePWrRooWCgoJ09uxZS8ztcgEAACisAhWl9u3bp169euXZ3rNnT8XHx99xUgAAACVJUY+B+vTpo4cfflgNGjRQw4YN9c4778jd3V3bt29XamqqPvnkE02fPl3dunVTQECAFi5cqG3btmn79u2SpLVr1+rQoUP64osv1LJlS/Xu3VtTpkzRnDlzlJGRIUmaN2+e6tatq2nTpsnPz08RERF64oknNGPGDEse06dP1/DhwzVkyBD5+/tr3rx5Klu2rBYsWCBJ+coFAACgsApUlEpOTs71McjZnJycdO7cuTtOCgAAoCQpzjFQZmamFi9erLS0NAUGBio+Pl7Xrl2zmqeqcePGql27tuLi4iRJcXFxatasmby9vS0xQUFBMpvNlqut4uLicsx1FRQUZDlGRkaG4uPjrWIcHBzUvXt3S0x+cslNenq6zGaz1QIAAHCzAhWlatSooQMHDuTZ/vPPP6tatWp3nBQAAEBJUhxjoP3798vd3V0uLi56/vnntWzZMvn7+yspKUnOzs6qUKGCVby3t7eSkpIkSUlJSVYFqez27LZbxZjNZl25ckXnz59XZmZmrjE3HuN2ueQmKipKnp6elqVWrVr5e1EAAECpUqCi1MMPP6y33npLV69ezdF25coVTZgwQY888kiRJQcAAFASFMcYqFGjRkpISNCOHTv0wgsvKCwsTIcOHSqqlO1q3LhxSk1NtSwnT560d0oAAKAEcipI8JtvvqnvvvtODRs2VEREhBo1aiRJOnz4sObMmaPMzEy98cYbxZIoAACAvRTHGMjZ2Vn169eXJAUEBGjXrl2aNWuW+vfvr4yMDKWkpFhdoZScnCwfHx9Jko+PT46n5GU/Ee/GmJufkpecnCwPDw+5ubnJ0dFRjo6OucbceIzb5ZIbFxcXubi4FODVAAAApVGBrpTy9vbWtm3b1LRpU40bN06PPfaYHnvsMb3++utq2rSptmzZkuMScAAAgLudLcZAWVlZSk9PV0BAgMqUKaOYmBhL25EjR5SYmKjAwEBJUmBgoPbv32/1lLx169bJw8ND/v7+lpgbj5Edk30MZ2dnBQQEWMVkZWUpJibGEpOfXAAAAAqrQFdKSVKdOnW0atUqXbp0Sb/++qsMw1CDBg1UsWLF4sgPAACgRCjKMdC4cePUu3dv1a5dW3/88YcWLVqk2NhYrVmzRp6enho2bJgiIyNVqVIleXh4aMSIEQoMDFT79u0l/fW0P39/fz3zzDOaOnWqkpKS9Oabbyo8PNxyhdLzzz+v2bNna+zYsRo6dKg2bNigr7/+WitXrrTkERkZqbCwMLVp00b333+/Zs6cqbS0NA0ZMkSS8pULAABAYRW4KJWtYsWKatu2bVHmAgAAUOIVxRjo7NmzCg0N1ZkzZ+Tp6anmzZtrzZo16tGjhyRpxowZcnBwUN++fZWenq6goCB99NFHlv0dHR21YsUKvfDCCwoMDFS5cuUUFhamyZMnW2Lq1q2rlStXavTo0Zo1a5Zq1qypf//73woKCrLE9O/fX+fOndP48eOVlJSkli1bavXq1VZXfd0uFwAAgMIqdFEKAAAAhfPJJ5/cst3V1VVz5szRnDlz8ozJvnLrVrp06aK9e/feMiYiIkIRERF3lAsAAEBhFGhOKQAAAAAAAKAoUJQCAAAAAACAzVGUAgAAAAAAgM3ZtSi1efNm9enTR9WrV5fJZNLy5ctvGR8bGyuTyZRjSUpKsk3CAAAAAAAAKBJ2LUqlpaWpRYsWBZ4488iRIzpz5oxl8fLyKqYMAQAAAAAAUBzs+vS93r17q3fv3gXez8vLSxUqVCj6hAAAAAAAAGATd+WcUi1btlS1atXUo0cPbd269Zax6enpMpvNVgsAAAAAAADs664qSlWrVk3z5s3Tt99+q2+//Va1atVSly5dtGfPnjz3iYqKkqenp2WpVauWDTMGAAAAAABAbux6+15BNWrUSI0aNbKsd+jQQf/73/80Y8YMff7557nuM27cOEVGRlrWzWYzhSkAAAAAAAA7u6uKUrm5//77tWXLljzbXVxc5OLiYsOMAAAAAAAAcDt31e17uUlISFC1atXsnQYAAAAAAAAKwK5XSl2+fFm//vqrZf3YsWNKSEhQpUqVVLt2bY0bN06nTp3SZ599JkmaOXOm6tatqyZNmujq1av697//rQ0bNmjt2rX26gIAAAAAAAAKwa5Fqd27d6tr166W9ey5n8LCwhQdHa0zZ84oMTHR0p6RkaExY8bo1KlTKlu2rJo3b67169dbHQMAAAAAAAAln12LUl26dJFhGHm2R0dHW62PHTtWY8eOLeasAAAAAAAAUNzu+jmlAAAAAAAAcPehKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAYGNRUVFq27atypcvLy8vL4WEhOjIkSNWMVevXlV4eLgqV64sd3d39e3bV8nJyVYxiYmJCg4OVtmyZeXl5aVXXnlF169ft4qJjY1V69at5eLiovr16ys6OjpHPnPmzJGvr69cXV3Vrl077dy5s8C5AAAAFBRFKQAAABvbtGmTwsPDtX37dq1bt07Xrl1Tz549lZaWZokZPXq0fvzxRy1dulSbNm3S6dOn9fjjj1vaMzMzFRwcrIyMDG3btk2ffvqpoqOjNX78eEvMsWPHFBwcrK5duyohIUGjRo3Ss88+qzVr1lhilixZosjISE2YMEF79uxRixYtFBQUpLNnz+Y7FwAAgMJwsncCAAAApc3q1aut1qOjo+Xl5aX4+Hh16tRJqamp+uSTT7Ro0SJ169ZNkrRw4UL5+flp+/btat++vdauXatDhw5p/fr18vb2VsuWLTVlyhS9+uqrmjhxopydnTVv3jzVrVtX06ZNkyT5+flpy5YtmjFjhoKCgiRJ06dP1/DhwzVkyBBJ0rx587Ry5UotWLBAr732Wr5yAQAAKAyulAIAALCz1NRUSVKlSpUkSfHx8bp27Zq6d+9uiWncuLFq166tuLg4SVJcXJyaNWsmb29vS0xQUJDMZrMOHjxoibnxGNkx2cfIyMhQfHy8VYyDg4O6d+9uiclPLjdLT0+X2Wy2WgAAAG5GUQoAAMCOsrKyNGrUKD3wwANq2rSpJCkpKUnOzs6qUKGCVay3t7eSkpIsMTcWpLLbs9tuFWM2m3XlyhWdP39emZmZucbceIzb5XKzqKgoeXp6WpZatWrl89UAAAClCUUpAAAAOwoPD9eBAwe0ePFie6dSZMaNG6fU1FTLcvLkSXunBAAASiDmlAIAALCTiIgIrVixQps3b1bNmjUt2318fJSRkaGUlBSrK5SSk5Pl4+Njibn5KXnZT8S7Mebmp+QlJyfLw8NDbm5ucnR0lKOjY64xNx7jdrnczMXFRS4uLgV4JQAAQGnElVIAAAA2ZhiGIiIitGzZMm3YsEF169a1ag8ICFCZMmUUExNj2XbkyBElJiYqMDBQkhQYGKj9+/dbPSVv3bp18vDwkL+/vyXmxmNkx2Qfw9nZWQEBAVYxWVlZiomJscTkJxcAAIDC4EopAAAAGwsPD9eiRYv0/fffq3z58pa5mTw9PeXm5iZPT08NGzZMkZGRqlSpkjw8PDRixAgFBgZannbXs2dP+fv765lnntHUqVOVlJSkN998U+Hh4ZarlJ5//nnNnj1bY8eO1dChQ7VhwwZ9/fXXWrlypSWXyMhIhYWFqU2bNrr//vs1c+ZMpaWlWZ7Gl59cAAAACsOuV0pt3rxZffr0UfXq1WUymbR8+fLb7hMbG6vWrVvLxcVF9evXV3R0dLHnCQAAUJTmzp2r1NRUdenSRdWqVbMsS5YsscTMmDFDjzzyiPr27atOnTrJx8dH3333naXd0dFRK1askKOjowIDA/X0008rNDRUkydPtsTUrVtXK1eu1Lp169SiRQtNmzZN//73vxUUFGSJ6d+/vz744AONHz9eLVu2VEJCglavXm01+fntcgEAACgMu14plZaWphYtWmjo0KF6/PHHbxt/7NgxBQcH6/nnn9eXX36pmJgYPfvss6pWrZrV4AoAAKAkMwzjtjGurq6aM2eO5syZk2dMnTp1tGrVqlsep0uXLtq7d+8tYyIiIhQREXFHuQAAABSUXYtSvXv3Vu/evfMdP2/ePNWtW1fTpk2TJPn5+WnLli2aMWMGRSkAAAAAAIC7yF010XlcXJy6d+9utS0oKEhxcXF57pOeni6z2Wy1AAAAAAAAwL7uqqJUUlKS1fwGkuTt7S2z2awrV67kuk9UVJQ8PT0tS61atWyRKgAAAAAAAG7hripKFca4ceOUmppqWU6ePGnvlAAAAAAAAEo9u84pVVA+Pj5KTk622pacnCwPDw+5ubnluo+Li4vlscgAAAAAAAAoGe6qK6UCAwMVExNjtW3dunUKDAy0U0YAAAAAAAAoDLsWpS5fvqyEhAQlJCRIko4dO6aEhAQlJiZK+uvWu9DQUEv8888/r99++01jx47V4cOH9dFHH+nrr7/W6NGj7ZE+AAAAAAAACsmuRandu3erVatWatWqlSQpMjJSrVq10vjx4yVJZ86csRSoJKlu3bpauXKl1q1bpxYtWmjatGn697//raCgILvkDwAAAAAAgMKx65xSXbp0kWEYebZHR0fnus/evXuLMSsAAAAAAAAUt7tqTikAAAAAAADcGyhKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAAAAAAwOYoSgEAAAAAAMDmKEoBAAAAAADA5ihKAQAA2NjmzZvVp08fVa9eXSaTScuXL7dqNwxD48ePV7Vq1eTm5qbu3bvr6NGjVjEXL17UoEGD5OHhoQoVKmjYsGG6fPmyVczPP/+sjh07ytXVVbVq1dLUqVNz5LJ06VI1btxYrq6uatasmVatWlXgXAAAAAqDohQAAICNpaWlqUWLFpozZ06u7VOnTtWHH36oefPmaceOHSpXrpyCgoJ09epVS8ygQYN08OBBrVu3TitWrNDmzZv13HPPWdrNZrN69uypOnXqKD4+Xu+//74mTpyo+fPnW2K2bdumgQMHatiwYdq7d69CQkIUEhKiAwcOFCgXAACAwnCydwIAAAClTe/evdW7d+9c2wzD0MyZM/Xmm2/q0UcflSR99tln8vb21vLlyzVgwAD98ssvWr16tXbt2qU2bdpIkv75z3/q4Ycf1gcffKDq1avryy+/VEZGhhYsWCBnZ2c1adJECQkJmj59uqV4NWvWLPXq1UuvvPKKJGnKlClat26dZs+erXnz5uUrFwAAgMLiSikAAIAS5NixY0pKSlL37t0t2zw9PdWuXTvFxcVJkuLi4lShQgVLQUqSunfvLgcHB+3YscMS06lTJzk7O1tigoKCdOTIEV26dMkSc+N5smOyz5OfXHKTnp4us9lstQAAANyMohQAAEAJkpSUJEny9va22u7t7W1pS0pKkpeXl1W7k5OTKlWqZBWT2zFuPEdeMTe23y6X3ERFRcnT09Oy1KpV6za9BgAApVGJKErNmTNHvr6+cnV1Vbt27bRz5848Y6Ojo2UymawWV1dXG2YLAACAWxk3bpxSU1Mty8mTJ+2dEgAAKIHsXpRasmSJIiMjNWHCBO3Zs0ctWrRQUFCQzp49m+c+Hh4eOnPmjGU5ceKEDTMGAAAoPj4+PpKk5ORkq+3JycmWNh8fnxxjpevXr+vixYtWMbkd48Zz5BVzY/vtcsmNi4uLPDw8rBYAAICb2b0oNX36dA0fPlxDhgyRv7+/5s2bp7Jly2rBggV57mMymeTj42NZbr6kHAAA4G5Vt25d+fj4KCYmxrLNbDZrx44dCgwMlCQFBgYqJSVF8fHxlpgNGzYoKytL7dq1s8Rs3rxZ165ds8SsW7dOjRo1UsWKFS0xN54nOyb7PPnJBQAAoLDsWpTKyMhQfHy81eSZDg4O6t69+y0nz7x8+bLq1KmjWrVq6dFHH9XBgwfzjGWiTQAAUNJcvnxZCQkJSkhIkPTXhOIJCQlKTEyUyWTSqFGj9Pbbb+uHH37Q/v37FRoaqurVqyskJESS5Ofnp169emn48OHauXOntm7dqoiICA0YMEDVq1eXJD311FNydnbWsGHDdPDgQS1ZskSzZs1SZGSkJY+RI0dq9erVmjZtmg4fPqyJEydq9+7dioiIkKR85QIAAFBYTvY8+fnz55WZmZnr5JmHDx/OdZ9GjRppwYIFat68uVJTU/XBBx+oQ4cOOnjwoGrWrJkjPioqSpMmTSqW/AEAAApj9+7d6tq1q2U9u1AUFham6OhojR07VmlpaXruueeUkpKiBx98UKtXr7aaR/PLL79URESEHnroITk4OKhv37768MMPLe2enp5au3atwsPDFRAQoCpVqmj8+PF67rnnLDEdOnTQokWL9Oabb+r1119XgwYNtHz5cjVt2tQSk59cAAAACsOuRanCCAwMtLpcvEOHDvLz89PHH3+sKVOm5IgfN26c1TeCZrOZJ8AAAAC76tKliwzDyLPdZDJp8uTJmjx5cp4xlSpV0qJFi255nubNm+s///nPLWP69eunfv363VEuAAAAhWHXolSVKlXk6OhY4Mkzb1SmTBm1atVKv/76a67tLi4ucnFxueNcAQAAAAAAUHTsOqeUs7OzAgICrCbPzMrKUkxMTL4nz8zMzNT+/ftVrVq14koTAAAAAAAARczut+9FRkYqLCxMbdq00f3336+ZM2cqLS1NQ4YMkSSFhoaqRo0aioqKkiRNnjxZ7du3V/369ZWSkqL3339fJ06c0LPPPmvPbgAAAAAAAKAA7F6U6t+/v86dO6fx48crKSlJLVu21OrVqy2TnycmJsrB4f8u6Lp06ZKGDx+upKQkVaxYUQEBAdq2bZv8/f3t1QUAAAAAAAAUkN2LUpIUERFhefTwzWJjY63WZ8yYoRkzZtggKwAAAAAAABQXu84pBQAAAAAAgNKJohQAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAAAAAABsjqIUAAAAAAAAbI6iFAAAAAAAAGyOohQAAAAAAABsjqIUAAAAAAAAbM7J3gkAAGzH97WV9k4BdnL83WB7pwAAAABY4UopAAAAAAAA2BxFKQAAAAAAANgcRSkAAAAAAADYHEUpAAAAAAAA2BxFKQAAAAAAANgcRSkAAAAAAADYHEUpAAAAAAAA2BxFKQAAAAAAANgcRSkAAAAAAADYHEUpAAAAAAAA2BxFKQAAAAAAANgcRSkAAAAAAADYnJO9EwAAAAAAAIA139dW2vX8x98NLvZzcKUUAAAAAAAAbK5EFKXmzJkjX19fubq6ql27dtq5c+ct45cuXarGjRvL1dVVzZo106pVq2yUKQAAQOlU0PEaAADA7di9KLVkyRJFRkZqwoQJ2rNnj1q0aKGgoCCdPXs21/ht27Zp4MCBGjZsmPbu3auQkBCFhITowIEDNs4cAACgdCjoeA0AACA/7F6Umj59uoYPH64hQ4bI399f8+bNU9myZbVgwYJc42fNmqVevXrplVdekZ+fn6ZMmaLWrVtr9uzZNs4cAACgdCjoeA0AACA/7DrReUZGhuLj4zVu3DjLNgcHB3Xv3l1xcXG57hMXF6fIyEirbUFBQVq+fHmu8enp6UpPT7esp6amSpLMZvMdZn/3ykr/094pwI5K83sf/P6XZqX5dz+774Zh2DmTu1NhxmvFMf6y998ve/8O0X/6b0+luf+lue8S/af/he9/fsdfdi1KnT9/XpmZmfL29rba7u3trcOHD+e6T1JSUq7xSUlJucZHRUVp0qRJObbXqlWrkFkDdzfPmfbOAIA98Lsv/fHHH/L09LR3GnedwozX7sXxV2n/HaL/9s7Avkpz/0tz3yX6T//v/Bi3G3/ZtShlC+PGjbO6siorK0sXL15U5cqVZTKZ7JgZ7MFsNqtWrVo6efKkPDw87J0OABvhd790MwxDf/zxh6pXr27vVEqNkjb+Ku1/A+g//S+t/S/NfZfoP/23b//zO/6ya1GqSpUqcnR0VHJystX25ORk+fj45LqPj49PgeJdXFzk4uJita1ChQqFTxr3BA8Pj1L5hwko7fjdL724QqrwCjNeK6njr9L+N4D+0//S2v/S3HeJ/tN/+/U/P+Mvu0507uzsrICAAMXExFi2ZWVlKSYmRoGBgbnuExgYaBUvSevWrcszHgAAAIVXmPEaAABAftj99r3IyEiFhYWpTZs2uv/++zVz5kylpaVpyJAhkqTQ0FDVqFFDUVFRkqSRI0eqc+fOmjZtmoKDg7V48WLt3r1b8+fPt2c3AAAA7lm3G68BAAAUht2LUv3799e5c+c0fvx4JSUlqWXLllq9erVlMs3ExEQ5OPzfBV0dOnTQokWL9Oabb+r1119XgwYNtHz5cjVt2tReXcBdxMXFRRMmTMhxSwGAexu/+8Cdud14raQr7X8D6D/9L639L819l+g//b87+m8yeD4yAAAAAAAAbMyuc0oBAAAAAACgdKIoBQAAAAAAAJujKAUAAAAAAACboygFAAAAAAAAm6MoBQAAAAAAAJtzsncCQHE6f/68FixYoLi4OCUlJUmSfHx81KFDBw0ePFhVq1a1c4YAAKConTlzRnPnztWWLVt05swZOTg46L777lNISIgGDx4sR0dHe6cIAADElVK4h+3atUsNGzbUhx9+KE9PT3Xq1EmdOnWSp6enPvzwQzVu3Fi7d++2d5oA7ODkyZMaOnSovdMAUAx2794tPz8/rVq1SteuXdPRo0cVEBCgcuXK6eWXX1anTp30xx9/2DvNYpWRkaGvv/5ao0eP1sCBAzVw4ECNHj1aS5cuVUZGhr3TQzGbPXu2QkNDtXjxYknS559/Ln9/fzVu3Fivv/66rl+/bucMi9/vv/+uy5cv59h+7do1bd682Q4Z2c6FCxe0ceNGXbx4UdJfX9K/9957mjx5sn755Rc7Z2cf9913n44ePWrvNGzOMAxt3LhR//rXv7RixQpdu3bN3inlymQYhmHvJIDi0L59e7Vo0ULz5s2TyWSyajMMQ88//7x+/vlnxcXF2SlDAPayb98+tW7dWpmZmfZOBUARe/DBB9WjRw9NmDBBkvTFF19o9uzZ2r59uy5duqRu3bqpU6dOmjVrlp0zLR6//vqrgoKCdPr0abVr107e3t6SpOTkZO3YsUM1a9bUTz/9pPr169s5U/tITk7Wxx9/rPHjx9s7lWLx9ttva+rUqerZs6e2bt2qUaNG6f3339fo0aPl4OCgGTNm6IUXXtCkSZPsnWqxOHPmjB599FHFx8fLZDLpqaee0kcffSR3d3dJf/38q1evfs9+/u/cuVM9e/aU2WxWhQoVtG7dOvXr109OTk7KysrS6dOntWXLFrVu3dreqRaLDz/8MNftkZGRGjt2rHx8fCRJL730ki3TspmHH35YX331lTw9PXXx4kU9/PDD2rlzp6pUqaILFy6oYcOG2rx5c4m7W4iiFO5Zbm5u2rt3rxo3bpxr++HDh9WqVStduXLFxpkBKG4//PDDLdt/++03jRkz5p4dlAKlWdmyZXXgwAHdd999kqSsrCy5urrq5MmT8vb21rp16zR48GCdOnXKzpkWjx49eqhcuXL67LPP5OHhYdVmNpsVGhqqK1euaM2aNXbK0L7u9S8l6tevr6lTp+rxxx/Xvn37FBAQoE8//VSDBg2SJC1btkxjx469Z68aCQsL05EjRzR79mylpKTotddek8lk0tq1a1WxYkUlJyerWrVqysrKsneqxaJHjx7y9fXV9OnT9fHHH2vWrFnq1auX/vWvf0mShg4dqkuXLmnZsmV2zrR4ODg4qEaNGnJysp6l6MSJE6pevbrKlCkjk8mk3377zU4ZFi8HBwclJSXJy8tLL774ojZt2qQVK1aobt26+v333xUSEqK2bdtq7ty59k7VCnNK4Z7l4+OjnTt35lmU2rlzp+XbQwD3lpCQEJlMJt3qe5ebr6AEcG/w8vLSmTNnLEWp5ORkXb9+3VKgadCggeW2lnvR1q1btXPnzhwFKUny8PDQlClT1K5dOztkZhs///zzLduPHDlio0zs4/Tp02rTpo0kqUWLFnJwcFDLli0t7a1bt9bp06ftlF3xW79+vZYtW2Z5DbZu3ap+/fqpW7duiomJkXRvf/7Hx8frww8/VPny5TVy5Ei9+uqrGj58uKU9IiJCf/vb3+yYYfF67rnntGPHDi1atEh+fn6W7WXKlNHatWvl7+9vx+xsa8OGDZo6darq1q0rSapZs6bee+89q/dDSUFRCvesl19+Wc8995zi4+P10EMPWV2+HhMTo3/961/64IMP7JwlgOJQrVo1ffTRR3r00UdzbU9ISFBAQICNswJgCyEhIXr++ef1/vvvy8XFRVOmTFHnzp3l5uYm6a+iRI0aNeycZfGpUKGCjh8/rqZNm+bafvz4cVWoUMG2SdlQy5Yt8/xSInv7vVyU8PHx0aFDh1S7dm0dPXpUmZmZOnTokJo0aSJJOnjwoLy8vOycZfFJTU1VxYoVLesuLi767rvv1K9fP3Xt2lVffPGFHbMrfhkZGZa/dWXKlFHZsmVVpUoVS3v2bVz3qnnz5mnZsmUKCgrS2LFjFRERYe+UbC7779ulS5dUr149q7b69euXyKI0RSncs8LDw1WlShXNmDFDH330keUybUdHRwUEBCg6OlpPPvmknbMEUBwCAgIUHx+fZ1HqdldRAbh7vf322zpz5oz69OmjzMxMBQYGWv1D1GQyKSoqyo4ZFq9nn31WoaGheuutt3L9Uu7tt9/WiBEj7Jxl8alUqZKmTp2qhx56KNf2gwcPqk+fPjbOynYGDRqk0NBQPfroo4qJidHYsWP18ssv68KFCzKZTHrnnXf0xBNP2DvNYnPffffp559/VoMGDSzbnJyctHTpUvXr10+PPPKIHbMrfrVq1dJvv/0mX19fSdLixYtVrVo1S/uZM2esilT3oscee0z333+/QkNDtXLlSi1cuNDeKdnU4MGD5eLiomvXrunYsWOWgrQkJSUllcgvJShK4Z7Wv39/9e/fX9euXdP58+cl/fUNQZkyZeycGYDi9MorrygtLS3P9vr162vjxo02zAiArbi7u2vJkiW6evWqrl+/bpngOFvPnj3tlJltTJ48WeXKldP777+vMWPGWL41NwxDPj4+evXVVzV27Fg7Z1l8AgICdPr0adWpUyfX9pSUlHv6S4lJkybJzc1NcXFxGj58uF577TW1aNFCY8eO1Z9//qk+ffpoypQp9k6z2PTu3Vvz589X3759rbZnF6b69u2rkydP2im74jdgwACdPXvWsh4cHGzV/sMPP+j++++3dVo2V6NGDa1fv17vvvuuWrVqdU//zt8oLCzM8v+PPvqo/vzzT6v2b7/91up23pKCic4BAACAe9CxY8eUlJQk6a/burLnFrmXLVu2TGlpaXr66adzbb906ZJ++OEHq3+84d5x/fp1/fnnn7nOqZbdfurUqTyLlve6P//8U46OjnJxcbF3KjYTHx+vLVu2KDQ01OrWztIoLS1Njo6OcnV1tXcqVihKAQAAAKXEyZMnNWHCBC1YsMDeqQA2V9rf//Sf/pfE/jvYOwEAAAAAtnHx4kV9+umn9k7Dbk6ePKmhQ4faOw3YSWl//9N/+l8S+8+cUgAAAMA94ocffrhl+2+//WajTEqm7H+UlbQrBVA0Svv7n/7T/1spqf3n9j0ApUp0dLRGjRqllJQUe6cCAECRc3BwuO0TRk0mk+WpxPea/PyjbMyYMfds/0u70v7+p//0/27sP7fvAbhnDB48WCEhITm2x8bGymQyKSUlRf3799d///tfS9vEiRNL5FMoAAAojGrVqum7775TVlZWrsuePXvsnWKxCgkJ0WOPPaaQkJBcl8jISHuniGJU2t//9J/+3439pygFoFRxc3OTl5eXvdMAAKBYBAQEKD4+Ps/2232Lfre7W/9RhqJR2t//9J/+3439pygFoFSJjo5WhQoVLP8/adIk7du3TyaTSSaTSdHR0XbNDwCAO/HKK6+oQ4cOebbXr19fGzdutGFGtnW3/qMMRaO0v//pP/2/G/vPnFIA7hmDBw9WSkqKli9fbrU9NjZWXbt21aVLl7R8+XLLnFJXrlzRW2+9pdWrV2v9+vWSJE9PT7m5udkhewAAcKf+85//KC0tTb169cq1PS0tTbt371bnzp1tnBkAIDc8fQ/APWXFihVyd3e32pbXZH5ubm5yd3eXk5OTfHx8bJEeAAAoRh07drxle7ly5ShIAUAJQlEKwD2la9eumjt3rtW2HTt26Omnn7ZTRgAAAACA3FCUAnBPKVeunOrXr2+17ffff7dTNgAAAACAvDDROYBSzdnZOc/b+wAAAAAAxYeiFIBSzdfXV8eOHVNCQoLOnz+v9PR0e6cEAABgNyaTKcdDY250/PhxmUwmJSQk2CwnW4qNjZXJZFJKSoq9UwFKBYpSAEq1vn37qlevXuratauqVq2qr776yt4pAQCAUm7w4MEymUwymUxydnZW/fr1NXnyZF2/fr3IzjFx4kS1bNkyx/YzZ86od+/eRXae3JSUwk+XLl00atQou+YAlHbMKQXgnhEdHZ3r9i5dusgwDEl/DfIGDx5saXNxcdE333xjg+wAAADyr1evXlq4cKHS09O1atUqhYeHq0yZMho3blyxnpcnEgOwJa6UAgAAAIASxsXFRT4+PqpTp45eeOEFde/eXT/88IMkKT09XS+//LJq1KihcuXKqV27doqNjbXsGx0drQoVKmj58uVq0KCBXF1dFRQUpJMnT1raJ02apH379lmuyMr+cu/m2/d27typVq1aydXVVW3atNHevXtz5HrgwAH17t1b7u7u8vb21jPPPKPz588Xuu/57d+aNWvk5+cnd3d39erVS2fOnLHEXL9+XS+99JIqVKigypUr69VXX1VYWJhCQkIk/fVF5aZNmzRr1izLa3D8+HHL/vHx8WrTpo3Kli2rDh066MiRI4XuD4C8UZQCAAAAgBLOzc1NGRkZkqSIiAjFxcVp8eLF+vnnn9WvXz/16tVLR48etcT/+eefeuedd/TZZ59p69atSklJ0YABAyRJ/fv315gxY9SkSROdOXNGZ86cUf/+/XOc8/Lly3rkkUfk7++v+Ph4TZw4US+//LJVTEpKirp166ZWrVpp9+7dWr16tZKTk/Xkk08Wuq/57d8HH3ygzz//XJs3b1ZiYqJVbu+9956+/PJLLVy4UFu3bpXZbLYqts2aNUuBgYEaPny45TWoVauWpf2NN97QtGnTtHv3bjk5OWno0KGF7g+AvHH7HgAAAACUUIZhKCYmRmvWrNGIESOUmJiohQsXKjExUdWrV5ckvfzyy1q9erUWLlyof/zjH5Kka9euafbs2WrXrp0k6dNPP5Wfn5927typ+++/X+7u7nJycrrl7XqLFi1SVlaWPvnkE7m6uqpJkyb6/fff9cILL1hiZs+erVatWlnOK0kLFixQrVq19N///lcNGzYsUH8L0r958+apXr16kv4qZE2ePNlynH/+858aN26cHnvsMUueq1atsrR7enrK2dlZZcuWzfU1eOedd9S5c2dJ0muvvabg4GBdvXpVrq6uBeoPgFujKAUAAAAAJcyKFSvk7u6ua9euKSsrS0899ZQmTpyo2NhYZWZm5ij2pKenq3LlypZ1JycntW3b1rLeuHFjVahQQb/88ovuv//+fOXwyy+/qHnz5laFmMDAQKuYffv2aePGjXJ3d8+x///+978CF6X279+fr/6VLVvWUpCSpGrVquns2bOSpNTUVCUnJ1v109HRUQEBAcrKyspXHs2bN7c6tiSdPXtWtWvXLlB/ANwaRSkAAAAAKGG6du2quXPnytnZWdWrV5eT01//dLt8+bIcHR0VHx8vR0dHq31yKwwVt8uXL6tPnz567733crRlF3MKerz89K9MmTJWbSaTyfJgm6Jw4/FNJpMk5bugBSD/KEoBAAAAQAlTrlw51a9fP8f2Vq1aKTMzU2fPnlXHjh3z3P/69evavXu35WqhI0eOKCUlRX5+fpIkZ2dnZWZm3jIHPz8/ff7551a3rW3fvt0qpnXr1vr222/l6+trKZzdifz271Y8PT3l7e2tXbt2qVOnTpKkzMxM7dmzRy1btrTE5ec1AFC8mOgcAAAAAO4SDRs21KBBgxQaGqrvvvtOx44d086dOxUVFaWVK1da4sqUKaMRI0Zox44dio+P1+DBg9W+fXtLkcrX11fHjh1TQkKCzp8/r/T09Bzneuqpp2QymTR8+HAdOnRIq1at0gcffGAVEx4erosXL2rgwIHatWuX/ve//2nNmjUaMmTIbQs++/fvV0JCgmXZt29fvvt3OyNGjFBUVJS+//57HTlyRCNHjtSlS5csVz1lvwY7duzQ8ePHdf78ea6EAuyAohQAAAAA3EUWLlyo0NBQjRkzRo0aNVJISIh27dplNd9R2bJl9eqrr+qpp57SAw88IHd3dy1ZssTS3rdvX/Xq1Utdu3ZV1apV9dVXX+U4j7u7u3788Uft379frVq10htvvJHjNr3q1atr69atyszMVM+ePdWsWTONGjVKFSpUkIPDrf+52alTJ7Vq1cqyBAQE5Lt/t/Pqq69q4MCBCg0NVWBgoNzd3RUUFGQ1P9bLL78sR0dH+fv7q2rVqkpMTMz38QEUDZNRlDfeAgAAAADsKjo6WqNGjVJKSoq9UykxsrKy5OfnpyeffFJTpkyxdzoA/j/mlAIAAAAA3FNOnDihtWvXqnPnzkpPT9fs2bN17NgxPfXUU/ZODcANuH0PAAAAAHBPcXBwUHR0tNq2basHHnhA+/fv1/r16y0TvQMoGbh9DwAAAAAAADbHlVIAAAAAAACwOYpSAAAAAAAAsDmKUgAAAAAAALA5ilIAAAAAAACwOYpSAAAAAAAAsDmKUgAAAAAAALA5ilIAAAAAAACwOYpSAAAAAAAAsDmKUgAAAAAAALC5/wfXXm+pJRzzpAAAAABJRU5ErkJggg==",
            "text/plain": [
              "<Figure size 1200x800 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "# Number of unique values for peptides and alleles\n",
        "num_unique_peptides = train_data['peptide'].nunique()\n",
        "num_unique_alleles = train_data['allele'].nunique()\n",
        "print(f\"Number of unique peptides: {num_unique_peptides}\")\n",
        "print(f\"Number of unique alleles: {num_unique_alleles}\")\n",
        "\n",
        "# Analyze the 'hit' column\n",
        "hit_counts = train_data['hit'].value_counts()\n",
        "\n",
        "# Peptide length distribution\n",
        "train_data['peptide_length'] = train_data['peptide'].apply(len)\n",
        "peptide_length_distribution = train_data['peptide_length'].value_counts()\n",
        "\n",
        "plt.figure(figsize=(12, 8))\n",
        "\n",
        "# Hit distribution\n",
        "plt.subplot(2, 2, 3)\n",
        "hit_counts.plot(kind='bar')\n",
        "plt.title('Train Dataset Hit Distribution')\n",
        "plt.xlabel('Hit')\n",
        "plt.ylabel('Count')\n",
        "\n",
        "# Peptide length distribution\n",
        "plt.subplot(2, 2, 4)\n",
        "peptide_length_distribution.plot(kind='bar')\n",
        "plt.title('Train Dataset Peptide Length Distribution')\n",
        "plt.xlabel('Peptide Length')\n",
        "plt.ylabel('Count')\n",
        "\n",
        "plt.tight_layout()\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zJm633BcR69a"
      },
      "source": [
        "We can observe several important results. Firstly, nearly all the peptides are unique, but there are only 130 distinct alleles. The peptide length is almost uniformly distributed between 9 and 15 amino acids. Lastly, it's crucial to note that the dataset is unbalanced. The distribution of the hit label is 95% for class 0 and 5% for class 1, which can present several challenges when building a classifier.\n",
        "- Most machine learning algorithms assume balanced data, leading to bias towards the majority class.\n",
        "- Accuracy is unreliable with imbalanced data, as high accuracy can be achieved by always predicting the majority class, failing to reflect true performance on the minority class.\n",
        "- Models trained on imbalanced data may overfit to the majority class, resulting in poor generalization and performance on new minority class examples.\n",
        "\n",
        "There is many techniques to adress unbalanced dataset. On the first part of the dataset we will observe how good perform classical machine learning models without resampling the dataset. We will use metrics such as **precision**, **recall**, **F1-score** or also **ROC-AUC** to evaluate them.\n",
        "\n",
        "In the second part of this work, we will test these models as well as a Deep Learning model on a undersampled balanced dataset.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EUZiKa46ca6Z"
      },
      "source": [
        "## 2. Method Explanation\n",
        "There is two approaches to predict MHCI presentation based on the peptide and the MHC allele. The first one is to build a specific model for each existing HLAs. The advantage of this approach is that we don't need to encode the allele data. However, this approach takes a long time to train and cannot handle unseen HLAs.\n",
        "\n",
        "The second approach is to use the sequences of both the peptides and the HLAs to construct a single pan-specific model for an entire class of HLAs as in (Reynisson et al., 2020). As opposed to the first approach, we train only one model that will take an encoding of the HLAs, and will therfore be able to handle unseen alleles. The challenge here is to find the right encoding for HLAs.\n",
        "\n",
        "In this work, we will use the second approach because the training dataset does not include all existing alleles, and this method allows for better generalization.\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kzsRU05szanX"
      },
      "source": [
        "### Feature Engineering\n",
        "#### Peptides\n",
        "Peptide are represented by a sequence of characters, each one encoding a specific amino acid. Since there is 20 differents amino acids, there is 20 different encoding characters.\n",
        "The amino acids are categorical data, which means that we need to encode them in order to feed a model. We can use a one hot encoder to perfom this encoding.\n",
        "As a result we get the following vectors:\n",
        "- A: [1,0,0,0,...]\n",
        "- C: [0,1,0,0,...]\n",
        "- E: [0,0,1,0,...]\n",
        "\n",
        "We then concatenate all the one hot encoded vector of each amino acid. Since, we observed that there was peptides of different lengths, we chose to encode all of them in a vector of size equal to the maximum peptide length. This allows us to preserve the structure information contain in the sequence.\n",
        "\n",
        "As a result, we get a feature vector of shape (20*`max_peptide_length`).\n",
        "\n",
        "We also save the initial peptide as it might be a usefull information for the mdoels.\n",
        "\n",
        "#### HLA allele\n",
        "Based on the specification on the [nomenclature of HLA Alleles](https://hla.alleles.org/nomenclature/naming.html), we treat the HLA allele as categorical data. Moreover, we decided distinguish the gene and the subtype of the allele.\n",
        "\n",
        "There is two ways of handling these categorical data. The first one is to manually encode the data. The second approach is to let the model learn an representation of these data called *embedding*. We will test both approaches.\n",
        "\n",
        "The training dataset does not include all possible genes or allele subtypes. Therefore, we create a vocabulary based on the genes and subtypes present in the training data. Each unique gene or subtype is assigned an index from 1 to the number of unique values, which will serve as their one-hot encoding. Any unseen values encountered in the test set will be assigned an index of 0."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tUYAFJF_zeLt"
      },
      "source": [
        "### Models\n",
        "We will use 3 different models:\n",
        "\n",
        "#### Random Forest\n",
        "It is an ensemble learning method for classification, that constructs a multitude of decision trees at training time. Here are the reason why we choose Random Forest:\n",
        "- Works well with big datasets and high dimensionality data (which is our case)\n",
        "- Easily capture non-linear pattern\n",
        "- Computationally efficient and low training time because each decision-tree can be train in parallel\n",
        "- Less prone to overfit as it tends to reduce the variance\n",
        "\n",
        "#### Extreme Gradient Boosting (XGBoost)\n",
        "XGBoost is also an esemble learning method which is based on gradient boosting. Gradient boosting methods train iteratively a decision tree based on the accuracy of the previous one.\n",
        "- Well suited for large datasets, more than Random Forest\n",
        "- Highly performant with very low training time thanks to many optimizations\n",
        "\n",
        "\n",
        "#### Custom Deep Learning Model\n",
        "Since the increasing amount of data about MHCI-peptide presentation, artificials neural networks (ANNs) are becoming more and more popular for this type of problem. In the recent years, ANNs were found among the top-performed methods (Luo et al., 2015).\n",
        "\n",
        "Various neural network architectures are suitable for this problem. We opted to develop our own convolutional neural network (CNN) architecture. CNNs are effective for this type of problem due to their capacity of learning spatial structures, which is beneficial for capturing peptide structure information. Additionally, CNNs are relatively simple architectures and easy to implement.\n",
        "\n",
        "Our architecture is based on DeepSeqPan (Liu et al., 2019) and the DCNN from Han et al., 2017.\n",
        "\n",
        "It consist of 3 modules. The first one, `PeptideEncoder` use a CNN to encode petpide data and capture its spatial structure. The second one, `HLAEncoder` is a multi layer perceptron (MLP) that learn an embedding of the HLAs. Finally, we concatenate them in the `NetMHC` and pass it throughout a MLP."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "id": "X3eeJWWhjT8P"
      },
      "outputs": [],
      "source": [
        "amino_acids = 'ACEDGFIHKMLNQPSRTWVY'\n",
        "\n",
        "# mapping a unique integer to each amino acid\n",
        "aa2int = dict((c, i) for i, c in enumerate(amino_acids))\n",
        "\n",
        "def one_hot_encode_peptide(peptide, max_peptide_length):\n",
        "    \"\"\"One-hot encodes a peptide sequence\n",
        "\n",
        "    Args:\n",
        "        peptide (str): The peptide sequence to encode.\n",
        "        max_peptide_length (int): Maximum peptide length\n",
        "\n",
        "    Returns:\n",
        "        numpy.ndarray: One-hot encoded matrix of shape (max_peptide_length * len(amino_acids)).\n",
        "    \"\"\"\n",
        "    # Encode each character in the peptide\n",
        "    integer_encoded = [aa2int.get(char, len(amino_acids)) for char in peptide]\n",
        "\n",
        "    # Pad sequence if shorter than max_peptide_length\n",
        "    if len(integer_encoded) < max_peptide_length:\n",
        "        integer_encoded += [len(amino_acids)] * (max_peptide_length - len(integer_encoded))\n",
        "    elif len(integer_encoded) > max_peptide_length:\n",
        "        integer_encoded = integer_encoded[:max_peptide_length]\n",
        "\n",
        "    # One-hot encoding\n",
        "    onehot_encoded = np.zeros((max_peptide_length, len(amino_acids)), dtype=np.int8)\n",
        "\n",
        "    for i, value in enumerate(integer_encoded):\n",
        "        if value < len(amino_acids):\n",
        "            onehot_encoded[i, value] = 1\n",
        "\n",
        "    return onehot_encoded"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "metadata": {
        "id": "7qgz1SYGp_hb"
      },
      "outputs": [],
      "source": [
        "def build_hla_gene_vocabulary(alleles):\n",
        "    '''Build a vocabulary of each unique gene encountered\n",
        "\n",
        "    Args:\n",
        "        alleles (pd.Series): Serie of string of the pattern 'Yxx:xx',\n",
        "        where Y is a char representig the gene and each x a digit representig the subtype\n",
        "\n",
        "    Returns:\n",
        "        dict: Dictionary of unique genes\n",
        "    '''\n",
        "    splitted_alleles = alleles.str.split('-', expand=True)[1]\n",
        "    hla_genes = splitted_alleles.str[0]\n",
        "    # +1 to reserve 0 index for the unseen data\n",
        "    vocabulary = {gene: i+1 for i, gene in enumerate(hla_genes.unique())}\n",
        "    return vocabulary\n",
        "\n",
        "def build_hla_subtype_vocabulary(alleles):\n",
        "    '''\n",
        "    Build a vocabulary of each unique subtype encountered\n",
        "\n",
        "    Args:\n",
        "        alleles (pd.Series): Serie of string of the pattern 'Yxx:xx',\n",
        "        where Y is a char representig the gene and each x a digit representig the subtype\n",
        "\n",
        "    Returns:\n",
        "        dict: Dictionary of unique subtype\n",
        "    '''\n",
        "    splitted_alleles = alleles.str.split('-', expand=True)[1]\n",
        "    subtypes = splitted_alleles.str[1:]\n",
        "    # +1 to reserve 0 index for the unseen data\n",
        "    vocabulary = {subtype: i+1 for i, subtype in enumerate(subtypes.unique())}\n",
        "    return vocabulary"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 20,
      "metadata": {
        "id": "U-UDI6LSL4RQ"
      },
      "outputs": [],
      "source": [
        "def encode_data(data, gene_vocab, subtype_vocab):\n",
        "    '''Encode the data about MHC-Peptide presentation\n",
        "\n",
        "    Args:\n",
        "        data (pd.DataFrame): The original data\n",
        "        gene_vocab (dict): gene vocabulary\n",
        "        subtype_vocab (dict): subtype vocabulary\n",
        "\n",
        "    Returns:\n",
        "        pd.DataFrame: The encoded data\n",
        "    '''\n",
        "    encoded_data = pd.DataFrame()\n",
        "\n",
        "    # Extract gene and subtype from HLA allele\n",
        "    splitted_allele = data['allele'].str.split('-', expand=True)[1]\n",
        "    gene = splitted_allele.str[0]\n",
        "    subtype = splitted_allele.str[1:]\n",
        "\n",
        "    # Build encoded DataFrame\n",
        "    encoded_data = pd.DataFrame()\n",
        "    encoded_data['encoded_peptide'] = data['peptide'].apply(lambda x: one_hot_encode_peptide(x, 15).flatten())\n",
        "    encoded_data['peptide_length'] = data['peptide'].apply(lambda x: len(x))\n",
        "    encoded_data['encoded_gene'] = gene.apply(lambda x: gene_vocab.get(x, 0))\n",
        "    encoded_data['encoded_subtype'] = subtype.apply(lambda x: subtype_vocab.get(x, 0))\n",
        "    encoded_data['hit'] = data['hit']\n",
        "\n",
        "    return encoded_data\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 21,
      "metadata": {
        "id": "6PZbwvOMMLW-"
      },
      "outputs": [],
      "source": [
        "def encoded_data2numpy(train_data_encoded, test_data_encoded):\n",
        "    '''Convert encoded data to numpy array\n",
        "\n",
        "    Args:\n",
        "        train_data_encoded (pd.DataFrame): Encoded training data\n",
        "        test_data_encoded (pd.DataFrame): Encoded test data\n",
        "\n",
        "    Returns:\n",
        "        np.ndarray: X_train, X_test, y_train, y_test\n",
        "    '''\n",
        "    X_train = np.hstack([\n",
        "        np.stack(train_data_encoded['encoded_peptide'].to_numpy()),\n",
        "        train_data_encoded['peptide_length'].to_numpy().reshape(-1,1),\n",
        "        train_data_encoded['encoded_gene'].to_numpy().reshape(-1,1),\n",
        "        train_data_encoded['encoded_subtype'].to_numpy().reshape(-1,1)\n",
        "    ])\n",
        "\n",
        "    X_test = np.hstack([\n",
        "        np.stack(test_data_encoded['encoded_peptide'].to_numpy()),\n",
        "        test_data_encoded['peptide_length'].to_numpy().reshape(-1,1),\n",
        "        test_data_encoded['encoded_gene'].to_numpy().reshape(-1,1),\n",
        "        test_data_encoded['encoded_subtype'].to_numpy().reshape(-1,1)\n",
        "    ])\n",
        "\n",
        "    y_train = train_data_encoded['hit'].to_numpy()\n",
        "    y_test  = test_data_encoded['hit'].to_numpy()\n",
        "\n",
        "    return X_train, X_test, y_train, y_test"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 22,
      "metadata": {
        "id": "KDUy5lMm8BpC"
      },
      "outputs": [],
      "source": [
        "def undersample_balance_data(data):\n",
        "    '''Balance the data by undersampling it such that both classes are evenly distributed\n",
        "\n",
        "    Args:\n",
        "        data (pd.DataFrame): Data containing the 'hit' column\n",
        "\n",
        "    Returns:\n",
        "        pd.DataFrame: balanced sample of the data\n",
        "    '''\n",
        "    class_0 = data[data['hit'] == 0]\n",
        "    class_1 = data[data['hit'] == 1]\n",
        "\n",
        "    n_samples = min(len(class_0), len(class_1))\n",
        "\n",
        "    class_0_balanced = class_0.sample(n_samples, random_state=42)\n",
        "    class_1_balanced = class_1.sample(n_samples, random_state=42)\n",
        "\n",
        "    data_balanced = pd.concat([class_0_balanced, class_1_balanced], ignore_index=True)\n",
        "    return data_balanced"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 23,
      "metadata": {
        "id": "tLJIASei6eMK"
      },
      "outputs": [],
      "source": [
        "models = {\n",
        "    'random_forest': RandomForestClassifier(random_state=42),\n",
        "    'xg_boost': xgb.XGBClassifier(random_state=42)\n",
        "}\n",
        "\n",
        "def train_models(models_dict, X_train, y_train):\n",
        "    for name, model in models_dict.items():\n",
        "        start_time = time.time()\n",
        "        model.fit(X_train, y_train)\n",
        "        end_time = time.time()\n",
        "        print(f'{name} trained in {end_time - start_time} seconds')\n",
        "\n",
        "def test_models(models_dict, X_test, y_test):\n",
        "    for name, model in models_dict.items():\n",
        "        y_pred = model.predict(X_test)\n",
        "        test_accuracy = accuracy_score(y_test, y_pred)\n",
        "        print(f\"{name} Test Accuracy: {test_accuracy}\")\n",
        "        print(\"Classification Report:\")\n",
        "        print(classification_report(y_test, y_pred))\n",
        "        print(\"ROC Curve\")\n",
        "        ax = plt.gca()\n",
        "        rfc_disp = RocCurveDisplay.from_estimator(model, X_test, y_test, ax=ax, alpha=0.8)\n",
        "        plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 24,
      "metadata": {
        "id": "Suj4GcCi6awd"
      },
      "outputs": [],
      "source": [
        "class PeptideEncoder(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(PeptideEncoder, self).__init__()\n",
        "        self.conv1 = nn.Conv2d(1, 32, kernel_size=3, padding=1)\n",
        "        self.bn1 = nn.BatchNorm2d(32)\n",
        "        self.conv2 = nn.Conv2d(32, 1, kernel_size=3, padding=1)\n",
        "        self.bn2 = nn.BatchNorm2d(1)\n",
        "        self.leaky_relu = nn.LeakyReLU()\n",
        "\n",
        "    def forward(self, x):\n",
        "        out = self.conv1(x)\n",
        "        out = self.bn1(out)\n",
        "        out = self.leaky_relu(out)\n",
        "        out = self.conv2(out)\n",
        "        out = self.bn2(out)\n",
        "        out = self.leaky_relu(out)\n",
        "        return out\n",
        "\n",
        "class HLAEncoder(nn.Module):\n",
        "    def __init__(\n",
        "        self,\n",
        "        gene_vocab_size,\n",
        "        gene_embed_size,\n",
        "        subtype_vocab_size,\n",
        "        subtype_embed_size\n",
        "    ):\n",
        "        super(HLAEncoder, self).__init__()\n",
        "        self.gene_embedding = nn.Embedding(gene_vocab_size, gene_embed_size)\n",
        "        self.subtype_embedding = nn.Embedding(subtype_vocab_size, subtype_embed_size)\n",
        "        self.fc1 = nn.Linear(gene_embed_size + subtype_embed_size, 256)\n",
        "        self.dropout = nn.Dropout(p=0.5)\n",
        "        self.fc2 = nn.Linear(256, 212)\n",
        "        self.leaky_relu = nn.LeakyReLU()\n",
        "\n",
        "    def forward(self, gene_idx, subtype_idx):\n",
        "        gene_embed = self.gene_embedding(gene_idx).squeeze(1)\n",
        "        subtype_embed = self.subtype_embedding(subtype_idx).squeeze(1)\n",
        "        combined = torch.cat((gene_embed, subtype_embed), dim=1)\n",
        "\n",
        "        out = self.fc1(combined)\n",
        "        out = self.leaky_relu(out)\n",
        "        out = self.dropout(out)\n",
        "        out = self.fc2(out)\n",
        "        out = self.leaky_relu(out)\n",
        "        return out\n",
        "\n",
        "class NetMHC(nn.Module):\n",
        "    def __init__(\n",
        "        self,\n",
        "        gene_vocab_size,\n",
        "        gene_embed_size,\n",
        "        subtype_vocab_size,\n",
        "        subtype_embed_size\n",
        "    ):\n",
        "        super(NetMHC, self).__init__()\n",
        "        self.peptide_encoder = PeptideEncoder()\n",
        "        self.hla_encoder = HLAEncoder(gene_vocab_size, gene_embed_size, subtype_vocab_size, subtype_embed_size)\n",
        "        self.fc1 = nn.Linear(512, 128)\n",
        "        self.dropout = nn.Dropout(p=0.5)\n",
        "        self.fc2 = nn.Linear(128, 1)\n",
        "        self.leaky_relu = nn.LeakyReLU()\n",
        "        self.sigmoid = nn.Sigmoid()\n",
        "\n",
        "    def forward(self, x):\n",
        "        peptides, genes, subtypes = x\n",
        "        peptide_encoded = self.peptide_encoder(peptides)\n",
        "        peptide_encoded = peptide_encoded.view(-1, 300)\n",
        "        hla_encoded = self.hla_encoder(genes, subtypes)\n",
        "        combined = torch.cat((peptide_encoded, hla_encoded), dim=1)\n",
        "\n",
        "        out = self.fc1(combined)\n",
        "        out = self.leaky_relu(out)\n",
        "        out = self.dropout(out)\n",
        "        out = self.fc2(out)\n",
        "        out = self.sigmoid(out)\n",
        "\n",
        "        return out"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 25,
      "metadata": {
        "id": "mAintoXk6lmh"
      },
      "outputs": [],
      "source": [
        "class MHCDataset(Dataset):\n",
        "    def __init__(self, features, labels):\n",
        "\n",
        "        self.gene_features = features['encoded_gene'].to_numpy().reshape(-1, 1)\n",
        "        self.subtype_features = features['encoded_gene'].to_numpy().reshape(-1, 1)\n",
        "        self.peptide_features = np.stack(features['encoded_peptide'].to_numpy()).reshape(-1, 1, 15, 20)\n",
        "\n",
        "        self.labels = labels.to_numpy()\n",
        "\n",
        "        assert len(self.peptide_features) == len(self.gene_features) == len(self.subtype_features)== len(self.labels)\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.peptide_features)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        return (\n",
        "            torch.tensor(self.peptide_features[idx], dtype=torch.float32),\n",
        "            torch.tensor(self.gene_features[idx], dtype=torch.long),\n",
        "            torch.tensor(self.subtype_features[idx], dtype=torch.long),\n",
        "            torch.tensor(self.labels[idx], dtype=torch.float32)\n",
        "        )"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pBu1cQoc63hq"
      },
      "source": [
        "## 3. Experiment Description\n",
        "\n",
        "### Datasets\n",
        "\n",
        "Given the fact that we are dealing with unbalanced data, we train the Random Forest and XGBoost models on two distinct datasets. The first dataset is a random sample from the full training set, which remains unbalanced. The second dataset is a balanced subset of the full dataset. Both datasets are of equal size.\n",
        "\n",
        "Regarding our neural network model, we will not train it on the unbalanced dataset because preliminary experiments showed that the loss function was too low, preventing the model from training effectively.\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GNFG_TFA8JwR"
      },
      "source": [
        "### Metrics\n",
        "We will use the following metrics to evaluate our models:\n",
        "\n",
        "\n",
        "In the following, $tp$ denote true positive, $fp$ denote false positive, $tn$ denote true negative and $fn$ denote false negative.\n",
        "\n",
        "- **precision**: $\\frac{tp}{tp+fp}$ (How accurate the model is among the predicted positives)\n",
        "- **recall**: $\\frac{tp}{tp+fn}$ (How the model correctly identify True Positives)\n",
        "- **f1-score**: $2*\\frac{precision*recall}{precision+recall}$ (Balance between precision and recall)\n",
        "- **ROC AUC**:  Area Under the Receiver Operating Characteristic Curve (How well the model distinguishes between positive and negative classes)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 26,
      "metadata": {
        "id": "xrxuRlCINS8i"
      },
      "outputs": [],
      "source": [
        "# Get a sample of the full dataset (we take the same size of the balanced dataset we will use afterwards)\n",
        "train_data_sample = train_data.sample(90832, random_state=42)\n",
        "\n",
        "# Drop the '*' character to align with the test data format\n",
        "train_data_sample['allele'] = train_data_sample['allele'].str.replace('*', '', regex=False)\n",
        "\n",
        "gene_vocab = build_hla_gene_vocabulary(train_data_sample['allele'])\n",
        "subtype_vocab = build_hla_subtype_vocabulary(train_data_sample['allele'])\n",
        "\n",
        "train_data_encoded = encode_data(train_data_sample, gene_vocab, subtype_vocab)\n",
        "test_data_encoded = encode_data(test_data, gene_vocab, subtype_vocab)\n",
        "\n",
        "X_train, X_test, y_train, y_test = encoded_data2numpy(train_data_encoded, test_data_encoded)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 27,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "sE3PK_Otw8oy",
        "outputId": "99d4245e-0f65-4d99-a12a-ead901b9ea83"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "random_forest trained in 24.304145097732544 seconds\n",
            "xg_boost trained in 4.035407781600952 seconds\n",
            "random_forest Test Accuracy: 0.9554305330812215\n",
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.96      1.00      0.98    900725\n",
            "           1       0.85      0.09      0.16     45416\n",
            "\n",
            "    accuracy                           0.96    946141\n",
            "   macro avg       0.90      0.54      0.57    946141\n",
            "weighted avg       0.95      0.96      0.94    946141\n",
            "\n",
            "ROC Curve\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "xg_boost Test Accuracy: 0.9596286388603813\n",
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.96      0.99      0.98    900725\n",
            "           1       0.71      0.27      0.39     45416\n",
            "\n",
            "    accuracy                           0.96    946141\n",
            "   macro avg       0.84      0.63      0.68    946141\n",
            "weighted avg       0.95      0.96      0.95    946141\n",
            "\n",
            "ROC Curve\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "train_models(models, X_train, y_train)\n",
        "test_models(models, X_test, y_test)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 28,
      "metadata": {
        "id": "Y8tSHsEZDQm9"
      },
      "outputs": [],
      "source": [
        "# Get a balanced sample of the full dataset\n",
        "train_data_sample = undersample_balance_data(train_data)\n",
        "\n",
        "# Drop the '*' character to align with the test data format\n",
        "train_data_sample['allele'] = train_data_sample['allele'].str.replace('*', '', regex=False)\n",
        "\n",
        "gene_vocab = build_hla_gene_vocabulary(train_data_sample['allele'])\n",
        "subtype_vocab = build_hla_subtype_vocabulary(train_data_sample['allele'])\n",
        "\n",
        "train_data_encoded = encode_data(train_data_sample, gene_vocab, subtype_vocab)\n",
        "test_data_encoded = encode_data(test_data, gene_vocab, subtype_vocab)\n",
        "\n",
        "X_train, X_test, y_train, y_test = encoded_data2numpy(train_data_encoded, test_data_encoded)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 29,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "-TJUytTJDsDU",
        "outputId": "08064f31-efef-4044-8c6a-029691266b19"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "random_forest trained in 184.6155140399933 seconds\n",
            "xg_boost trained in 19.85626196861267 seconds\n",
            "random_forest Test Accuracy: 0.9324138791152693\n",
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.99      0.94      0.96    900725\n",
            "           1       0.39      0.74      0.51     45416\n",
            "\n",
            "    accuracy                           0.93    946141\n",
            "   macro avg       0.69      0.84      0.74    946141\n",
            "weighted avg       0.96      0.93      0.94    946141\n",
            "\n",
            "ROC Curve\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "xg_boost Test Accuracy: 0.8932844047557393\n",
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.99      0.90      0.94    900725\n",
            "           1       0.28      0.78      0.41     45416\n",
            "\n",
            "    accuracy                           0.89    946141\n",
            "   macro avg       0.63      0.84      0.68    946141\n",
            "weighted avg       0.95      0.89      0.92    946141\n",
            "\n",
            "ROC Curve\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "train_models(models, X_train, y_train)\n",
        "test_models(models, X_test, y_test)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 32,
      "metadata": {
        "id": "z8b5GJQfL4EU"
      },
      "outputs": [],
      "source": [
        "# Training Hyperparameters\n",
        "batch_size = 32\n",
        "learning_rate = 0.001\n",
        "n_epochs = 20\n",
        "\n",
        "# Extract training and validation sets\n",
        "features_train, features_val, labels_train, labels_val = train_test_split(train_data_encoded, train_data_sample['hit'], test_size=0.2, random_state=42)\n",
        "\n",
        "train_set = MHCDataset(features_train, labels_train)\n",
        "val_set = MHCDataset(features_val, labels_val)\n",
        "\n",
        "train_loader = DataLoader(train_set, batch_size=batch_size, shuffle=True)\n",
        "val_loader = DataLoader(val_set, batch_size=batch_size, shuffle=False)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 33,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OlTfXJJGD5w1",
        "outputId": "2c6f3241-4f59-40b8-9b83-c66b3f198083"
      },
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.88  val_loss: 0.01, val_accuracy: 0.91\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.91  val_loss: 0.01, val_accuracy: 0.92\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.91  val_loss: 0.01, val_accuracy: 0.92\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.91  val_loss: 0.01, val_accuracy: 0.92\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.91  val_loss: 0.01, val_accuracy: 0.92\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.92\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.92\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.92\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.92\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.92\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": []
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "  train_loss: 0.01, train_accuracy: 0.92  val_loss: 0.01, val_accuracy: 0.93\n",
            "NetMHC Training time: 1099.5886719226837 seconds\n"
          ]
        }
      ],
      "source": [
        "# Get device\n",
        "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
        "\n",
        "# +1 to handle unseen data\n",
        "model = NetMHC(len(gene_vocab)+1, 3, len(subtype_vocab)+1, 150).to(device)\n",
        "criterion = nn.BCELoss()\n",
        "optimizer = optim.Adam(model.parameters(), lr=learning_rate)\n",
        "\n",
        "\n",
        "# NetMHC Training\n",
        "train_loss_hist, train_acc_hist = [], []\n",
        "val_loss_hist, val_acc_hist = [], []\n",
        "\n",
        "start_time = time.time()\n",
        "\n",
        "for epoch in range(n_epochs):\n",
        "    model.train()\n",
        "    train_correct = 0\n",
        "    train_loss = 0.0\n",
        "\n",
        "    for peptide_features, gene_features, subtype_features, labels in \\\n",
        "        tqdm(train_loader, desc=f'Epoch {epoch+1}/{n_epochs} - Training', leave=False):\n",
        "\n",
        "        peptide_features = peptide_features.to(device)\n",
        "        gene_features = gene_features.to(device)\n",
        "        subtype_features = subtype_features.to(device)\n",
        "        labels = labels.to(device)\n",
        "\n",
        "        x = (peptide_features, gene_features, subtype_features)\n",
        "        y = labels.unsqueeze(1)\n",
        "\n",
        "        optimizer.zero_grad()\n",
        "\n",
        "        # Forward pass\n",
        "        logit = model.forward(x)\n",
        "        loss = criterion(logit, y)\n",
        "\n",
        "        # Backward pass and optimization\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "        train_correct += (logit.round() == y).sum().item()\n",
        "        train_loss += loss.item()\n",
        "\n",
        "    train_avg_loss = train_loss / len(train_set)\n",
        "    train_loss_hist.append(train_avg_loss)\n",
        "    train_avg_acc = train_correct / len(train_set)\n",
        "    train_acc_hist.append(train_avg_acc)\n",
        "\n",
        "\n",
        "    # Eval\n",
        "    model.eval()\n",
        "    val_correct = 0\n",
        "    val_loss = 0.0\n",
        "\n",
        "    with torch.no_grad():\n",
        "        for peptide_features, gene_features, subtype_features, labels in \\\n",
        "            tqdm(val_loader, desc=f'Epoch {epoch}/{n_epochs} - Validation', leave=False):\n",
        "\n",
        "            peptide_features = peptide_features.to(device)\n",
        "            gene_features = gene_features.to(device)\n",
        "            subtype_features = subtype_features.to(device)\n",
        "            labels = labels.to(device)\n",
        "\n",
        "            x = (peptide_features, gene_features, subtype_features)\n",
        "            y = labels.unsqueeze(1)\n",
        "\n",
        "            # Forward pass\n",
        "            logit = model.forward(x)\n",
        "            loss = criterion(logit, y)\n",
        "\n",
        "            val_correct += (logit.round() == y).sum().item()\n",
        "            val_loss += loss.item()\n",
        "\n",
        "    val_avg_loss = val_loss / len(val_set)\n",
        "    val_loss_hist.append(val_avg_loss)\n",
        "    val_avg_acc = val_correct / len(val_set)\n",
        "    val_acc_hist.append(val_avg_acc)\n",
        "\n",
        "\n",
        "    print(f'  train_loss: {train_avg_loss:.2f}, train_accuracy: {train_avg_acc:.2f}' \\\n",
        "          f'  val_loss: {val_avg_loss:.2f}, val_accuracy: {val_avg_acc:.2f}')\n",
        "\n",
        "# Save the model\n",
        "torch.save(model.state_dict(), os.path.join(ROOT_DIR, 'netmhc_balanced.pth'))\n",
        "\n",
        "end_time = time.time()\n",
        "print(f'NetMHC Training time: {end_time - start_time} seconds')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 34,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 904
        },
        "id": "WY7jBBy0PJyn",
        "outputId": "ad17f679-045d-427c-9cdd-63495221f98c"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "epochs = range(1, n_epochs + 1)\n",
        "\n",
        "# Plot the loss graph\n",
        "plt.figure()\n",
        "plt.plot(epochs, train_loss_hist, label='Train')\n",
        "plt.plot(epochs, val_loss_hist, label='Validation')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Loss')\n",
        "plt.legend(['Train', 'Validation'], loc='upper right')\n",
        "plt.title('Training and Validation Loss')\n",
        "plt.show()\n",
        "\n",
        "# Plot the accuracy graph\n",
        "plt.figure()\n",
        "plt.plot(epochs, train_acc_hist, label='Train')\n",
        "plt.plot(epochs, val_acc_hist, label='Validation')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "\n",
        "plt.legend(['Train', 'Validation'], loc='upper left')\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 35,
      "metadata": {
        "id": "2nzeB9eWN9t6"
      },
      "outputs": [],
      "source": [
        "test_set = MHCDataset(test_data_encoded, test_data_encoded['hit'])\n",
        "test_loader = DataLoader(test_set, batch_size=batch_size, shuffle=False)\n",
        "\n",
        "model.eval()\n",
        "\n",
        "y_probas = []\n",
        "\n",
        "with torch.no_grad():\n",
        "    for peptide_features, gene_features, subtype_features, _ in test_loader:\n",
        "      peptide_features = peptide_features.to(device)\n",
        "      gene_features = gene_features.to(device)\n",
        "      subtype_features = subtype_features.to(device)\n",
        "\n",
        "      x = (peptide_features, gene_features, subtype_features)\n",
        "\n",
        "      # Forward pass\n",
        "      logit = model(x)\n",
        "\n",
        "      y_probas.extend(logit.cpu().numpy())"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 37,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 680
        },
        "id": "qZA8XSRrOaPy",
        "outputId": "1b7dceb4-44ea-47e0-8aea-bdca964cacf3"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "NetMHC Test Accuracy: 0.9119370157302136\n",
            "Classification Report:\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.99      0.92      0.95    900725\n",
            "           1       0.33      0.80      0.47     45416\n",
            "\n",
            "    accuracy                           0.91    946141\n",
            "   macro avg       0.66      0.86      0.71    946141\n",
            "weighted avg       0.96      0.91      0.93    946141\n",
            "\n",
            "ROC Curve\n"
          ]
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "test_accuracy = accuracy_score(y_test, np.round(y_probas))\n",
        "print(f\"NetMHC Test Accuracy: {test_accuracy}\")\n",
        "print(\"Classification Report:\")\n",
        "print(classification_report(y_test, np.round(y_probas)))\n",
        "print(\"ROC Curve\")\n",
        "# Compute ROC curve and ROC area\n",
        "fpr, tpr, _ = roc_curve(y_test, y_probas)\n",
        "roc_auc = auc(fpr, tpr)\n",
        "\n",
        "# Plot ROC curve\n",
        "plt.figure()\n",
        "plt.plot(fpr, tpr, lw=2, label='ROC curve (area = %0.2f)' % roc_auc)\n",
        "plt.xlim([0.0, 1.0])\n",
        "plt.ylim([0.0, 1.05])\n",
        "plt.xlabel('False Positive Rate')\n",
        "plt.ylabel('True Positive Rate')\n",
        "plt.title('Receiver Operating Characteristic')\n",
        "plt.legend(loc=\"lower right\")\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CZGQcOc6bDub"
      },
      "source": [
        "## 4. Result Analysis\n",
        "\n",
        "### Unbalanced Dataset\n",
        "First, the **training time** is 24s for Random Forest and 4s for XGBoost, so XGBoost is training 6 times faster.\n",
        "\n",
        "Then, we notice that **precision**, **recall** and **f1-score** are close to 1 for the negative class ('0' label) but significantly lower for the positive class ('1' label). This indicates that the model is really good at predicting true negative but struggle to predict the true positive. The imbalance in the dataset could be a reason for this issue.\n",
        "\n",
        "Because the metrics are the same for both model regarding the '0' class, we summarize the metrics **only for '1' class** in the following table:\n",
        "\n",
        "|               | precision | recall | f1-score | ROC AUC |\n",
        "| --------------| --------- | ------ | -------- | ------- |\n",
        "| Random Forest | 0.85      | 0.09   | 0.16     | 0.88    |\n",
        "| XGBoost       | 0.71      | 0.27   | 0.39     | 0.90    |\n",
        "\n",
        "As a result, wee see that the XGBoost classifier is performing better than the Random Forest classifier on the Unbalanced Dataset, as well as being faster to train."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "P497loPEw422"
      },
      "source": [
        "### Balanced Dataset\n",
        "\n",
        "First, the training time for Random Forest is 185 seconds, while for XGBoost it is 20 seconds, making XGBoost nearly 10 times faster than Random Forest. NetMHC takes 1100 seconds, which is significantly longer.\n",
        "\n",
        "Then, the three models have almost the same metrics for the negative class. Although the metrics for the positive class remain lower than those for the negative class, they are significantly improved compared to the unbalanced dataset. This indicates that dataset balance plays a crucial role.\n",
        "\n",
        "The following table summarize the metrics for the positive class:\n",
        "\n",
        "|               | precision | recall | f1-score | ROC AUC |\n",
        "| --------------| --------- | ------ | -------- | ------- |\n",
        "| Random Forest | 0.39      | 0.74   | 0.51     | 0.92    |\n",
        "| XGBoost       | 0.28      | 0.78   | 0.41     | 0.91    |\n",
        "| NetMHC        | 0.33      | 0.80   | 0.47     | 0.93    |\n",
        "\n",
        "\n",
        "For Random Forest and XGBoost, precision is halved with the balanced dataset, but recall is significantly higher. Overall, the F1-score and ROC AUC are slightly improved with the balanced dataset. The top-performing models are either NetMHC or Random Forest, depending on the metric of interest.\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pRLrm3MJ5VW7"
      },
      "source": [
        "### Which dataset to choose ?\n",
        "\n",
        "Given the results, we recognize a precision-recall trade-off. If you want to focus on the **correctness** of positive predictions, then you should train your model on the raw dataset or a sample of it. If you want to focus on capturing **all relevant instances**, then you should train your models on a balanced dataset.\n",
        "\n",
        "### Which model to choose ?\n",
        "\n",
        "Overall, the model with the highest ROC AUC is the NetMHC model trained on the balanced dataset. Along with the Random Forest model trained on the balanced dataset, these are the best-performing models when considering all metrics.\n",
        "\n",
        "Nevertheless, the choice of model ultimately depends on the specific use case."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QDwiE8GdBrnU"
      },
      "source": [
        "### Further improvements\n",
        "\n",
        "\n",
        "\n",
        "1.   **Adressing class imbalance**:\n",
        "\n",
        "     Class imbalance significantly affects model performance. In this work, we address this issue by undersampling the dataset, although this approach results in data loss. We could oversample the dataset using synthetic data. However, this method must be used cautiously, especially since it is a biological application.\n",
        "\n",
        "2.  **Feature engineering:**\n",
        "    \n",
        "    The way we encoded the petpide sequence only preserve information about its spatial structure/sequence. There is other methods in the litterature such as BLOSUM matrix to add some information about amino acid similarity.\n",
        "    Regarding the HLA allele encoding, other methods than one hot encoding exists, such as mapping HLA to an amino acid pseudo sequence.\n",
        "\n",
        "3.  **Model**:\n",
        "\n",
        "    We focused only on 3 different models. We could invesigate further their hyperparameters tuning. Other NNs architecutre are used in the state-of-the-art such as RNNs or attention-based NNs (Kalemati et al., 2023) to leverage the sequence information contained in the data.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4IolH4aayJVe"
      },
      "source": [
        "## References\n",
        "- Reynisson, B., Alvarez, B., Paul, S., Peters, B., & Nielsen, M. (2020). NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data. In Nucleic Acids Research (Vol. 48, Issue W1, pp. W449–W454). Oxford University Press (OUP). https://doi.org/10.1093/nar/gkaa379\n",
        "\n",
        "- Luo, H., Ye, H., Ng, H. W., Shi, L., Tong, W., Mendrick, D. L., & Hong, H. (2015). Machine Learning Methods for Predicting HLA-Peptide Binding Activity. In Bioinformatics and Biology Insights (Vol. 9s3, p. BBI.S29466). SAGE Publications. https://doi.org/10.4137/bbi.s29466\n",
        "\n",
        "- Liu, Z., Cui, Y., Xiong, Z., Nasiri, A., Zhang, A., & Hu, J. (2019). DeepSeqPan, a novel deep convolutional neural network model for pan-specific class I HLA-peptide binding affinity prediction. In Scientific Reports (Vol. 9, Issue 1). Springer Science and Business Media LLC. https://doi.org/10.1038/s41598-018-37214-1\n",
        "\n",
        "- Han, Y., & Kim, D. (2017). Deep convolutional neural networks for pan-specific peptide-MHC class I binding prediction. In BMC Bioinformatics (Vol. 18, Issue 1). Springer Science and Business Media LLC. https://doi.org/10.1186/s12859-017-1997-x\n",
        "\n",
        "- Kalemati, M., Darvishi, S. & Koohi, S. CapsNet-MHC predicts peptide-MHC class I binding based on capsule neural networks. Commun Biol 6, 492 (2023). https://doi.org/10.1038/s42003-023-04867-2"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "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": "3.10.12"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}