phenotype_prediction.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "# Phenotype Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from sup_src import utility\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns; sns.set()\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "**phenotype_prediction.ipynb** </br>\n",
    "In this notebook you are able to load our example data. You will then run one of our MLP experiments resulting in an average accuracy score across multiple seeds. Next you will run the data on the DA model and compare the two results on a box plot. In addition you are able to plot the class accuracy for both models, similar to figure 3D and supplementary figure 7 of the manuscript. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "First we load the training and test data. The data consists of gene expression values as described in the method section of the paper."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "df_sra_train = pd.read_csv('data/sra_train_tissue.csv', index_col='run_accession')\n",
    "df_sra_test = pd.read_csv('data/sra_test_tissue.csv', index_col='run_accession')\n",
    "df_gtex = pd.read_csv('data/gtex_tissue.csv', index_col='Unnamed: 0')\n",
    "df_tcga = pd.read_csv('data/tcga_tissue.csv', index_col='bigwig_file')\n",
    "df_bias_injection = pd.concat([df_sra_train,df_sra_test])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "# MLP\n",
    "- classes are labeled as strings and have to be converted to integer encoding\n",
    "- a tissue to integer mapping is returned (*tissue_mapping*)\n",
    "- data is randomized and extracted as floats\n",
    "- class weights are determined for model training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "df_train, df_test, tissue_mapping = utility.class_to_int(pd.concat([df_gtex, df_bias_injection]), df_tcga)\n",
    "\n",
    "train = df_train.sample(frac=1).values.astype('float32')\n",
    "test = df_test.sample(frac=1).values.astype('float32')\n",
    "class_weights = utility.get_class_weights(df_train)\n",
    "assert len(class_weights) == 16"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "- train and test data, as well as the tissue mapping and the class weighs are apped to the experiment\n",
    "- model type (e.g. mlp), the phenotype that is tested (e.g. tissue) has to be specified\n",
    "- experiments can be saved (save=True) and we need to specify the number of seeds that we want to run\n",
    "- the experiment outputs the sample and class accuracy for each run as well as their averages\n",
    "- *mlp_epochs=10* is the fault value for training time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.8307, 0.8602, 0.8555]\n",
      "[0.7669, 0.8285, 0.792]\n",
      "0.8488\n",
      "0.7958000000000002\n"
     ]
    }
   ],
   "source": [
    "utility.run_experiment(train, test, mapping=tissue_mapping, class_weights=class_weights, model_type='mlp', \n",
    "                       pheno_type='tissue', name='mlp_gtex_sra_tcga_tissue', save=True, seeds=3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "# DA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "For the DA model we first need to specify the configurations:\n",
    "<ul>\n",
    "    <li>source_layers: a list of integers representing the number of nodes to be used per layer for the source and bias mapper, [512] will create one layer with 512 nodes</li>\n",
    "    <li>classifier_layers: a list of integers representing the number of nodes to be used per layer for the classifier layer, [] will only create a single output layer with n=classes</li>\n",
    "    <li>lr: learning rate applied in the second training cycle\n",
    "    <li>classes: number of classes in the data</li>\n",
    "    <li>batch_size: number of samples per batch</li>\n",
    "    <li>margin: size of margin applied in triplet loss\n",
    "    <li>print: True / False, if test accuracy should be printed after every epoch during the second training cycle    \n",
    "</ul>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "config = {'mapper_layers': [512],\n",
    "          'classifier_layers': [],\n",
    "          'lr': 0.0005,\n",
    "          'classes': 16,\n",
    "          'batch_size': 64,\n",
    "          'margin': 5,\n",
    "          'print': True}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "- dataset preparation and running the experiment works as described above\n",
    "- in addition to *mlp_epochs* we can now also specify the *da_epochs*\n",
    "- running a DA model the experiment will output the sample accuracy for each *da_epochs*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "df_train, df_test, df_bias, tissue_mapping = utility.class_to_int(df_gtex, df_tcga, df_bias_injection,\n",
    "                                                                 da_mode=True)\n",
    "train = df_train.sample(frac=1).values.astype('float32')\n",
    "test = df_test.sample(frac=1).values.astype('float32')\n",
    "bias = df_bias.sample(frac=1).values.astype('float32')\n",
    "class_weights = utility.get_class_weights(df_train)\n",
    "assert len(class_weights) == 16"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.6407699443413729\n",
      "0.7809601113172542\n",
      "0.8188775510204082\n",
      "0.8494897959183674\n",
      "0.8577226345083488\n",
      "0.8693181818181818\n",
      "0.8679267161410018\n",
      "0.8649118738404453\n",
      "0.8711734693877551\n",
      "0.8688543599257885\n",
      "0.6665120593692022\n",
      "0.7912801484230055\n",
      "0.8278061224489796\n",
      "0.8580705009276438\n",
      "0.8554035250463822\n",
      "0.8741883116883117\n",
      "0.8646799628942486\n",
      "0.873956400742115\n",
      "0.877435064935065\n",
      "0.8755797773654916\n",
      "0.6865723562152134\n",
      "0.8144712430426716\n",
      "0.8470547309833024\n",
      "0.864448051948052\n",
      "0.8658395176252319\n",
      "0.8719851576994434\n",
      "0.8781307977736549\n",
      "0.875347866419295\n",
      "0.8828849721706865\n",
      "0.87847866419295\n",
      "[0.8689, 0.8756, 0.8785]\n",
      "[0.8057, 0.8194, 0.8128]\n",
      "0.8743333\n",
      "0.8126333333333333\n"
     ]
    }
   ],
   "source": [
    "utility.run_experiment(train, test, bias, tissue_mapping, class_weights, config, model_type='da', name='da_gtex_sra', seeds=3, mlp_epochs=10, da_epochs=10, save=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "Collapsed": "false"
   },
   "source": [
    "# Box Plot\n",
    "If the models have been saved, we can load them here and plot the results. </br>\n",
    "Add the names of the experiments you want to plot to the list below."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "Collapsed": "false"
   },
   "outputs": [],
   "source": [
    "plot_this = ['mlp_gtex_sra_tcga_tissue','da_gtex_sra']\n",
    "\n",
    "res_sample = {}\n",
    "res_tissue = {}\n",
    "res_tcga = pd.DataFrame()\n",
    "\n",
    "for each in plot_this:\n",
    "    sample = []\n",
    "    tissue = []\n",
    "    sample_, tissue_, _ = utility.load_and_predict(each, test, tissue_mapping, 'models/', seeds=3, classes=16)\n",
    "    sample.append(sample_)\n",
    "    tissue.append(tissue_)\n",
    "        \n",
    "    res_sample[each] = sample[0]\n",
    "    res_tissue[each] = tissue[0]\n",
    "    \n",
    "    for i in range(len(sample[0])):\n",
    "        tmp_dic = {'accuracy': float(sample[0][i]),\n",
    "                  'type': 'sample',\n",
    "                  'model': each}\n",
    "        res_tcga = res_tcga.append(tmp_dic, ignore_index=True)\n",
    "        \n",
    "        tmp_dic = {'accuracy': float(tissue[0][i]),\n",
    "          'type': 'class',\n",
    "          'model': each}\n",
    "        res_tcga = res_tcga.append(tmp_dic, ignore_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax = sns.boxplot(x='model', y='accuracy', hue='type', data=res_tcga)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Per Class Accuracy\n",
    "Per class accuracy allows us to see where a model performed better.</br>\n",
    "Rename the models to be loaded accordingly. Don't forget to change the number of seeds you have run."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "list_models = []\n",
    "seeds = 3\n",
    "for i in range(seeds):\n",
    "    list_models.append(tf.keras.models.load_model('models/mlp_gtex_sra_tcga_tissue_'+str(i)+'.h5', \n",
    "                                               compile=False))\n",
    "\n",
    "res_mlp = utility.mlp_report_per_class(test, list_models[0], 16, tissue_mapping)\n",
    "res_mlp['run'] = 0\n",
    "\n",
    "for i in range(1,seeds):\n",
    "    df_tmp = utility.mlp_report_per_class(test, list_models[i], 16, tissue_mapping)\n",
    "    df_tmp['run'] = i\n",
    "    res_mlp = pd.concat([res_mlp, df_tmp])\n",
    "\n",
    "res_mlp['data'] = 'MLP G+S-T'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "list_models = []\n",
    "for i in range(seeds):\n",
    "    list_models.append(tf.keras.models.load_model('models/da_gtex_sra_'+str(i)+'.h5', \n",
    "                                               compile=False))\n",
    "\n",
    "res_da = utility.mlp_report_per_class(test, list_models[0], 16, tissue_mapping)\n",
    "res_da['run'] = 0\n",
    "\n",
    "for i in range(1,seeds):\n",
    "    df_tmp = utility.mlp_report_per_class(test, list_models[i], 16, tissue_mapping)\n",
    "    df_tmp['run'] = i\n",
    "    res_da = pd.concat([res_da, df_tmp])\n",
    "\n",
    "res_da['data'] = 'DA G+S-T'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "res = pd.concat([res_mlp, res_da])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Text(0, 0, 'OVARY'),\n",
       " Text(0, 0, 'BRAIN'),\n",
       " Text(0, 0, 'ESOPHAGUS'),\n",
       " Text(0, 0, 'PANCREAS'),\n",
       " Text(0, 0, 'THYROID'),\n",
       " Text(0, 0, 'STOMACH'),\n",
       " Text(0, 0, 'BREAST'),\n",
       " Text(0, 0, 'SKIN'),\n",
       " Text(0, 0, 'LUNG'),\n",
       " Text(0, 0, 'TESTIS'),\n",
       " Text(0, 0, 'BONE MARROW'),\n",
       " Text(0, 0, 'ADRENAL GLAND'),\n",
       " Text(0, 0, 'UTERUS'),\n",
       " Text(0, 0, 'PROSTATE'),\n",
       " Text(0, 0, 'LIVER'),\n",
       " Text(0, 0, 'KIDNEY')]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax = sns.barplot(x='tissue', y='accuracy', hue='data', data=res)\n",
    "ax.set_xticklabels(ax.get_xticklabels(), rotation=90)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}