4-Model_fitting.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-04T00:51:19.367987Z",
     "start_time": "2020-05-04T00:51:19.359990Z"
    }
   },
   "source": [
    "__GitHub repo__: https://git.io/JfsKA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We noted in the previous notebook (add link) that a number of datapoints had no signal (a.k.a. \"bad PCRs\"). After we exclude these, is the remaining data sufficient to predict drug classes?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-02T03:13:27.181406Z",
     "start_time": "2020-05-02T03:13:26.755409Z"
    }
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import pathlib\n",
    "df = pd.read_csv(pathlib.Path.cwd().joinpath('Clean_data_outputs','8-Clean_for_ML', 'HypoIII_for_ML_encoded.csv'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:05:39.278940Z",
     "start_time": "2020-04-22T16:05:33.586253Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import pathlib\n",
    "import pickle\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "from imblearn.over_sampling import SMOTE\n",
    "from plotnine import *\n",
    "from sklearn.ensemble import GradientBoostingClassifier, RandomForestClassifier\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.metrics import log_loss\n",
    "from sklearn.model_selection import LeaveOneGroupOut\n",
    "\n",
    "pd.set_option('mode.chained_assignment', None)\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:05:40.040578Z",
     "start_time": "2020-04-22T16:05:39.583579Z"
    }
   },
   "outputs": [],
   "source": [
    "df = pd.read_csv(pathlib.Path.cwd().joinpath('Clean_data_outputs','8-Clean_for_ML', 'HypoIII_for_ML_encoded.csv'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:05:40.977141Z",
     "start_time": "2020-04-22T16:05:40.885096Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Drug</th>\n",
       "      <th>Conc</th>\n",
       "      <th>Replicate</th>\n",
       "      <th>strain.0</th>\n",
       "      <th>strain.1</th>\n",
       "      <th>strain.10</th>\n",
       "      <th>strain.100</th>\n",
       "      <th>strain.101</th>\n",
       "      <th>strain.102</th>\n",
       "      <th>strain.103</th>\n",
       "      <th>...</th>\n",
       "      <th>strain.91</th>\n",
       "      <th>strain.92</th>\n",
       "      <th>strain.93</th>\n",
       "      <th>strain.94</th>\n",
       "      <th>strain.95</th>\n",
       "      <th>strain.96</th>\n",
       "      <th>strain.97</th>\n",
       "      <th>strain.98</th>\n",
       "      <th>strain.99</th>\n",
       "      <th>Class</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>drug_0</td>\n",
       "      <td>2xMIC</td>\n",
       "      <td>1</td>\n",
       "      <td>0.359612</td>\n",
       "      <td>-0.911401</td>\n",
       "      <td>1.554297</td>\n",
       "      <td>0.069421</td>\n",
       "      <td>-2.972960</td>\n",
       "      <td>1.185445</td>\n",
       "      <td>0.421721</td>\n",
       "      <td>...</td>\n",
       "      <td>0.176432</td>\n",
       "      <td>0.449538</td>\n",
       "      <td>0.131855</td>\n",
       "      <td>-0.794105</td>\n",
       "      <td>1.381503</td>\n",
       "      <td>-0.938048</td>\n",
       "      <td>-2.939168</td>\n",
       "      <td>1.492050</td>\n",
       "      <td>-0.387570</td>\n",
       "      <td>Class 4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>drug_0</td>\n",
       "      <td>2xMIC</td>\n",
       "      <td>2</td>\n",
       "      <td>0.922680</td>\n",
       "      <td>0.156397</td>\n",
       "      <td>1.587909</td>\n",
       "      <td>0.587814</td>\n",
       "      <td>0.183369</td>\n",
       "      <td>0.565838</td>\n",
       "      <td>1.133654</td>\n",
       "      <td>...</td>\n",
       "      <td>0.702362</td>\n",
       "      <td>0.783339</td>\n",
       "      <td>0.747126</td>\n",
       "      <td>0.113741</td>\n",
       "      <td>1.673887</td>\n",
       "      <td>-0.256643</td>\n",
       "      <td>-0.483839</td>\n",
       "      <td>1.684535</td>\n",
       "      <td>1.228144</td>\n",
       "      <td>Class 4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>drug_0</td>\n",
       "      <td>2xMIC</td>\n",
       "      <td>3</td>\n",
       "      <td>1.003276</td>\n",
       "      <td>0.171781</td>\n",
       "      <td>1.870643</td>\n",
       "      <td>0.381407</td>\n",
       "      <td>-0.678374</td>\n",
       "      <td>1.030591</td>\n",
       "      <td>1.469478</td>\n",
       "      <td>...</td>\n",
       "      <td>0.911449</td>\n",
       "      <td>0.944320</td>\n",
       "      <td>0.480790</td>\n",
       "      <td>-0.326351</td>\n",
       "      <td>1.691710</td>\n",
       "      <td>0.168192</td>\n",
       "      <td>-0.834700</td>\n",
       "      <td>1.609109</td>\n",
       "      <td>1.397204</td>\n",
       "      <td>Class 4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>drug_0</td>\n",
       "      <td>2xMIC</td>\n",
       "      <td>4</td>\n",
       "      <td>1.825450</td>\n",
       "      <td>0.914839</td>\n",
       "      <td>1.976140</td>\n",
       "      <td>1.693167</td>\n",
       "      <td>0.974058</td>\n",
       "      <td>2.142402</td>\n",
       "      <td>2.025873</td>\n",
       "      <td>...</td>\n",
       "      <td>1.264043</td>\n",
       "      <td>1.414149</td>\n",
       "      <td>1.483434</td>\n",
       "      <td>0.833600</td>\n",
       "      <td>2.156906</td>\n",
       "      <td>1.275261</td>\n",
       "      <td>1.326924</td>\n",
       "      <td>2.090184</td>\n",
       "      <td>1.211598</td>\n",
       "      <td>Class 4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>drug_0</td>\n",
       "      <td>2xMIC</td>\n",
       "      <td>5</td>\n",
       "      <td>1.862371</td>\n",
       "      <td>0.871132</td>\n",
       "      <td>1.937450</td>\n",
       "      <td>1.371979</td>\n",
       "      <td>0.936343</td>\n",
       "      <td>1.996058</td>\n",
       "      <td>2.160803</td>\n",
       "      <td>...</td>\n",
       "      <td>1.460437</td>\n",
       "      <td>1.147487</td>\n",
       "      <td>1.533885</td>\n",
       "      <td>1.618346</td>\n",
       "      <td>2.086125</td>\n",
       "      <td>0.782927</td>\n",
       "      <td>0.709278</td>\n",
       "      <td>2.169011</td>\n",
       "      <td>1.499092</td>\n",
       "      <td>Class 4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 429 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     Drug   Conc  Replicate  strain.0  strain.1  strain.10  strain.100  \\\n",
       "0  drug_0  2xMIC          1  0.359612 -0.911401   1.554297    0.069421   \n",
       "1  drug_0  2xMIC          2  0.922680  0.156397   1.587909    0.587814   \n",
       "2  drug_0  2xMIC          3  1.003276  0.171781   1.870643    0.381407   \n",
       "3  drug_0  2xMIC          4  1.825450  0.914839   1.976140    1.693167   \n",
       "4  drug_0  2xMIC          5  1.862371  0.871132   1.937450    1.371979   \n",
       "\n",
       "   strain.101  strain.102  strain.103  ...  strain.91  strain.92  strain.93  \\\n",
       "0   -2.972960    1.185445    0.421721  ...   0.176432   0.449538   0.131855   \n",
       "1    0.183369    0.565838    1.133654  ...   0.702362   0.783339   0.747126   \n",
       "2   -0.678374    1.030591    1.469478  ...   0.911449   0.944320   0.480790   \n",
       "3    0.974058    2.142402    2.025873  ...   1.264043   1.414149   1.483434   \n",
       "4    0.936343    1.996058    2.160803  ...   1.460437   1.147487   1.533885   \n",
       "\n",
       "   strain.94  strain.95  strain.96  strain.97  strain.98  strain.99    Class  \n",
       "0  -0.794105   1.381503  -0.938048  -2.939168   1.492050  -0.387570  Class 4  \n",
       "1   0.113741   1.673887  -0.256643  -0.483839   1.684535   1.228144  Class 4  \n",
       "2  -0.326351   1.691710   0.168192  -0.834700   1.609109   1.397204  Class 4  \n",
       "3   0.833600   2.156906   1.275261   1.326924   2.090184   1.211598  Class 4  \n",
       "4   1.618346   2.086125   0.782927   0.709278   2.169011   1.499092  Class 4  \n",
       "\n",
       "[5 rows x 429 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:05:41.672728Z",
     "start_time": "2020-04-22T16:05:41.598678Z"
    }
   },
   "outputs": [],
   "source": [
    "df = df[~df['Class'].isna()]\n",
    "df = df.sort_values(by=['Class', 'Drug']).reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:05:42.975113Z",
     "start_time": "2020-04-22T16:05:42.326244Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 400x400 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (141446632312)>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(ggplot(df, aes(x='Class', fill='Class'))\n",
    " +geom_bar()\n",
    " +theme_minimal()\n",
    " +theme(figure_size=(4,4), axis_text_x=element_text(angle=90), legend_position='none')\n",
    " +xlab('')\n",
    " +ylab('Count')\n",
    " +scale_fill_brewer(type='qual', palette='Set1'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Logistic Regression with LASSO"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Implement a logistic regression model with LASSO.\n",
    "- Split data into training and validation using LeaveOneGroupOut (LOGO). Each group is a datapoints from a single drug. This allows us to mimic a test situation where we obtain a single drug with an unknown mechanism of action  \n",
    "- Use LASSO regularization to select features\n",
    "- Use SMOTE (Synthetic Minority Over-sampling TEchnique) to correct for class imbalance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-20T16:06:43.804302Z",
     "start_time": "2020-04-20T16:06:43.798305Z"
    }
   },
   "source": [
    "### Select LASSO strength"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First we will iterate over a number of regularization strengths. Note that lower the value of C, greater is the regularization strength"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 442,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T15:36:40.282844Z",
     "start_time": "2020-04-22T15:36:40.257795Z"
    }
   },
   "outputs": [],
   "source": [
    "def logistic_reg_vanilla(df: pd.DataFrame, C: float) -> float:\n",
    "    \"\"\"\n",
    "    Fits a logistic regression model using leave out group out for validation and SMOTE for oversampling. Returns cross entropy loss\n",
    "    \"\"\"\n",
    "    X = df.filter(regex=('\\.')).values\n",
    "    y = df['Class'].values\n",
    "    groups = df['Drug'].values\n",
    "    n_classes = df['Class'].nunique()\n",
    "    predictions = np.full([len(y), n_classes], np.nan)\n",
    "    logo = LeaveOneGroupOut()\n",
    "    for train_ind, test_ind in logo.split(X, y, groups):\n",
    "        X_train, y_train= SMOTE().fit_sample(X[train_ind], y[train_ind])\n",
    "        X_test, y_test = X[test_ind], y[test_ind]   \n",
    "        model = LogisticRegression(penalty='l1', C=C, solver='liblinear', multi_class='ovr')\n",
    "        model.fit(X_train, y_train)\n",
    "        predictions[test_ind] = model.predict_proba(X_test)\n",
    "    ce = log_loss(y, predictions)\n",
    "    return ce\n",
    "        \n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 443,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T15:48:27.557865Z",
     "start_time": "2020-04-22T15:36:50.616557Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "C : 0.01, cross-entropy: 1.2845940758134553\n",
      "C : 0.1, cross-entropy: 0.8253421506289884\n",
      "C : 1, cross-entropy: 0.9628848417094924\n",
      "C : 10, cross-entropy: 1.0211248488802362\n",
      "C : 100, cross-entropy: 1.1537051581755318\n"
     ]
    }
   ],
   "source": [
    "output_metrics = {}\n",
    "for C in [0.01, 0.1, 1, 10, 100]:\n",
    "    output_metrics[C] = logistic_reg_vanilla(df, C)\n",
    "    print(f'C : {C}, cross-entropy: {output_metrics[C]}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "C = 0.1 seems to give the best results, hence we will use it for the next step"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fit model with selected lasso strength"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:05:44.623413Z",
     "start_time": "2020-04-22T16:05:44.585418Z"
    }
   },
   "outputs": [],
   "source": [
    "def logistic_reg(df: pd.DataFrame, C: float) -> (pd.DataFrame, pd.DataFrame):\n",
    "    \"\"\"\n",
    "    Adapts logisitic_reg_vanilla to output feature importances and predictions.\n",
    "    Predictions are returned as probability estimates for each class. \n",
    "    The importances are converted to asbolute values and averaged over each LOGO iteration.\n",
    "    \"\"\"\n",
    "    X = df.filter(regex=('\\.')).values\n",
    "    y = df['Class'].values\n",
    "    groups = df['Drug'].values\n",
    "    n_classes = df['Class'].nunique()\n",
    "    predictions = np.full([len(y), n_classes], np.nan)\n",
    "    logo = LeaveOneGroupOut()\n",
    "    feature_importances  = np.zeros([n_classes, X.shape[1]])\n",
    "    \n",
    "    for train_ind, test_ind in logo.split(X, y, groups):\n",
    "        X_train, y_train= SMOTE().fit_sample(X[train_ind], y[train_ind])\n",
    "        X_test, y_test = X[test_ind], y[test_ind]   \n",
    "        model = LogisticRegression(penalty='l1', C=C, solver='liblinear', multi_class='ovr')\n",
    "        model.fit(X_train, y_train)\n",
    "        predictions[test_ind] = model.predict_proba(X_test)\n",
    "        feature_importances += model.coef_\n",
    "        \n",
    "    #format predictions and feature importances before output\n",
    "    predictions = pd.DataFrame(predictions, columns= model.classes_)\n",
    "    predictions = predictions.merge(df[['Drug', 'Class']], left_index=True, right_index=True)\n",
    "    \n",
    "    feature_importances = np.abs(feature_importances.T)\n",
    "    feature_importances = feature_importances/len(groups)\n",
    "    feature_importances = pd.DataFrame(feature_importances, columns = model.classes_)\n",
    "    feature_importances['Strain'] = df.filter(regex=('\\.')).columns\n",
    "    feature_importances = feature_importances[['Strain']+list(model.classes_)]\n",
    "    return predictions, feature_importances\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:14:10.056793Z",
     "start_time": "2020-04-22T18:12:28.997190Z"
    }
   },
   "outputs": [],
   "source": [
    "predictions, feature_importances = logistic_reg(df, C=0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:15:58.203102Z",
     "start_time": "2020-04-22T18:15:58.173101Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Class 1</th>\n",
       "      <th>Class 2</th>\n",
       "      <th>Class 3</th>\n",
       "      <th>Class 4</th>\n",
       "      <th>Class 5</th>\n",
       "      <th>Class 6</th>\n",
       "      <th>Drug</th>\n",
       "      <th>Class</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.716816</td>\n",
       "      <td>0.020018</td>\n",
       "      <td>0.006671</td>\n",
       "      <td>0.009591</td>\n",
       "      <td>0.238210</td>\n",
       "      <td>0.008693</td>\n",
       "      <td>drug_12</td>\n",
       "      <td>Class 1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.598741</td>\n",
       "      <td>0.022168</td>\n",
       "      <td>0.004032</td>\n",
       "      <td>0.058308</td>\n",
       "      <td>0.269807</td>\n",
       "      <td>0.046944</td>\n",
       "      <td>drug_12</td>\n",
       "      <td>Class 1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.161857</td>\n",
       "      <td>0.025684</td>\n",
       "      <td>0.014003</td>\n",
       "      <td>0.056966</td>\n",
       "      <td>0.720838</td>\n",
       "      <td>0.020652</td>\n",
       "      <td>drug_12</td>\n",
       "      <td>Class 1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.743223</td>\n",
       "      <td>0.021980</td>\n",
       "      <td>0.002067</td>\n",
       "      <td>0.037902</td>\n",
       "      <td>0.186578</td>\n",
       "      <td>0.008251</td>\n",
       "      <td>drug_12</td>\n",
       "      <td>Class 1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.408251</td>\n",
       "      <td>0.017638</td>\n",
       "      <td>0.005930</td>\n",
       "      <td>0.101342</td>\n",
       "      <td>0.442526</td>\n",
       "      <td>0.024312</td>\n",
       "      <td>drug_12</td>\n",
       "      <td>Class 1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Class 1   Class 2   Class 3   Class 4   Class 5   Class 6     Drug  \\\n",
       "0  0.716816  0.020018  0.006671  0.009591  0.238210  0.008693  drug_12   \n",
       "1  0.598741  0.022168  0.004032  0.058308  0.269807  0.046944  drug_12   \n",
       "2  0.161857  0.025684  0.014003  0.056966  0.720838  0.020652  drug_12   \n",
       "3  0.743223  0.021980  0.002067  0.037902  0.186578  0.008251  drug_12   \n",
       "4  0.408251  0.017638  0.005930  0.101342  0.442526  0.024312  drug_12   \n",
       "\n",
       "     Class  \n",
       "0  Class 1  \n",
       "1  Class 1  \n",
       "2  Class 1  \n",
       "3  Class 1  \n",
       "4  Class 1  "
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:16:30.155914Z",
     "start_time": "2020-04-22T18:16:30.134921Z"
    }
   },
   "outputs": [],
   "source": [
    "def plot_specialized_confusion_matrix(predictions: pd.DataFrame) -> None:\n",
    "    \"\"\"\n",
    "    Takes in dataframe predictions and re-formats for plotting. Plots the final matrix\n",
    "    \"\"\"\n",
    "    #average predictions by class to create a confusion matrix\n",
    "    confusion = predictions.groupby('Class').mean().reset_index()\n",
    "    #format confusion for plotting\n",
    "    class_names = [col for col in confusion if col!='Class']\n",
    "    confusion = confusion.rename(columns={col: col[-1] for col in confusion[class_names]})\n",
    "    confusion['Class'] = confusion['Class'].replace({x: x[-1] for x in class_names})\n",
    "    confusion = confusion.set_index('Class')\n",
    "    \n",
    "    #plot\n",
    "    plt.figure(figsize=(5,5))\n",
    "    sns.set(font_scale=1.2)\n",
    "    heat = sns.heatmap(confusion, annot=True, fmt='1.2f', square=True, linewidths=.5, cbar=False)\n",
    "    plt.xlabel(\"Predicted Class\")\n",
    "    plt.ylabel(\"True Class\")\n",
    "    plt.show(heat)\n",
    "    return None\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:16:33.133850Z",
     "start_time": "2020-04-22T18:16:32.658850Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_specialized_confusion_matrix(predictions)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Class 1 is the one predicted with the best confidence (also has the most datapoints)\n",
    "- Class 2 is misclassified into Class 1\n",
    "- The other classes are predicted with ~50% confidence"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Find a minimal set of features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:20:01.843919Z",
     "start_time": "2020-04-22T16:20:01.812919Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Strain</th>\n",
       "      <th>Class 1</th>\n",
       "      <th>Class 2</th>\n",
       "      <th>Class 3</th>\n",
       "      <th>Class 4</th>\n",
       "      <th>Class 5</th>\n",
       "      <th>Class 6</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>strain.0</td>\n",
       "      <td>0.006386</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000041</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>strain.1</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>strain.10</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>strain.100</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>strain.101</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000027</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02210</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Strain   Class 1  Class 2   Class 3  Class 4  Class 5  Class 6\n",
       "0    strain.0  0.006386      0.0  0.000041      0.0      0.0  0.00000\n",
       "1    strain.1  0.000000      0.0  0.000000      0.0      0.0  0.00000\n",
       "2   strain.10  0.000000      0.0  0.000000      0.0      0.0  0.02011\n",
       "3  strain.100  0.000000      0.0  0.000000      0.0      0.0  0.00000\n",
       "4  strain.101  0.000000      0.0  0.000027      0.0      0.0  0.02210"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_importances.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since logistic regression performs one vs all classification, we have a different set of feature importances for each Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:20:04.221920Z",
     "start_time": "2020-04-22T16:20:04.174921Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>rank</th>\n",
       "      <th>Class</th>\n",
       "      <th>feature_importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Class 1</td>\n",
       "      <td>0.111136</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Class 1</td>\n",
       "      <td>0.091930</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>Class 1</td>\n",
       "      <td>0.075371</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>Class 1</td>\n",
       "      <td>0.070978</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>Class 1</td>\n",
       "      <td>0.068950</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   rank    Class  feature_importance\n",
       "0     0  Class 1            0.111136\n",
       "1     1  Class 1            0.091930\n",
       "2     2  Class 1            0.075371\n",
       "3     3  Class 1            0.070978\n",
       "4     4  Class 1            0.068950"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Create a df with the features for each class ranked and melt it to enable ggplotting\n",
    "feature_importances_ranked = pd.DataFrame({'rank': range(feature_importances.shape[0])})\n",
    "feature_importances_ranked[class_names] = feature_importances[class_names].apply(lambda col: col.sort_values(ascending=False).values, axis=0)\n",
    "feature_importances_ranked = feature_importances_ranked.melt(id_vars = 'rank', value_name='feature_importance', var_name='Class')\n",
    "feature_importances_ranked.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:30:56.020045Z",
     "start_time": "2020-04-22T16:30:55.573427Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (141448964956)>"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(ggplot(feature_importances_ranked, aes(x='rank', y='feature_importance', color='Class'))\n",
    "+geom_line(size=1.0)\n",
    "+theme_light()\n",
    "+theme(figure_size=(10,3))\n",
    "+scale_x_continuous(breaks=list(range(0,451, 50)))\n",
    "+xlab('Rank')\n",
    "+ylab('Feature Importance')\n",
    "+scale_color_brewer(type='qual', palette='Set1'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From the graph: If we extract the top 75 features for each class and merge them, we will obtain a set of required features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:20:16.375460Z",
     "start_time": "2020-04-22T16:20:16.365076Z"
    }
   },
   "outputs": [],
   "source": [
    "def extract_useful_features(col, n):\n",
    "    largest = col.nlargest(n)\n",
    "    largest = largest[largest!=0]\n",
    "    return largest.index.to_list()    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:20:17.723636Z",
     "start_time": "2020-04-22T16:20:17.679642Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of features in the minimum feature set: 227\n"
     ]
    }
   ],
   "source": [
    "feature_importances = feature_importances.set_index('Strain')\n",
    "minimal_features = list(set(feature_importances[class_names].apply(extract_useful_features, args=(75,)).sum()))\n",
    "print(f\"Number of features in the minimum feature set: {len(minimal_features)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Within the minimal set, what is the distribution of feature importances?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:20:25.119252Z",
     "start_time": "2020-04-22T16:20:25.098263Z"
    }
   },
   "outputs": [],
   "source": [
    "#create df with minimal_features and sum the importances across classes\n",
    "feature_importances = feature_importances.reset_index()\n",
    "feature_importances_filt = feature_importances[feature_importances['Strain'].isin(minimal_features)]\n",
    "feature_importances_filt['total_importance'] = feature_importances_filt[class_names].sum(axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:28:51.415226Z",
     "start_time": "2020-04-22T18:28:51.397230Z"
    }
   },
   "outputs": [],
   "source": [
    "def plot_importances(df:pd.DataFrame, strain_col:str, imp_col:str)->ggplot:\n",
    "    \"\"\"\n",
    "    Takes in a DataFrame with columns corresponding to strain and importances.\n",
    "    Displays a column plot with indicated x and y co-ordinates for the top 50 strains\n",
    "    \"\"\"\n",
    "    #make 'Strain' a categorical column for ggplotting\n",
    "    strains_sorted = df.sort_values(by=imp_col, ascending=False)[strain_col].values\n",
    "    df[strain_col] = pd.Categorical(df[strain_col], strains_sorted, ordered=True)\n",
    "    \n",
    "    return (ggplot(df.iloc[1:50], aes(x=strain_col, y=imp_col, fill=imp_col))\n",
    "    +geom_col()\n",
    "    +theme_light()\n",
    "    +theme(figure_size=(10,3), axis_text_x=element_text(angle=90), legend_position='none')\n",
    "    +ylab(\"Feature importance\") \n",
    "    +xlab(\"\")\n",
    "    +scale_fill_cmap(name='plasma')\n",
    "    )\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:28:54.420235Z",
     "start_time": "2020-04-22T18:28:53.190234Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (-9223371895405515411)>"
      ]
     },
     "execution_count": 90,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "plot_importances(feature_importances_filt, 'Strain', 'total_importance')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T16:27:10.075805Z",
     "start_time": "2020-04-22T16:27:10.062807Z"
    }
   },
   "source": [
    "- Strain.110 is disproportionately important!\n",
    "\n",
    "- This is in concordance with domain knowledge \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since logistic regression only models linear relationships, I decided to also fit the model to a Random Forest estimator that can use non-linear relationships between the data for prediction"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Select hyperparameters"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I decided to focus on two hyperparameters, n_estimators and max_depth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 452,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T15:51:11.221820Z",
     "start_time": "2020-04-22T15:51:11.202812Z"
    }
   },
   "outputs": [],
   "source": [
    "def random_forest_vanilla(df: pd.DataFrame, n_estimators: int, max_depth: int) -> int:\n",
    "    \"\"\"\n",
    "    Fits a random forest with the specified n_estimators and max_depth. \n",
    "    Uses LOGO for CV and SMOTE for correcting class imbalances. \n",
    "    Returns the cross-entropy loss.\n",
    "    \"\"\"\n",
    "    X = df.filter(regex=('\\.')).values\n",
    "    y = df['Class'].values\n",
    "    groups = df['Drug'].values\n",
    "    n_classes = df['Class'].nunique()\n",
    "    logo = LeaveOneGroupOut()\n",
    "    predictions = np.full([len(y), n_classes], np.nan)\n",
    "    for train_ind, test_ind in logo.split(X, y, groups):\n",
    "        X_train, y_train= SMOTE().fit_sample(X[train_ind], y[train_ind])\n",
    "        X_test, y_test = X[test_ind], y[test_ind]   \n",
    "        model = RandomForestClassifier(n_estimators=n_estimators, max_depth=max_depth)\n",
    "        model.fit(X_train, y_train)\n",
    "        predictions[test_ind] = model.predict_proba(X_test)    \n",
    "    ce = log_loss(y, predictions)\n",
    "    return ce\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 368,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-20T21:06:12.654870Z",
     "start_time": "2020-04-20T20:53:30.849135Z"
    },
    "deletable": false,
    "editable": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "n_estimators: 5, max_depth: 5, cross-entropy-loss: 1.7722560290498246\n",
      "n_estimators: 5, max_depth: 10, cross-entropy-loss: 3.498907229429203\n",
      "n_estimators: 5, max_depth: 20, cross-entropy-loss: 6.470737181711466\n",
      "n_estimators: 10, max_depth: 5, cross-entropy-loss: 1.5031186031691626\n",
      "n_estimators: 10, max_depth: 10, cross-entropy-loss: 1.8925465201445033\n",
      "n_estimators: 10, max_depth: 20, cross-entropy-loss: 3.324013160437168\n",
      "n_estimators: 25, max_depth: 5, cross-entropy-loss: 1.3997813348738954\n",
      "n_estimators: 25, max_depth: 10, cross-entropy-loss: 1.2908506512001783\n",
      "n_estimators: 25, max_depth: 20, cross-entropy-loss: 1.7138841040204775\n",
      "n_estimators: 50, max_depth: 5, cross-entropy-loss: 1.3830235292899293\n",
      "n_estimators: 50, max_depth: 10, cross-entropy-loss: 1.1874792738568833\n",
      "n_estimators: 50, max_depth: 20, cross-entropy-loss: 1.420956505898267\n"
     ]
    }
   ],
   "source": [
    "for n_estimators in [5, 10, 25, 50]:\n",
    "    for max_depth in [5, 10, 20]:\n",
    "        print (f\"n_estimators: {n_estimators}, max_depth: {max_depth}, cross-entropy-loss: {random_forest_vanilla(df, n_estimators, max_depth)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "n_estimators = 50 and max_depth = 10 gives the lowest loss"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fit model on selected hyperparameters "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:32:28.690125Z",
     "start_time": "2020-04-22T18:32:28.666124Z"
    }
   },
   "outputs": [],
   "source": [
    "def random_forest(df: pd.DataFrame, n_estimators: int, max_depth: int) -> (int, pd.DataFrame, pd.DataFrame):\n",
    "    \"\"\"\n",
    "    Adapts random_forest_vanilla to output cross entropy loss, predictions and feature importances. \n",
    "    The importances are averaged over each LOGO iteration.\n",
    "    \"\"\"\n",
    "    X = df.filter(regex=('\\.')).values\n",
    "    y = df['Class'].values\n",
    "    groups = df['Drug'].values\n",
    "    n_classes = df['Class'].nunique()\n",
    "    predictions = np.full([len(y), n_classes], np.nan)\n",
    "    logo = LeaveOneGroupOut()\n",
    "    feature_importances  = np.zeros(X.shape[1])\n",
    "    for train_ind, test_ind in logo.split(X, y, groups):\n",
    "        X_train, y_train= SMOTE().fit_sample(X[train_ind], y[train_ind])\n",
    "        X_test, y_test = X[test_ind], y[test_ind]   \n",
    "        model = RandomForestClassifier(n_estimators=n_estimators, max_depth=max_depth)\n",
    "        model.fit(X_train, y_train)\n",
    "        predictions[test_ind] = model.predict_proba(X_test)\n",
    "        feature_importances += model.feature_importances_\n",
    "        \n",
    "    #format predictions and feature importances before output\n",
    "    loss = log_loss(y, predictions)\n",
    "    predictions = pd.DataFrame(predictions, columns= model.classes_)\n",
    "    predictions = predictions.merge(df[['Drug', 'Class']], left_index=True, right_index=True)\n",
    "    feature_importances = feature_importances/len(groups)\n",
    "    feature_importances = pd.Series(feature_importances, index=df.filter(regex=('\\.')).columns)\n",
    "    return loss, predictions, feature_importances\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:35:28.624298Z",
     "start_time": "2020-04-22T18:32:30.005125Z"
    }
   },
   "outputs": [],
   "source": [
    "rf_log_loss, rf_predictions, rf_feature_importances = random_forest(df, n_estimators=50, max_depth=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:35:30.649508Z",
     "start_time": "2020-04-22T18:35:30.387506Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_confusion_matrix(rf_predictions)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- Compared to logistic regression, the Random Forest model performs worse. \n",
    "- There appears to be substantial bias towards Class 1 in spite of correcting class imbalancing with SMOTE. \n",
    "- Neither models can accurately classify Class 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Find important features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T18:36:13.145366Z",
     "start_time": "2020-04-22T18:36:12.305370Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (141448460008)>"
      ]
     },
     "execution_count": 101,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_feature_importances_df = pd.DataFrame(rf_feature_importances).reset_index()\n",
    "rf_feature_importances_df.columns=['Strain', 'importances']\n",
    "plot_importances(rf_feature_importances_df, 'Strain', 'importances')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Strain 110 is the most useful feature, similar to logistic regression. The second and subsequent most important features differ between logistic regression and random forest"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "While the distribution of feature importances informs us about which are the top n features, it's difficult to assess how big the subset of features selected for the next experiment should be. For example, should one use the top 20 features? The top 100?\n",
    "\n",
    "One way to answer this is by recursive feature elimination where you iterate through and fit datasets with an increasingly smaller number of features and estimate loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T19:03:53.150342Z",
     "start_time": "2020-04-22T19:03:53.141341Z"
    }
   },
   "outputs": [],
   "source": [
    "def RFE_rf(df: pd.DataFrame, n_estimators:int, max_depth:int)->(list, list):\n",
    "    \"\"\"\n",
    "    Takes in the chemical genomics dataframe and subjects it to recursive feature elimination by fitting the data \n",
    "    to a random forest with the specified n_estimators and max_depth. \n",
    "    Returns a list of the least important features and the cross entropy loss at each step\n",
    "    \"\"\"\n",
    "    features_to_remove=[]\n",
    "    log_losses = []\n",
    "    all_features = df.filter(regex='\\.').columns\n",
    "    while len(all_features)-len(features_to_remove)>1:\n",
    "        if len(features_to_remove)%50==0:\n",
    "            print (f\"{len(features_to_remove)} features removed\")\n",
    "        df_dropped = df.drop(columns=features_to_remove)\n",
    "        log_loss, _, feature_importances = random_forest(df_dropped, n_estimators, max_depth)\n",
    "        log_losses.append(log_loss)\n",
    "        features_to_remove.append(feature_importances.idxmin())\n",
    "    return features_to_remove[:-1], log_losses                         "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 404,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T09:45:32.894135Z",
     "start_time": "2020-04-21T21:58:49.374571Z"
    },
    "deletable": false,
    "editable": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 features removed\n",
      "50 features removed\n",
      "100 features removed\n",
      "150 features removed\n",
      "200 features removed\n",
      "250 features removed\n",
      "300 features removed\n",
      "350 features removed\n",
      "400 features removed\n"
     ]
    }
   ],
   "source": [
    "features_to_remove, log_losses = RFE_rf(df, n_estimators=50, max_depth=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 405,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T13:20:27.987004Z",
     "start_time": "2020-04-22T13:20:27.967997Z"
    },
    "deletable": false,
    "editable": false
   },
   "outputs": [],
   "source": [
    "#save files\n",
    "with open('Model_outputs/features_RFE.pkl', 'wb') as f:\n",
    "    pickle.dump(features_to_remove, f)\n",
    "with open('Model_outputs/log_loss_RFE.pkl', 'wb') as f:\n",
    "    pickle.dump(log_losses, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T19:05:38.527003Z",
     "start_time": "2020-04-22T19:05:38.516966Z"
    }
   },
   "outputs": [],
   "source": [
    "#read files\n",
    "with open('Model_outputs/features_RFE.pkl', 'rb') as f:\n",
    "    features_to_remove = pickle.load(f)\n",
    "with open('Model_outputs/log_loss_RFE.pkl', 'rb') as f:\n",
    "    log_losses = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T19:16:28.994284Z",
     "start_time": "2020-04-22T19:16:28.987592Z"
    }
   },
   "outputs": [],
   "source": [
    "loss_df = pd.DataFrame({'Number_of_features_removed': range(len(log_losses)), 'cross_entropy_loss': log_losses})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T19:12:15.344385Z",
     "start_time": "2020-04-22T19:12:14.851386Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1000x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (141452079708)>"
      ]
     },
     "execution_count": 112,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(ggplot(loss_df, aes(x='Number_of_features_removed', y='cross_entropy_loss', color='cross_entropy_loss'))\n",
    "+geom_point()\n",
    "+theme_light()\n",
    "+theme(figure_size=(10,3), legend_position='none')\n",
    "+scale_color_cmap(name='plasma_r')\n",
    "+scale_x_continuous(breaks=list(range(0,430,25)))\n",
    "+xlab(\"Number of features removed\")\n",
    "+ylab(\"Cross-entropy loss\")\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T21:36:54.512718Z",
     "start_time": "2020-04-22T21:36:54.504719Z"
    }
   },
   "source": [
    "- The loss decreases as features are removed, upto ~400 features, after which we see an increase in the loss.\n",
    "- A subset of the top ~25 features likely provides the best predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If we remove the lowest 400 features, what are the features that remain? How much of a difference in the loss is incurred by removing each?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T21:17:50.094819Z",
     "start_time": "2020-04-22T21:17:50.078821Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['strain.110', 'strain.416', 'strain.393', 'strain.425', 'strain.331', 'strain.433', 'strain.232', 'strain.428', 'strain.46', 'strain.63', 'strain.44', 'strain.17', 'strain.240', 'strain.61', 'strain.170', 'strain.424', 'strain.420', 'strain.23', 'strain.418', 'strain.236', 'strain.233', 'strain.244', 'strain.415', 'strain.303', 'strain.148']\n"
     ]
    }
   ],
   "source": [
    "all_features = df.filter(regex='\\.').columns\n",
    "minimal_features_rf = list(set(all_features) - set(features_to_remove[:400]))\n",
    "print(minimal_features_rf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-04-22T21:20:57.841313Z",
     "start_time": "2020-04-22T21:20:57.832307Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Strains identified as important by RF but not by LR: {'strain.425', 'strain.420', 'strain.148', 'strain.393'}\n"
     ]
    }
   ],
   "source": [
    "print(f\"Strains identified as important by RF but not by LR: {set(minimal_features_rf).difference(set(minimal_features))}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Overall conclusions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- We are able to predict drug classes based on strain sensitivity profiles\n",
    "- *Given a new drug, our pipeline can make predictions about its mechanism of action*\n",
    "\n",
    "\n",
    "- Logistic Regression works better than Random Forest\n",
    "- Prediction is strongest for data from class 1 (Accuracy: 70%), but lower for other classes\n",
    "- Class 2 has a very poor prediction accuracy. More data for drugs from this class would likely help improve predictions\n",
    "\n",
    "\n",
    "- Drug susceptibility of Strain.110 is a very strong predictor of drug class\n",
    "- The number of features required for accurate prediction might be between ~25-250. Since we have not covered all possible drug classes in this data, it would be prudent to keep ~250 strains in the next experiment, since they may be useful predictors for unseen drug classes"
   ]
  }
 ],
 "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": "3.7.3"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}