retrospective_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Table of Contents\n",
    "1. [Setup](#Setup)\n",
    "2. [Preprocessing](#Preprocessing)\n",
    "3. [ML Pipeline](#ML-Pipeline)\n",
    "    1. [Train-Test Split](#Train-Test-Split)\n",
    "    2. [Random Forest](#Train-Random-Forest)\n",
    "    3. [Deep Learning](#Train-Deep-Learning)\n",
    "    4. [Ensemble Model](#Ensemble-Model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Setup\n",
    "[Back to Top](#Table-of-Contents)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import sklearn\n",
    "import random\n",
    "from tqdm import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# ROOT_DIR - root directory\n",
    "ROOT_DIR = os.getcwd()+'/'\n",
    "\n",
    "# FEATURE_DIR - directory where feature dataframes are saved\n",
    "DATA_DIR = ROOT_DIR + 'dataframes/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of prESTs: 22603\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>prest_id</th>\n",
       "      <th>uniprot_id</th>\n",
       "      <th>conc_cf</th>\n",
       "      <th>aa_seq</th>\n",
       "      <th>nt_seq</th>\n",
       "      <th>aa_len</th>\n",
       "      <th>true_nt_seq</th>\n",
       "      <th>AAA</th>\n",
       "      <th>AAT</th>\n",
       "      <th>AAC</th>\n",
       "      <th>...</th>\n",
       "      <th>ronn_avg</th>\n",
       "      <th>ronn_results</th>\n",
       "      <th>ronn_frac</th>\n",
       "      <th>disopred_results</th>\n",
       "      <th>disopred_avg</th>\n",
       "      <th>disopred_pb_results</th>\n",
       "      <th>disopred_pb_avg</th>\n",
       "      <th>disopred_frac</th>\n",
       "      <th>disopred_pb_frac</th>\n",
       "      <th>expressed</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>140099</td>\n",
       "      <td>G3V537</td>\n",
       "      <td>2.91540</td>\n",
       "      <td>TYYAWKHELLGSGTCPALPPREVLGMEELEKLPEEQVAEEELECSA...</td>\n",
       "      <td>GACAAGCTTGCGGCCGCAACCTACTATGCCTGGAAGCATGAGCTGC...</td>\n",
       "      <td>144</td>\n",
       "      <td>ACCTACTATGCCTGGAAGCATGAGCTGCTGGGCTCTGGCACCTGCC...</td>\n",
       "      <td>0.006944</td>\n",
       "      <td>0.013889</td>\n",
       "      <td>0.013889</td>\n",
       "      <td>...</td>\n",
       "      <td>0.488125</td>\n",
       "      <td>tyyawkhellGSGTcPALPPREVLGMEELEKLPEEQVAEEELECSa...</td>\n",
       "      <td>0.493056</td>\n",
       "      <td>tyyawkhellgsgtCPALPPREVLGMEELEKLPEEQVAEEELECSA...</td>\n",
       "      <td>0.421597</td>\n",
       "      <td>..............^^^^^^--^^^^^^^^^^^^^^^^^^^-^^^^...</td>\n",
       "      <td>0.352917</td>\n",
       "      <td>0.465278</td>\n",
       "      <td>0.423611</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>140225</td>\n",
       "      <td>P12724</td>\n",
       "      <td>1.48770</td>\n",
       "      <td>SLHARPPQFTRAQWFAIQHISLNPPRCTIAMRAINNYRWRCKNQNT...</td>\n",
       "      <td>GACAAGCTTGCGGCCGCATCACTCCATGCCAGACCCCCACAGTTTA...</td>\n",
       "      <td>136</td>\n",
       "      <td>TCACTCCATGCCAGACCCCCACAGTTTACGAGGGCTCAGTGGTTTG...</td>\n",
       "      <td>0.007353</td>\n",
       "      <td>0.051471</td>\n",
       "      <td>0.058824</td>\n",
       "      <td>...</td>\n",
       "      <td>0.368750</td>\n",
       "      <td>SLharppqftraqwfaiqhislnpprctiamrainnyrwrcknqnt...</td>\n",
       "      <td>0.058824</td>\n",
       "      <td>slharppqftraqwfaiqhislnpprctiamrainnyrwrcknqnt...</td>\n",
       "      <td>0.034338</td>\n",
       "      <td>.................................................</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>140325</td>\n",
       "      <td>O43506</td>\n",
       "      <td>1.50290</td>\n",
       "      <td>FVGWWTHQRFVELVVVVDNIRYLFSQSNATTVQHEVFNVVNIVDSF...</td>\n",
       "      <td>GACAAGCTTGCGGCCGCATTTGTGGGCTGGTGGACCCATCAGCGGT...</td>\n",
       "      <td>137</td>\n",
       "      <td>TTTGTGGGCTGGTGGACCCATCAGCGGTTTGTTGAGCTGGTAGTGG...</td>\n",
       "      <td>0.014599</td>\n",
       "      <td>0.072993</td>\n",
       "      <td>0.021898</td>\n",
       "      <td>...</td>\n",
       "      <td>0.310584</td>\n",
       "      <td>fvgwwthqrfvelvvvvdnirylfsqsnattvqhevfnvvnivdsf...</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>fvgwwthqrfvelvvvvdnirylfsqsnattvqhevfnvvnivdsf...</td>\n",
       "      <td>0.015912</td>\n",
       "      <td>.................................................</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>140354</td>\n",
       "      <td>O95072</td>\n",
       "      <td>0.92003</td>\n",
       "      <td>RRRLLFWDKETQISPEKFQEQLQTRAHCWECPMVQPPERTIRGPAE...</td>\n",
       "      <td>GACAAGCTTGCGGCCGCACGTCGCCGGTTACTGTTCTGGGACAAGG...</td>\n",
       "      <td>134</td>\n",
       "      <td>CGTCGCCGGTTACTGTTCTGGGACAAGGAGACTCAGATCTCCCCGG...</td>\n",
       "      <td>0.014925</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.540448</td>\n",
       "      <td>rrrllfWDKETQISPEKFQEQLQTRAHCWEcPmvQPPERTIRGPae...</td>\n",
       "      <td>0.671642</td>\n",
       "      <td>Rrrllfwdketqispekfqeqlqtrahcwecpmvqppertirgpae...</td>\n",
       "      <td>0.271269</td>\n",
       "      <td>-................................................</td>\n",
       "      <td>0.191791</td>\n",
       "      <td>0.208955</td>\n",
       "      <td>0.201493</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>140500</td>\n",
       "      <td>C9J919</td>\n",
       "      <td>2.21830</td>\n",
       "      <td>MQQPPPQKVRINLQQQPPPLQIKSVPLPTLKMQTTLVPPTVESSPE...</td>\n",
       "      <td>GACAAGCTTGCGGCCGCAATGCAACAGCCTCCACCTCAGAAAGTTC...</td>\n",
       "      <td>140</td>\n",
       "      <td>ATGCAACAGCCTCCACCTCAGAAAGTTCGAATCAATTTACAGCAAC...</td>\n",
       "      <td>0.014286</td>\n",
       "      <td>0.021429</td>\n",
       "      <td>0.021429</td>\n",
       "      <td>...</td>\n",
       "      <td>0.587714</td>\n",
       "      <td>MQQPPPQKVRINLQQQPPPLQIKSVPLPTLKMQTTLVPPTVESSPE...</td>\n",
       "      <td>0.771429</td>\n",
       "      <td>MQQPPPQKVRINLQQQPpplQIKsvplPtLKMQTTLVPPTVESSPE...</td>\n",
       "      <td>0.489143</td>\n",
       "      <td>^^^^^^^^^^^^^^^^^...^^^....-.----^-------^^^^^...</td>\n",
       "      <td>0.453500</td>\n",
       "      <td>0.585714</td>\n",
       "      <td>0.471429</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 164 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    prest_id uniprot_id  conc_cf  \\\n",
       "1     140099     G3V537  2.91540   \n",
       "2     140225     P12724  1.48770   \n",
       "5     140325     O43506  1.50290   \n",
       "8     140354     O95072  0.92003   \n",
       "10    140500     C9J919  2.21830   \n",
       "\n",
       "                                               aa_seq  \\\n",
       "1   TYYAWKHELLGSGTCPALPPREVLGMEELEKLPEEQVAEEELECSA...   \n",
       "2   SLHARPPQFTRAQWFAIQHISLNPPRCTIAMRAINNYRWRCKNQNT...   \n",
       "5   FVGWWTHQRFVELVVVVDNIRYLFSQSNATTVQHEVFNVVNIVDSF...   \n",
       "8   RRRLLFWDKETQISPEKFQEQLQTRAHCWECPMVQPPERTIRGPAE...   \n",
       "10  MQQPPPQKVRINLQQQPPPLQIKSVPLPTLKMQTTLVPPTVESSPE...   \n",
       "\n",
       "                                               nt_seq  aa_len  \\\n",
       "1   GACAAGCTTGCGGCCGCAACCTACTATGCCTGGAAGCATGAGCTGC...     144   \n",
       "2   GACAAGCTTGCGGCCGCATCACTCCATGCCAGACCCCCACAGTTTA...     136   \n",
       "5   GACAAGCTTGCGGCCGCATTTGTGGGCTGGTGGACCCATCAGCGGT...     137   \n",
       "8   GACAAGCTTGCGGCCGCACGTCGCCGGTTACTGTTCTGGGACAAGG...     134   \n",
       "10  GACAAGCTTGCGGCCGCAATGCAACAGCCTCCACCTCAGAAAGTTC...     140   \n",
       "\n",
       "                                          true_nt_seq       AAA       AAT  \\\n",
       "1   ACCTACTATGCCTGGAAGCATGAGCTGCTGGGCTCTGGCACCTGCC...  0.006944  0.013889   \n",
       "2   TCACTCCATGCCAGACCCCCACAGTTTACGAGGGCTCAGTGGTTTG...  0.007353  0.051471   \n",
       "5   TTTGTGGGCTGGTGGACCCATCAGCGGTTTGTTGAGCTGGTAGTGG...  0.014599  0.072993   \n",
       "8   CGTCGCCGGTTACTGTTCTGGGACAAGGAGACTCAGATCTCCCCGG...  0.014925  0.000000   \n",
       "10  ATGCAACAGCCTCCACCTCAGAAAGTTCGAATCAATTTACAGCAAC...  0.014286  0.021429   \n",
       "\n",
       "         AAC    ...      ronn_avg  \\\n",
       "1   0.013889    ...      0.488125   \n",
       "2   0.058824    ...      0.368750   \n",
       "5   0.021898    ...      0.310584   \n",
       "8   0.000000    ...      0.540448   \n",
       "10  0.021429    ...      0.587714   \n",
       "\n",
       "                                         ronn_results  ronn_frac  \\\n",
       "1   tyyawkhellGSGTcPALPPREVLGMEELEKLPEEQVAEEELECSa...   0.493056   \n",
       "2   SLharppqftraqwfaiqhislnpprctiamrainnyrwrcknqnt...   0.058824   \n",
       "5   fvgwwthqrfvelvvvvdnirylfsqsnattvqhevfnvvnivdsf...   0.000000   \n",
       "8   rrrllfWDKETQISPEKFQEQLQTRAHCWEcPmvQPPERTIRGPae...   0.671642   \n",
       "10  MQQPPPQKVRINLQQQPPPLQIKSVPLPTLKMQTTLVPPTVESSPE...   0.771429   \n",
       "\n",
       "                                     disopred_results  disopred_avg  \\\n",
       "1   tyyawkhellgsgtCPALPPREVLGMEELEKLPEEQVAEEELECSA...      0.421597   \n",
       "2   slharppqftraqwfaiqhislnpprctiamrainnyrwrcknqnt...      0.034338   \n",
       "5   fvgwwthqrfvelvvvvdnirylfsqsnattvqhevfnvvnivdsf...      0.015912   \n",
       "8   Rrrllfwdketqispekfqeqlqtrahcwecpmvqppertirgpae...      0.271269   \n",
       "10  MQQPPPQKVRINLQQQPpplQIKsvplPtLKMQTTLVPPTVESSPE...      0.489143   \n",
       "\n",
       "                                  disopred_pb_results  disopred_pb_avg  \\\n",
       "1   ..............^^^^^^--^^^^^^^^^^^^^^^^^^^-^^^^...         0.352917   \n",
       "2   .................................................         0.000000   \n",
       "5   .................................................         0.000000   \n",
       "8   -................................................         0.191791   \n",
       "10  ^^^^^^^^^^^^^^^^^...^^^....-.----^-------^^^^^...         0.453500   \n",
       "\n",
       "    disopred_frac  disopred_pb_frac  expressed  \n",
       "1        0.465278          0.423611      False  \n",
       "2        0.000000          0.000000      False  \n",
       "5        0.000000          0.000000      False  \n",
       "8        0.208955          0.201493      False  \n",
       "10       0.585714          0.471429      False  \n",
       "\n",
       "[5 rows x 164 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "DF_prest = pd.read_csv(DATA_DIR+'DF_prest_filtered.csv',index_col=0)\n",
    "print 'Number of prESTs:', len(DF_prest)\n",
    "DF_prest.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Preprocessing\n",
    "[Back to Top](#Table-of-Contents)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There are very few highly expressed proteins when there are >4 prESTs from the same uniprot, as these were peptides that were difficult to express"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 12851/12851 [00:12<00:00, 1065.91it/s]\n"
     ]
    }
   ],
   "source": [
    "df_counts = pd.DataFrame(data=0,columns=['high','low'],index=set(DF_prest.groupby('uniprot_id').count().aa_seq))\n",
    "\n",
    "for name, group in tqdm(DF_prest.groupby('uniprot_id')):\n",
    "    n = len(group.conc_cf)\n",
    "    df_counts.loc[n,'high'] += sum(group.expressed==True)\n",
    "    df_counts.loc[n,'low'] += sum(group.expressed==False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f5c31310150>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqgAAAHWCAYAAAC7eHtwAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAAPYQAAD2EBqD+naQAAIABJREFUeJzs3Xt4VOW99vF7JiEBAjNkPCCmgKBBciIIQYiAQcsGWg0K\nCiZutNBiQY0aRFo8BBCBqiB6VUFB39qNtVVOFVDEWlHKRqi+reYwxKLBDVSIJWQmwymEZNb7By+z\nnY4ckgyZJ+H7uS4vyTxrrfmtmxFuV1ZmbJZlWQIAAAAMYY/0AAAAAMB3UVABAABgFAoqAAAAjEJB\nBQAAgFEoqAAAADAKBRUAAABGoaACAADAKBRUAAAAGIWCCgAAAKNQUAEAAGCURhXUefPmqWfPnpKk\nTz75RD179lR6errS09PVq1cvpaen67333gtsv2zZMo0YMUL9+vXTuHHj5Ha7A2s1NTWaMWOGsrKy\ndM011+iBBx6Q1+ttzHgAAABohhpcUEtLS7VmzRrZbLbAYwkJCSosLFRhYaGKiopUWFio4cOHS5I2\nbtyoRYsWaf78+dqyZYuysrI0adIkVVdXS5IWLlyo0tJSLV++XBs2bJBlWXr44YcbeXoAAABobhpU\nUC3L0qxZs/TTn/70rPdZvny5Ro8erbS0NMXExGjixImy2WzauHGj/H6/Vq1apXvvvVcdO3aUw+FQ\nfn6+PvroI+3fv78hIwIAAKCZalBB/cMf/qDY2FjdeOONQY8fOnRIeXl5GjBggLKysvTb3/42sFZS\nUqLk5OTA1zabTUlJSSouLtauXbt08OBBJSUlBda7d++u1q1bB90GAAAAgJYvur47VFRU6IUXXtDv\nfve7oMfbtWunK6+8UuPHj9dzzz2nv/71r3rggQfkcDg0evRoeb1eORyOoH2cTqe8Xq+8Xq9sNpuc\nTmfQusPhkMfjacBpAQAAoLmq9xXUJ598Urfeequ6d+8e9HhycrKWLVumjIwMRUdHa+DAgcrJydHq\n1atPeSzLsk77XJZlBd3jeiZnOh4AAADMV68rqFu3btVnn32mOXPmSDpzIUxISNCf/vQnSZLL5Qq5\nGlpVVaUePXrI5XLJsix5PB516tQpsO7z+RQfH3/W81VWHpbdfvaF9lyIirLL4Wgjn++o6ur8EZ3F\nFGQSikxCkUkoMglFJqHIJBSZBDMtj/j4uDNuU6+CunbtWlVWVmrIkCGSThRUy7KUmZmp6dOn68iR\nI8rNzQ1sX1ZWps6dO0uSUlNT5Xa7dfPNN0uS/H6/tm/frrFjx6pz585yOp1yu92Bgrpjxw4dP35c\naWlpZz2f32/J7zfjKmpdnV+1tZF/EZiETEKRSSgyCUUmocgkFJmEIpNgzSmPen2L/5FHHtGGDRu0\nZs0arVmzRkuXLpUkrVmzRu3atdPTTz+tjz/+WHV1ddqyZYtWr14dKKy5ublas2aNCgsLVV1drcWL\nFys2NlZZWVmy2+0aO3asXnzxRZWXl8vj8WjhwoUaNmyYXC5X+M8aAAAAxqrXFdT27durffv2ga9r\na2tls9l08cUX64c//KEeeeQRzZ49W+Xl5brwwgv12GOPaejQoZKkwYMH68EHH1R+fr4qKyuVlpam\npUuXKiYmRpJ0//3368iRI7rppptUV1en6667TjNnzgzjqQIAAKA5sFkt6CeL9u8/GOkRFB1tV3x8\nnDyew83mMvq5RiahyCQUmYQik1BkEopMQpFJMNPyuOii9mfcplEfdQoAAACEGwUVAAAARqGgAgAA\nwCgUVAAAABiFggoAAACjUFABAABgFAoqAAAAjFKvN+oHAACAVFNTI7e7uEmfMyUlLfABRy0dBRUA\nAKCe3O5i/WLharW/oEuTPN/BA7v19IPSVVf1Pet9Bg/up2eeeV7XXHNNyNp7763XK68s0YoVa854\nnHfffVsvvfS81qx5r14zNwYFFQAAoAHaX9BFHS5JjPQYDTJ8+I81fPiP67GH7ZzN8n24BxUAAABG\noaACAAC0UBUV+5WXN1np6enKzb1VX3+9U9KJb9vfdNPwwHYff/zfuuWWGzVsWJaefPIJ/Z//s0T3\n3Tcp6Fh/+ctHGjv2Jl1//UDNmTNTdXV152xuCioAAEALtXbtHzV9+qPaunWrLrjgAi1duug7qye+\nbV9RUaHHHvulcnPHaf36D5SWlq5Vq5bLZvvfb+sfOXJYxcWFeu21N7VkyW/0wQd/0pYtm8/Z3BRU\nAACAFmrEiBv0gx90Vtu2bTVo0LXavXtXyDZ///unatu2jW655TZFR0frhhtGqkuXrkHb1NTUaOLE\nSYqNba3ExCvVtWs37d79P+dsbgoqAABAC9Wp06WBX8fGxqqm5njINhUVFbr44o5BV0yTklKCtunQ\noYNiY1sHHev48dBjhQsFFQAAoIX6buk8NUtRUcFv7GS3//t+/BQ/AAAAmkh8vEv/+te3QY+Vlm6P\n0DQn8D6oAAAADXDwwO4mfq6Mc3Ls3r37yOv1aM2a1brhhpF67731+uc/9+iyy7qdk+c7GxRUnFY4\nPsotKsquwYMHhGkiAAAiLyUlTU8/2JTPmKGUlLR67XF2394/cZ/q9OkFevHF57VkySKNGHHiTfx3\n7PhHo4/dUDbLsqxz+gxNaP/+g5EeQdHRdsXHx8njOazaWn+kx2m0zz77W6M/yu3ggd16+Yk7dMUV\nyS0ik3Boaa+TcCCTUGQSikxCkUkoMgl2NnnU1tYqOvp/r1vOnTtLlmXpscceD/s8F13U/ozbcAUV\nZ9ScP8oNAACc3tGjRzVq1I/085/fq1GjbtWOHf/Q5s0fadq0RyM2EwUVAADgPNamTRs98cRTevHF\nX+vFF59Xhw7x+s///Il++MP/iNhMFFQAAIDzXL9+/dWv3+uRHiOAt5kCAACAUSioAAAAMAoFFQAA\nAEahoAIAAMAoFFQAAAAYhZ/iBwAAqKdwfNJifaWkpCkmJuasty8v36cxY0bqzTdXKz4++RxOFn4U\nVAAAgHpyu4v12B8fl6Ozq0mez7enUnM0U1dd1bde+53rjyQ9VyioAAAADeDo7JLr8osjPcZpNddP\ntOceVAAAgBbO5/Pp8ccLdNNNIzRsWJZ+8Yt8lZeX69ixal1//TX6+uudgW1Hj75BM2c+Evj67bfX\n6Gc/u6NJ56WgAgAAtFAnv8X/yCOPqLLygF577U299da7io1trRkzpis2trWSk1NVXFwoSdq3b69i\nY2OD7q8tKvpcGRlXN+ncFFQAAIAWrKrKqw8++ECTJ+fJ4XCqbdu2+tnPJumLL7arvHyfrrqqr0pK\niiRJn3/+d/Xu3Vdt2rRRefk+SVJh4WcUVAAAAITPyauoXbteFnjsBz/oLMuytG/fXvXt2y9wBbWw\n8DP16pWulJQ0FRV9roqKClVU7FevXr2bdGZ+SAoAAOA88H0/0W+z2ZSSkqaKiv3yer36/PPPdMcd\nE1RXV6vCws8UFRWt1NReio2NbdJZG3UFdd68eerZs2fg661bt2rMmDHq27evsrOztW7duqDtly1b\nphEjRqhfv34aN26c3G53YK2mpkYzZsxQVlaWrrnmGj3wwAPyer2NGQ8AAOC8V1tbJ8uy9D//83Xg\nsf/5n69ls9n0gx90VqtWrZSa2ksfffSBqquPKiHhB0pNTVdxcaGKiz9X3779mnzmBhfU0tJSrVmz\nJtDG//Wvf+mee+7R7bffrq1bt+qRRx5RQUFBoIRu3LhRixYt0vz587VlyxZlZWVp0qRJqq6uliQt\nXLhQpaWlWr58uTZs2CDLsvTwww+H4RQBAADOT5ZlyeVyadCgQVqyZLF8Pp98Pp9efnmx+vTJ0IUX\nXiRJuuqqvlqx4g9KTz/xrfzLLuumysoD+r//99Mmv/9UauC3+C3L0qxZs/TTn/5Uzz33nCRp3bp1\n6tatm0aNGiVJyszM1PXXX68VK1YoJSVFy5cv1+jRo5WWliZJmjhxopYtW6aNGzdqxIgRWrVqlebP\nn6+OHTtKkvLz83XDDTdo//79uuiii8JxrgAAAGHj21PZtM+VUf/9Tl5IfPrpp/XoowX6z/+8VXa7\nXf369dejj04NbNenTz+98spLGj16bOCxEz/dX6SePZv+U6gaVFD/8Ic/KDY2VjfeeGOgoG7fvl0p\nKSlB2yUnJ+vdd9+VJJWUlOiGG24IrNlsNiUlJam4uFhJSUk6ePCgkpKSAuvdu3dX69at5Xa7NWTI\nkIaMCUP462q1fft2+XxHVVfnb/Bx6vsRbwAAnCspKWmao5lN94QZJ56zPi65pJP+8pdPFB1tV3x8\nnObNe1q1td//93Bqapr+8pdPgh57+unnGjxuY9W7oFZUVOiFF17Q7373u6DHvV6vLrnkkqDHnE6n\nPB5PYN3hcISse71eeb1e2Ww2OZ3OoHWHwxHYH83XYe8+Ldr2XqM+Dq6hH/EGAMC5EBMTw99J51C9\nC+qTTz6pW2+9Vd27d9c333xzxu1P9xmwZ/r4Lcuy6vUZsna7TXZ7ZD9zNirKHvTv5i5c5xGOj4OL\nirIrOrpl5dpSXifhQCahyCQUmYQik1BkEqw55lGvgrp161Z99tlnmjNnjqTgghkfHx9ytdPr9crl\nOnHVzOVyhaxXVVWpR48ecrlcsixLHo9HnTp1Cqz7fD7Fx8ef9XwuV1y9Cu255HC0ifQIYWHSeTgc\nbRQfHxfpMcLKpHxNQSahyCQUmYQik1BkEqw55VGvgrp27VpVVlYG7gm1LEuWZSkzM1MTJkzQ22+/\nHbR9cXGx0tPTJUmpqalyu926+eabJUl+v1/bt2/X2LFj1blzZzmdTrnd7kBB3bFjh44fPx74oaqz\nUVl52IgrqA5Hm0bfb2kKn+9opEcI8PmOyuM5HOkxwqKlvU7CgUxCkUkoMglFJqHIJJhpeZzNxaZ6\nFdRHHnlE+fn5ga/Ly8t12223ac2aNaqrq9PSpUu1cuVKjRw5Ulu3btXmzZu1fPlySVJubq6mTp2q\nG2+8UVdeeaVeeeUVxcbGKisrS3a7XWPHjtWLL76o1NRUxcbGauHChRo2bFjgCuzZ8Pst+f2nv22g\nqdTV+U95I3JzYsIL+aSWkul3tcRzaiwyCUUmocgkFJmEIpNgzSmPehXU9u3bq3379oGva2trZbPZ\ndPHFJ+4tfOmllzRnzhzNnj1bCQkJmj9/vhITEyVJgwcP1oMPPqj8/HxVVlYqLS1NS5cuDfxU9v33\n368jR47opptuUl1dna677jrNnNmEPx0HAAAAIzTqo04TEhJUWloa+DojI0NvvfXWKbfPyclRTk7O\n9661atVKBQUFKigoaMxIAAAAaOaaz49zAQAA4LxAQQUAAIBRKKgAAAAwCgUVAAAARqGgAgAAwCgU\nVAAAABiFggoAAACjUFABAABgFAoqAAAAjEJBBQAAgFEoqAAAADAKBRUAAABGoaACAADAKBRUAAAA\nGIWCCgAAAKNQUAEAAGAUCioAAACMQkEFAACAUSioAAAAMAoFFQAAAEahoAIAAMAoFFQAAAAYhYIK\nAAAAo1BQAQAAYBQKKgAAAIxCQQUAAIBRKKgAAAAwCgUVAAAARqGgAgAAwCgUVAAAABiFggoAAACj\nUFABAABgFAoqAAAAjEJBBQAAgFEoqAAAADAKBRUAAABGoaACAADAKPUuqF988YXGjx+vjIwMDRo0\nSFOmTFFFRYU++eQT9ezZU+np6UpPT1evXr2Unp6u9957L7DvsmXLNGLECPXr10/jxo2T2+0OrNXU\n1GjGjBnKysrSNddcowceeEBerzc8ZwkAAIBmo14FtaamRj/72c80YMAAbd26VevWrVNFRYUef/xx\nSVJCQoIKCwtVWFiooqIiFRYWavjw4ZKkjRs3atGiRZo/f762bNmirKwsTZo0SdXV1ZKkhQsXqrS0\nVMuXL9eGDRtkWZYefvjhMJ8uAAAATFevglpdXa0pU6bo5z//uVq1aqX4+HgNGzZMO3bsOOO+y5cv\n1+jRo5WWlqaYmBhNnDhRNptNGzdulN/v16pVq3TvvfeqY8eOcjgcys/P10cffaT9+/c3+OQAAADQ\n/NSroDocDt16662y20/stnPnTv3xj3/UjTfeKEk6dOiQ8vLyNGDAAGVlZem3v/1tYN+SkhIlJycH\nvrbZbEpKSlJxcbF27dqlgwcPKikpKbDevXt3tW7dOug2AAAAALR80Q3Zae/evRo2bJj8fr/Gjh2r\nvLw8lZaW6sorr9T48eP13HPP6a9//aseeOABORwOjR49Wl6vVw6HI+g4TqdTXq9XXq9XNptNTqcz\naN3hcMjj8Zz1XHa7TXa7rSGnFDZRUfagfzd3Jp1HVJRd0dHmzNMYLe11Eg5kEopMQpFJKDIJRSbB\nmmMeDSqol156qUpKSrR7924VFBTooYce0jPPPKNly5YFthk4cKBycnK0evVqjR49+nuPY1nWaZ/H\nsizZbGdfOF2uuHptfy45HG0iPUJYmHQeDkcbxcfHRXqMsDIpX1OQSSgyCUUmocgkFJkEa055NKig\nntSlSxdNmTJFOTk5euyxxxQfHx+0npCQoD/96U+SJJfLFXI1tKqqSj169JDL5ZJlWfJ4POrUqVNg\n3efzhRzzdCorDxtxBdXhaCOf76jq6vwRnSUcfL6jkR4hwOc7Ko/ncKTHCIuW9joJBzIJRSahyCQU\nmYQik2Cm5XE2F5vqVVC3bdumWbNmacOGDYHHbDabbDabPvnkE1VWVio3NzewVlZWps6dO0uSUlNT\n5Xa7dfPNN0uS/H6/tm/frrFjx6pz585yOp1yu92Bgrpjxw4dP35caWlpZz2f32/J7z/9VdnTqamp\nkdtd3OD9pRMvgsGDB6iuzq/a2si/CBrLhBfySS0l0+9qiefUWGQSikxCkUkoMglFJsGaUx71Kqip\nqak6dOiQFixYoLy8PB05ckQvvPCCMjIyFB0draeeekpdu3ZV//79tW3bNq1evVpPP/20JCk3N1dT\np07VjTfeqCuvvFKvvPKKYmNjlZWVJbvdrrFjx+rFF19UamqqYmNjtXDhQg0bNkwul+ucnPj3cbuL\n9YuFq9X+gi4NPsbBA7v1sqONrrgi+cwbAwAAIES9Cmq7du306quvavbs2crMzFTbtm01YMAAzZ07\nVxdffLEeffRRzZ49W+Xl5brwwgv12GOPaejQoZKkwYMH68EHH1R+fr4qKyuVlpampUuXKiYmRpJ0\n//3368iRI7rppptUV1en6667TjNnzgz/GZ9B+wu6qMMliU3+vAAAADih3vegJiYm6rXXXvvetTFj\nxmjMmDGn3DcnJ0c5OTnfu9aqVSsVFBSooKCgviMBAACgBWk+7zcAAACA8wIFFQAAAEahoAIAAMAo\nFFQAAAAYhYIKAAAAo1BQAQAAYBQKKgAAAIxCQQUAAIBRKKgAAAAwCgUVAAAARqGgAgAAwCgUVAAA\nABiFggoAAACjUFABAABgFAoqAAAAjEJBBQAAgFEoqAAAADAKBRUAAABGoaACAADAKBRUAAAAGIWC\nCgAAAKNQUAEAAGAUCioAAACMQkEFAACAUSioAAAAMAoFFQAAAEahoAIAAMAoFFQAAAAYhYIKAAAA\no1BQAQAAYBQKKgAAAIxCQQUAAIBRKKgAAAAwCgUVAAAARqGgAgAAwCgUVAAAABil3gX1iy++0Pjx\n45WRkaFBgwZpypQpOnDggCRp69atGjNmjPr27avs7GytW7cuaN9ly5ZpxIgR6tevn8aNGye32x1Y\nq6mp0YwZM5SVlaVrrrlGDzzwgLxebyNPDwAAAM1NvQpqTU2Nfvazn2nAgAHaunWr1q1bp4qKCs2a\nNUv79+/XPffco9tvv11bt27VI488ooKCgkAJ3bhxoxYtWqT58+dry5YtysrK0qRJk1RdXS1JWrhw\noUpLS7V8+XJt2LBBlmXp4YcfDv8ZAwAAwGj1KqjV1dWaMmWKfv7zn6tVq1aKj4/XsGHDtGPHDq1b\nt07dunXTqFGjFBMTo8zMTF1//fVasWKFJGn58uUaPXq00tLSFBMTo4kTJ8pms2njxo3y+/1atWqV\n7r33XnXs2FEOh0P5+fn66KOPtH///nNy4gAAADBTvQqqw+HQrbfeKrv9xG47d+7UH//4R91www1y\nu91KSUkJ2j45OVnFxcWSpJKSEiUnJwfWbDabkpKSVFxcrF27dungwYNKSkoKrHfv3l2tW7cOug0A\nAAAALV90Q3bau3evhg0bJr/fr7FjxyovL0933XWXLrnkkqDtnE6nPB6PJMnr9crhcISse71eeb1e\n2Ww2OZ3OoHWHwxHY/2zY7TbZ7baGnJIkKSoqfD8zFs5jRZJJ5xEVZVd0tDnzNMbJXE3KN9LIJBSZ\nhCKTUGQSikyCNcc8GlRQL730UpWUlGj37t0qKCjQtGnTTrmtzXbqwmhZ1mmfx7Ks0+7/71yuuHpt\n/+8cjjYN3vdcHiuSTDoPh6ON4uPjIj1GWJmUrynIJBSZhCKTUGQSikyCNac8GlRQT+rSpYumTJmi\nnJwcDRkyJORqp9frlcvlkiS5XK6Q9aqqKvXo0UMul0uWZcnj8ahTp06BdZ/Pp/j4+LOep7LycKOu\noPp8Rxu87/cdq67OH7bjRUo4M2ksn++oPJ7DkR4jLKKi7HI42rSY10k4kEkoMglFJqHIJBSZBDMt\nj7O52FSvgrpt2zbNmjVLGzZsCDxms9lks9mUmZmpVatWBW1fXFys9PR0SVJqaqrcbrduvvlmSZLf\n79f27ds1duxYde7cWU6nU263O1BQd+zYoePHjystLe2s5/P7Lfn9p78qezrh/E2rq/OrtjbyL4LG\nMuGFfFJLyfS7WuI5NRaZhCKTUGQSikxCkUmw5pRHvW5GSE1N1aFDh7RgwQJVV1ersrJSL7zwgjIy\nMpSdna29e/dq5cqVqqmp0aZNm7R582bddtttkqTc3FytWbNGhYWFqq6u1uLFixUbG6usrCzZ7XaN\nHTtWL774osrLy+XxeLRw4UINGzYscAUWAAAA54d6XUFt166dXn31Vc2ePVuZmZlq27atBgwYoLlz\n58rlcumll17SnDlzNHv2bCUkJGj+/PlKTEyUJA0ePFgPPvig8vPzVVlZqbS0NC1dulQxMTGSpPvv\nv19HjhzRTTfdpLq6Ol133XWaOXNm+M8YAAAARqv3PaiJiYl67bXXvnctIyNDb7311in3zcnJUU5O\nzveutWrVSgUFBSooKKjvSAAAAGhBms/7DQAAAOC8QEEFAACAUSioAAAAMAoFFQAAAEZp1Bv1I5S/\nrlbbt28Py5vhpqSkBd7lAAAA4HxBQQ2zw959WrTtPTk6N+79W317KjVHM3XVVX3DNBkAAEDzQEE9\nBxydXXJdfnGkxwAAAGiWuAcVAAAARqGgAgAAwCgUVAAAABiFggoAAACjUFABAABgFAoqAAAAjEJB\nBQAAgFEoqAAAADAKBRUAAABGoaACAADAKBRUAAAAGIWCCgAAAKNQUAEAAGAUCioAAACMQkEFAACA\nUaIjPQBwPqqpqdGnn26Xz3dUdXX+Bh8nJSVNMTExYZwMAIDIo6ACEVBSUqyHV86Uo7Orwcfw7anU\nHM3UVVf1DeNkAABEHgUViBBHZ5dcl18c6TEAADAO96ACAADAKBRUAAAAGIWCCgAAAKNQUAEAAGAU\nCioAAACMQkEFAACAUSioAAAAMAoFFQAAAEahoAIAAMAoFFQAAAAYhYIKAAAAo9S7oO7du1d5eXnq\n37+/Bg0apIcffliHDh3SN998o549eyo9PV3p6enq1auX0tPT9eqrrwb2Xb9+vUaOHKk+ffrolltu\n0ZYtW4KO/eyzz2ro0KHq37+/7rrrLu3Zs6fxZwgAAIBmpd4FdfLkyXI6ndq0aZNWrVqlL7/8Uk89\n9ZQkyWazqbCwUIWFhSoqKlJhYaEmTJggSSotLdX06dM1bdo0bdu2TePHj1deXp6+/fZbSdJrr72m\nd955Ry+//LI+/PBDde3aVXl5eWE8VQAAADQH9SqoBw8eVFpamqZOnarWrVurY8eOGjVqlD799NMz\n7rty5UoNGTJEgwcPVkxMjLKzs9WjRw+tXbtWkrR8+XJNmDBB3bp1U9u2bTVlyhSVlZWpqKioYWcG\nAACAZqleBbV9+/aaO3euXC5X4LG9e/eqY8eOkiTLsvTLX/5SgwYN0jXXXKOFCxeqrq5OkuR2u5Wc\nnBx0vOTkZBUXF+vYsWP66quvlJSUFFiLi4tT165dVVxc3OCTAwAAQPPTqB+SKi4u1uuvv667775b\nMTEx6tOnj4YNG6ZNmzZpyZIlWrt2rRYvXixJ8ng8cjgcQfs7nU55PB5VVVXJsiw5nc7vXQcAAMD5\nI7qhO/7tb3/TPffco2nTpmnAgAGSpN///veB9bS0NE2aNElLly7Vfffd973HsCxLNpvtlM9hWVa9\nZrLbbbLbT328M4mKMutNDaKi7IqOjuxMJmViQh7h0pjX6Xe1pExOvtZMes1FGpmEIpNQZBKKTII1\nxzwaVFA//PBDTZs2TTNmzNDIkSNPuV1CQoL2798vSXK5XCFXQ6uqquRyudShQwfZ7fZTrp8tlyvu\ntIX3TByONg3e91xwONooPj4u4jOYwoQ8wqVdu9ZhOU5LyuQkk15zpiCTUGQSikxCkUmw5pRHvQvq\n3//+d02fPl3PP/+8MjMzA49v3bpVhYWFmjx5cuCxsrIyJSQkSJJSU1PldruDjlVcXKzs7GzFxMQo\nMTFRJSUlysjIkCT5fD7t3r1b6enpZz1bZeXhRl2Z8vmONnjfc8HnOyqP53DEZzCFCXmEy6FD1WE5\nTkvKJCrKLoejjXy+o6qr80d6HCOQSSgyCUUmocgkmGl5nM2FlXoV1Lq6OhUUFOihhx4KKqfSiftF\nFy1apISEBP3oRz9SaWmpfvOb32jixImSpLFjx2rMmDHatGmTMjMztXbtWu3atUvZ2dmSpNzcXC1d\nulSDBw9Wx44dtWDBAqWkpCglJeWs5/P7Lfn99bstIPj8Iv+b9l11dX7V1kZ2JpMyMSGPcGnM6/S7\nWlImJ7XEc2osMglFJqHIJBSZBGtOedSroH722WfauXOn5syZoyeeeEI2my1wH+mGDRv07LPP6oUX\nXlBBQYH63yXwAAAgAElEQVQcDofuvPNO/eQnP5EkJSYmasGCBZo3b5727dunK664QkuWLNEFF1wg\nScrJyVFFRYXuvPNOHTlyRP3799evf/3r8J8xAAAAjFavgpqRkaHS0tJTrnfq1ElDhw495frQoUNP\nu56Xl8eb88N4NTU1crsb9/ZnX375jzBNAwBAy9Pgn+IHzldud7F+sXC12l/QpcHH+Hbnp+qSHcah\nAABoQSioQAO0v6CLOlyS2OD9Dx7YI+mf4RsIAIAWpPm8IRYAAADOCxRUAAAAGIWCCgAAAKNQUAEA\nAGAUCioAAACMQkEFAACAUSioAAAAMAoFFQAAAEahoAIAAMAoFFQAAAAYhYIKAAAAo1BQAQAAYBQK\nKgAAAIxCQQUAAIBRKKgAAAAwCgUVAAAARqGgAgAAwCgUVAAAABiFggoAAACjUFABAABgFAoqAAAA\njEJBBQAAgFEoqAAAADAKBRUAAABGoaACAADAKBRUAAAAGIWCCgAAAKNQUAEAAGAUCioAAACMQkEF\nAACAUSioAAAAMAoFFQAAAEahoAIAAMAoFFQAAAAYhYIKAAAAo9S7oO7du1d5eXnq37+/Bg0apIcf\nfliHDh2SJJWWluqOO+5QRkaGhg8frldffTVo3/Xr12vkyJHq06ePbrnlFm3ZsiVo/dlnn9XQoUPV\nv39/3XXXXdqzZ08jTg0AAADNUb0L6uTJk+V0OrVp0yatWrVKX375pZ566ikdO3ZMkydPVmZmpv77\nv/9bzz77rJYsWaI///nPkk6U1+nTp2vatGnatm2bxo8fr7y8PH377beSpNdee03vvPOOXn75ZX34\n4Yfq2rWr8vLywnu2AAAAMF69CurBgweVlpamqVOnqnXr1urYsaNGjRqlTz/9VB999JFqa2t19913\nq3Xr1kpOTtaYMWP05ptvSpJWrlypIUOGaPDgwYqJiVF2drZ69OihtWvXSpKWL1+uCRMmqFu3bmrb\ntq2mTJmisrIyFRUVhf+sAQAAYKx6FdT27dtr7ty5crlcgcf27dunjh07yu1268orr5TNZgusJScn\nq7i4WJLkdruVnJwcdLyT68eOHdNXX32lpKSkwFpcXJy6du0a2B8AAADnh+jG7FxcXKzXX39dixcv\n1rvvviuHwxG03qFDB1VVVUmSPB5PyLrT6dRXX32lqqoqWZYlp9MZsu7xeM56HrvdJrvdduYNTyEq\nyqyfGYuKsis6OrIzmZSJCXmcnMMUpmQSDidzNSnfSCOTUGQSikxCkUmw5phHgwvq3/72N91zzz16\n6KGHlJmZqXfffTdkG8uygq6oNmS9PlyuuNMe70wcjjYN3vdccDjaKD4+LuIzmMKEPE7OYQpTMgkn\nk/I1BZmEIpNQZBKKTII1pzwaVFA//PBDTZs2TTNmzNDIkSMlSfHx8dq1a1fQdl6vVx06dJAkuVyu\nkKuhVVVVcrlc6tChg+x2+ynXz1Zl5eFGXUH1+Y42eN9zwec7Ko/ncMRnMIUJeZycwxSmZBIOUVF2\nORxt5PMdVV2dP9LjGIFMQpFJKDIJRSbBTMvjbC6s1Lug/v3vf9f06dP1/PPPKzMzM/B4Wlqa3njj\nDfn9ftntJy4hFxcXq1evXpKk1NRUud3uoGMVFxcrOztbMTExSkxMVElJiTIyMiRJPp9Pu3fvVnp6\n+lnP5vdb8vvrd9X1u0z4Tfuuujq/amsjO5NJmZiQx8k5TGFKJuHUEs+pscgkFJmEIpNQZBKsOeVR\nr5sR6urqVFBQEPi2/ndde+21ateunRYvXqzq6moVFhZq1apVuv322yVJY8eO1ccff6xNmzappqZG\nK1eu1K5du5SdnS1Jys3N1bJly1RWVqZDhw5pwYIFSklJUUpKSphOFQAAAM1Bva6gfvbZZ9q5c6fm\nzJmjJ554QjabLXAf6YYNG7RkyRLNmDFDS5cu1YUXXqipU6fq2muvlSQlJiZqwYIFmjdvnvbt26cr\nrrhCS5Ys0QUXXCBJysnJUUVFhe68804dOXJE/fv3169//evwnzEAAACMVq+CmpGRodLS0tNu8/vf\n//6Ua0OHDtXQoUNPuZ6Xl8eb8wMAAJznms/7DQAAAOC8QEEFAACAUSioAAAAMAoFFQAAAEahoAIA\nAMAoFFQAAAAYhYIKAAAAo1BQAQAAYBQKKgAAAIxCQQUAAIBRKKgAAAAwCgUVAAAARqGgAgAAwCgU\nVAAAABiFggoAAACjUFABAABgFAoqAAAAjEJBBQAAgFEoqAAAADAKBRUAAABGoaACAADAKBRUAAAA\nGIWCCgAAAKNQUAEAAGAUCioAAACMQkEFAACAUSioAAAAMAoFFQAAAEahoAIAAMAoFFQAAAAYhYIK\nAAAAo1BQAQAAYBQKKgAAAIxCQQUAAIBRKKgAAAAwCgUVAAAARql3Qd28ebMGDhyoqVOnBj3+xz/+\nUUlJSUpPT1d6erp69eql9PR0FRcXB7Z59tlnNXToUPXv31933XWX9uzZE1jz+XzKz8/XwIEDNXjw\nYD322GOqqalpxKkBAACgOapXQX3llVc0b948XXbZZd+73q9fPxUWFqqwsFBFRUUqLCxUWlqaJOm1\n117TO++8o5dfflkffvihunbtqry8vMC+jz76qKqrq7V+/XqtXr1aZWVlmj9/fsPPDAAAAM1SvQpq\n69attWLFCnXp0qXeT7R8+XJNmDBB3bp1U9u2bTVlyhSVlZWpqKhIBw4c0AcffKCpU6fK6XTqoosu\n0j333KPVq1errq6u3s8FAACA5qteBXXcuHFq167dKdfLy8v105/+VFdffbX+4z/+Q2vXrpUkHTt2\nTF999ZWSkpIC28bFxalr164qLi5WaWmpoqKilJiYGFhPSUnR4cOHtXPnzvqeEwAAAJqx6HAdyOVy\n6bLLLtODDz6oyy+/XO+//75+8YtfqGPHjurWrZssy5LT6Qzax+l0yuPxyOl0qn379iFrkuTxeM56\nBrvdJrvd1uBziIoy62fGoqLsio6O7EwmZWJCHifnMIUpmYTDyVxNyjfSyCQUmYQik1BkEqw55hG2\ngpqVlaWsrKzA1z/+8Y/1/vvva/Xq1SE/UHWSZVlnPK7NdvaF0+WKq9f2/87haNPgfc8Fh6ON4uPj\nIj6DKUzI4+QcpjAlk3AyKV9TkEkoMglFJqHIJFhzyiNsBfX7JCQkyO12q0OHDrLb7SFXQ6uqquRy\nueRyueTz+WRZVqBger1eSSeuzJ6tysrDjbqC6vMdbfC+54LPd1Qez+GIz2AKE/I4OYcpTMkkHKKi\n7HI42sjnO6q6On+kxzECmYQik1BkEopMgpmWx9lcWAlbQX3jjTfkdDr1ox/9KPBYWVmZOnfurJiY\nGCUmJqqkpEQZGRmSTryt1O7du9W7d2916tRJkvTFF18E7lMtKiqS0+lUt27dznoGv9+S33/mq7Kn\nYsJv2nfV1flVWxvZmUzKxIQ8Ts5hClMyCaeWeE6NRSahyCQUmYQik2DNKY+w3YxQU1OjuXPnqqSk\nRLW1tXr77be1efNm5ebmSpJyc3O1bNkylZWV6dChQ1qwYIGSk5OVnJys+Ph4DR8+XM8995w8Ho/K\ny8u1ePFijRkzRnZ787lfAgAAAI1XryuovXr1ks1mU21trSTp/fffl81mU2Fhoe68804dOXJEDzzw\ngCoqKvSDH/xAixYtClwRzcnJUUVFRWC7/v376/nnnw8c+/HHH9fMmTP1wx/+UK1atVJ2drby8/PD\neKoAAABoDupVUIuKik67PnnyZE2ePPmU63l5eUFvzv9d7dq10zPPPFOfcQAAANAC8f1zAAAAGIWC\nCgAAAKNQUAEAAGAUCioAAACMQkEFAACAUSioAAAAMAoFFQAAAEahoAIAAMAoFFQAAAAYhYIKAAAA\no1BQAQAAYBQKKgAAAIxCQQUAAIBRKKgAAAAwCgUVAAAARqGgAgAAwCgUVAAAABiFggoAAACjUFAB\nAABgFAoqAAAAjEJBBQAAgFEoqAAAADAKBRUAAABGoaACAADAKBRUAAAAGIWCCgAAAKNER3oAAJCk\nmpoaffrpdvl8R1VX52/wcVJS0hQTExPGyQAATY2CCsAIJSXFenjlTDk6uxp8DN+eSs3RTF11Vd8w\nTgYAaGoUVADGcHR2yXX5xZEeAwAQYdyDCgAAAKNQUAEAAGAUCioAAACMQkEFAACAUSioAAAAMAoF\nFQAAAEahoAIAAMAoFFQAAAAYpd4FdfPmzRo4cKCmTp0asrZ+/XqNHDlSffr00S233KItW7YErT/7\n7LMaOnSo+vfvr7vuukt79uwJrPl8PuXn52vgwIEaPHiwHnvsMdXU1DTglAAAANCc1augvvLKK5o3\nb54uu+yykLXS0lJNnz5d06ZN07Zt2zR+/Hjl5eXp22+/lSS99tpreuedd/Tyyy/rww8/VNeuXZWX\nlxfY/9FHH1V1dbXWr1+v1atXq6ysTPPnz2/c2QEAAKDZqVdBbd26tVasWKEuXbqErK1cuVJDhgzR\n4MGDFRMTo+zsbPXo0UNr166VJC1fvlwTJkxQt27d1LZtW02ZMkVlZWUqKirSgQMH9MEHH2jq1Kly\nOp266KKLdM8992j16tWqq6sLz5kCAACgWahXQR03bpzatWv3vWtut1vJyclBjyUnJ6u4uFjHjh3T\nV199paSkpMBaXFycunbtquLiYpWWlioqKkqJiYmB9ZSUFB0+fFg7d+6sz4gAAABo5qLDdSCPxyOH\nwxH0mNPp1FdffaWqqipZliWn0xmy7vF45HQ61b59+5C1k8c9W3a7TXa7rYFnIEVFmfUzY1FRdkVH\nR3YmkzIxIY+Tc5jClEzCoTH/7X5XS8rk5GvNpNdcpJFJKDIJRSbBmmMeYSuo38eyLNlsp/5Lx7Ks\nMx7jdPv/O5crrl7b/zuHo02D9z0XHI42io+Pi/gMpjAhj5NzmMKUTMKhXbvWYTlOS8rkJJNec6Yg\nk1BkEopMgjWnPMJWUF0uV8jVzqqqKrlcLnXo0EF2u/2U6y6XSz6fL6jQer3ewHHPVmXl4UZdhfH5\njjZ433PB5zsqj+dwxGcwhQl5nJzDFKZkEg6HDlWH5TgtKZOoKLscjjby+Y6qrs4f6XGMQCahyCQU\nmQQzLY+zuYgQtoKampoqt9sd9FhxcbGys7MVExOjxMRElZSUKCMjQ9KJt5XavXu3evfurU6dOkmS\nvvjii8B9qkVFRXI6nerWrdtZz+D3W/L7z3xV9lRM+E37rro6v2prIzuTSZmYkMfJOUxhSibh0Jj/\ndr+rJWVyUks8p8Yik1BkEopMgjWnPMJ2M8LYsWP18ccfa9OmTaqpqdHKlSu1a9cuZWdnS5Jyc3O1\nbNkylZWV6dChQ1qwYIGSk5OVnJys+Ph4DR8+XM8995w8Ho/Ky8u1ePFijRkzRnZ787lfAgAAAI1X\nryuovXr1ks1mU21trSTp/fffl81mU2FhoRITE7VgwQLNmzdP+/bt0xVXXKElS5boggsukCTl5OSo\noqJCd955p44cOaL+/fvr+eefDxz78ccf18yZM/XDH/5QrVq1UnZ2tvLz88N4qgAAAGgO6lVQi4qK\nTrs+dOhQDR069JTreXl5QW/O/13t2rXTM888U59xAAAA0ALx/XMAAAAYhYIKAAAAo1BQAQAAYJRz\n+kb9AM4PNTU1cruLG3WML7/8R5imAQA0dxRUAI3mdhfrFwtXq/0FXRp8jG93fqou2WEcCgDQbFFQ\nAYRF+wu6qMMliQ3e/+CBPZL+Gb6BAADNFvegAgAAwCgUVAAAABiFggoAAACjUFABAABgFAoqAAAA\njEJBBQAAgFEoqAAAADAKBRUAAABGoaACAADAKBRUAAAAGIWCCgAAAKNQUAEAAGAUCioAAACMQkEF\nAACAUSioAAAAMAoFFQAAAEahoAIAAMAoFFQAAAAYhYIKAAAAo1BQAQAAYBQKKgAAAIxCQQUAAIBR\nKKgAAAAwCgUVAAAARqGgAgAAwCgUVAAAABiFggoAAACjUFABAABgFAoqAAAAjEJBBQAAgFHCWlB7\n9uypXr16KT09PfDvOXPmSJK2bt2qMWPGqG/fvsrOzta6deuC9l22bJlGjBihfv36ady4cXK73eEc\nDQAAAM1EdDgPZrPZ9N5776lTp05Bj+/fv1/33HOPZsyYoRtuuEF/+9vfdPfdd6t79+5KSUnRxo0b\ntWjRIr3yyiu68sor9V//9V+aNGmS/vznP6t169bhHBEAAACGC+sVVMuyZFlWyOPr1q1Tt27dNGrU\nKMXExCgzM1PXX3+9VqxYIUlavny5Ro8erbS0NMXExGjixImy2WzauHFjOMcDAABAMxD2e1AXLFig\n6667Tv369dOMGTN05MgRud1upaSkBG2XnJys4uJiSVJJSYmSk5MDazabTUlJSYF1AAAAnD/C+i3+\n3r17a+DAgXrqqae0Z88eTZkyRY8//ri8Xq8uueSSoG2dTqc8Ho8kyev1yuFwhKx7vd56Pb/dbpPd\nbmvw/FFRZv3MWFSUXdHRkZ3JpExMyOPkHKYgk1CmZBIOJ3M1Kd9II5NQZBKKTII1xzzCWlDfeOON\nwK+7d++uqVOn6u6771ZGRsb3bm+znbpMft+tAmficsWd9phn4nC0afC+54LD0Ubx8XERn8EUJuRx\ncg5TkEkoUzIJJ5PyNQWZhCKTUGQSrDnlEdaC+u8SEhJUV1cnu90euFp6ktfrlcvlkiS5XK6Q9aqq\nKvXo0aNez1dZebhRV1B9vqMN3vdc8PmOyuM5HPEZTGFCHifnMAWZhDIlk3CIirLL4Wgjn++o6ur8\nkR7HCGQSikxCkUkw0/I4m4sIYSuopaWlWrt2rX75y18GHisrK1NsbKyysrK0evXqoO2Li4uVnp4u\nSUpNTZXb7dbNN98sSfL7/dq+fbvGjBlTrxn8fkt+f/2vvJ5kwm/ad9XV+VVbG9mZTMrEhDxOzmEK\nMgllSibh1BLPqbHIJBSZhCKTYM0pj7DdjOByufTmm2/q5ZdfVk1Njb7++mv9+te/1m233aaRI0dq\n7969WrlypWpqarRp0yZt3rxZt912myQpNzdXa9asUWFhoaqrq7V48WLFxsZqyJAh4RoPAAAAzUTY\nrqB27NhRS5cu1YIFC/Tiiy8qNjZWo0aN0pQpU9SqVSu99NJLmjNnjmbPnq2EhATNnz9fiYmJkqTB\ngwfrwQcfVH5+viorK5WWlqalS5cqJiYmXOMBAACgmQjrPagZGRlBPyj172tvvfXWKffNyclRTk5O\nOMcBAABAM9R83m8AAAAA5wUKKgAAAIxCQQUAAIBRKKgAAAAwCgUVAAAARqGgAgAAwCgUVAAAABgl\nrO+DCgAIn5qaGn366fZGf352SkoaH3wCoFmhoAKAoUpKivXwyplydHY1+Bi+PZWao5m66qq+YZwM\nAM4tCioAGMzR2SXX5RdHegwAaFLcgwoAAACjUFABAABgFAoqAAAAjEJBBQAAgFEoqAAAADAKBRUA\nAABGoaACAADAKBRUAAAAGIWCCgAAAKNQUAEAAGAUCioAAACMQkEFAACAUSioAAAAMEp0pAcAgJao\npqZGbndxo47x5Zf/CNM0ANC8UFAB4Bxwu4v1i4Wr1f6CLg0+xrc7P1WX7DAOBQDNBAUVAM6R9hd0\nUYdLEhu8/8EDeyT9M3wDAUAzwT2oAAAAMAoFFQAAAEahoAIAAMAo3IMKAGg2ampq9Omn2+XzHVVd\nnb/Bx0lJSVNMTEwYJwMQThRUAECzUVJSrIdXzpSjs6vBx/DtqdQczdRVV/UN42QAwomCCgBoVhyd\nXXJdfnGkxwBwDnEPKgAAAIxCQQUAAIBR+BY/AKBJ8PGvAM4WBRUA0CT4+FcAZ8uogvrNN99o9uzZ\n+vzzzxUXF6cf//jHeuihhyI9FgAgTPj41/DjrbfQEhlVUO+77z6lpaVp48aNOnDggO666y5deOGF\nGj9+fKRHAwDASLz1VihKe/NnTEEtLi7Wjh07tGzZMsXFxSkuLk4TJkzQsmXLKKgAAJwGb70VjNLe\n/BlTULdv366EhAS1a9cu8FhycrK+/vprHT58WHFxcRGcDgAANCeU9ubNmILq9XrlcDiCHuvQoUNg\n7WwKqt1uk91ua/AMUVF2HTywu8H7S9KRqnL59lQ26hjSif9zi+pvV3R0ZN8JzJRMTMlDIpPvQyah\nyCQUmYT6+9//1uhj7NjxBZn8GzIJZrfb1K5dax06VC2/32rUsfr0aZoryjbLsho3aZgsWbJEf/7z\nn7VixYrAY7t27dKIESP0wQcf6NJLL43gdAAAAGgqkf/fgv/P5XLJ4/EEPVZVVSWbzab4+PgITQUA\nAICmZkxBTU1N1d69e+X1egOPFRUV6fLLL1ebNm0iOBkAAACakjEFNSkpSb169dIzzzyjQ4cOqays\nTL/97W91++23R3o0AAAANCFj7kGVpG+//VYFBQX65JNP1K5dO+Xm5uree++N9FgAAABoQkYVVAAA\nAMCYb/EDAAAAEgUVAAAAhqGgAgAAwCgUVAAAABiFggoAAACjUFABAABgFAoqmty+ffsiPYIRPB6P\nDh06FOkxjLJ+/XodO3Ys0mMY4fjx4/r222/FOwH+r5qaGpWXl8vv90d6FBjO5/PJ5/NFegw0AgUV\nYXXo0CEVFBRoxIgR+slPfqJt27aFbDNixIgITBY5FRUVuu+++zR8+HA9//zzsixLU6dOVWZmpvr1\n66c77rhD//rXvyI9phGeeOIJVVVVRXqMJjdnzpzAr48cOaLp06frqquu0pAhQ9S7d2/NmzdPx48f\nj+CETe+bb77R3LlzJUmVlZXKy8tT7969dd111yk9PV2PPfaYjh49GuEpm1ZSUpKmTp2qb7/9NtKj\nGKOyslL5+fkaPny4nnzySdXU1Oi+++7T1Vdfrf79++vOO+/U/v37Iz1mk3vvvfc0ceJEDRo0SKmp\nqerdu7eGDx+u6dOn64svvoj0eGeFgnqOpaenR3qEJjV37lyVlpbqjjvuUGpqqiZPnqzXX389aJvz\n7YrQ3Llz5fV6NW7cOH344Yd6/PHHVVFRoTfeeENvvPGGWrduraeeeirSYzaZnj17Kikp6Xv/8Xg8\nuvbaa5WUlBTpMZvUihUrAr9+8skntX37di1evFhvv/225s+fr48//ljPPvtsBCdsegUFBYFfz5o1\nS/v379fSpUu1fv16LV68WGVlZXriiSciOGHTi46OVo8ePXTzzTfrmWee0YEDByI9UsT96le/0sGD\nBzVhwgT94x//0F133SW/36+33npLb775ptq1a6cnn3wy0mM2qd/85jd6+umndf3112v69OlKT0/X\no48+qqlTpyomJka5ubn64IMPIj3mmVk4p9LS0iI9QpMaOHCgVV5eHvi6qKjIuvrqq601a9YEHuvV\nq1ckRouYgQMHWpWVlZZlWVZZWZmVlJRk7dq1K7B+4MAB65prronUeE1u7ty51tVXX2299NJL1j//\n+c/AP3v27LGuvvpq6/PPP7f++c9/RnrMJvXdPyf69etnlZWVBa3v3LnTGjBgQFOPFVG9e/e2Dh06\nZFmWZV199dVBf65YlmVVVFRYffr0icRoEXPyz85vvvnG+uUvf2n16dPHmj59uvWnP/0pkNX5ZuDA\ngZbX67Us68RromfPnta//vWvwHplZaU1cODASI0XEddee6315ZdfBr7etWuXddtttwW+3rx5szVi\nxIhIjFYv0ZEuyM3Z1KlTz7hNXV1dE0xijuPHj8vpdAa+TktL0+LFizVp0iRddNFFyszMPO+uoFZX\nV6tdu3aSpO7du8tut6tLly6B9bZt2+rIkSORGq/JPfLIIxo1apRmzZqlLVu2aNasWerevbskyWaz\n6ZJLLlHHjh0jPGXTstlsgV/HxcWpc+fOQesJCQnn3b25cXFxqqioUFxcnC688EJFRUUFrR8+fFjR\n0efnX2GXXnqpnnzySeXl5en111/Xr371K+3bt08dOnTQhRdeqHXr1kV6xCZz7NgxxcXFSZLatGkj\n6cSfqSe1adPmvLs95vDhw0F/hlx88cX66quvAl9nZmaqvLw8EqPVC9/ib4Rt27apvLxcMTExp/zn\nfNOvXz898cQTqqysDDzWt29fzZ8/X/n5+Vrx/9q5e5fWwSgM4E8HwUURERR0cfGtGmKDgkWd7GBR\nOzjbRXByFRH7D+gkIsUP0EFwEAoWOoo4uFS0UihKwMFBUFE3i1AD5jhcbrmlXq7DpW8kzw+y5GR4\nCCE55LxJKlXxMPYDwzCwu7tb/rDj6Oioop5MJmEYho5o2nR3d+Pg4ADRaBTxeByrq6solUq6Y2kj\nInh8fMTDwwMsy0Imk6mo7+3toaurS1M6PaanpzE3N4fj42PMzMxgaWkJ5+fnsG0bh4eHmJ2dxeTk\npO6YWnV0dGBxcREnJydIp9NYWFhALBbTHaumenp6kEwmcXt7i/X1dbS3t2N7exsA4Loutra2oJTS\nnLK2ent7sbGxAeDXvWVzc7P8EuDt7Q1ra2vo7OzUGfF7NL/B/dFOT08lGo1KsVj86zF+G2ff399L\nLBaTRCJRVbu8vJTx8XEJBoMakulj27YMDQ1JKpWqqo2Njcnw8LDc3NxoSOYNLy8vMj8/L6OjoxIK\nhapGuX6glJJgMChKKVFKVYzjVlZWxLIsubi40JhQj/39fYlEIuXz8nvr7++X5eVleX9/1x2xpvy2\nZOw7bNuWkZERUUrJ1NSU3N3dycTEhIRCITFNU8LhsFxfX+uOWVOFQkHC4bCYpil9fX0yODgouVxO\nRETS6bREIhG5urrSnPLfAiI+m7f+Zzs7O6ivr0c8Hv+ybpomCoVCjVPpVywW0dDQULX/4+MD+Xwe\nAwMDGlLp4zgOSqUSGhsbK/afnZ3BMIzyEgA/y2azyGQySCQSX147fmXbNpqbm3237OFPT09P5V9u\ntbS0oK2trWrk7we5XM53987vcF0Xz8/PaG1tRSAQgOM4yGazcF0XlmWhqalJd8Sae319RT6fBwBY\nls+3ZX4AAABTSURBVFV+9jiOg7q6uh8xyWSDSkRERESewjWoREREROQpbFCJiIiIyFPYoBIRERGR\np7BBJSIiIiJPYYNKRERERJ7CBpWIiIiIPIUNKhERERF5yicmE1qSDLYw9gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f5c04c4fdd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAqAAAAHWCAYAAACorzuEAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f5c316f4c10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_counts.plot(kind='bar')\n",
    "df_counts[df_counts.index > 5].plot(kind='bar')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Only keep proteins that have <5 prESTs per protein"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "low_num_uniprots = DF_prest.groupby('uniprot_id').count().aa_seq[DF_prest.groupby('uniprot_id').count().aa_seq < 5].index.tolist()\n",
    "DF_prest_filtered = DF_prest[DF_prest.uniprot_id.isin(low_num_uniprots)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "prest_id\n",
      "1    6354\n",
      "2    4122\n",
      "3    1712\n",
      "4     511\n",
      "Name: aa_seq, dtype: int64\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAawAAAGXCAYAAADvdWTCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niFSBhUVERKrAwiIiIlVgYRERkSr8P0rSLOB6qMRuAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f5c3125d6d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Number of prESTs per protein\n",
    "pie_var = DF_prest_filtered.groupby('uniprot_id').count().groupby('prest_id').count().aa_seq\n",
    "pie = (pie_var).plot(kind='pie',fontsize=10,figsize=(5,5),label='before')\n",
    "print(pie_var)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ML Pipeline\n",
    "[Back to Top](#Table-of-Contents)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This pipeline mirrors the classification workflow, except for splitting training and testing datasets."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train-Test Split\n",
    "[Back to Top](#Table-of-Contents)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Get the number of prESTs per each uniprot\n",
    "uniprot_counts = DF_prest_filtered.groupby('uniprot_id').count().prest_id\n",
    "\n",
    "# Add all uniprots with a single prEST to the training set\n",
    "training_uniprots = uniprot_counts[uniprot_counts == 1].index.tolist()\n",
    "\n",
    "# Randomly pick 1/4 of other uniprots and add them to training set\n",
    "random.seed(10)\n",
    "other_uniprots = uniprot_counts[uniprot_counts > 1].index.tolist()\n",
    "training_uniprots += random.sample(other_uniprots,len(other_uniprots)/4)\n",
    "\n",
    "# Add all remaining uniprots to test set\n",
    "testing_uniprots = list(set(uniprot_counts.index.tolist()) - set(training_uniprots))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of proteins: 12699\n",
      "Number of training proteins: 7940\n",
      "Number of testing proteins: 4759\n"
     ]
    }
   ],
   "source": [
    "print 'Total number of proteins:', len(DF_prest_filtered.uniprot_id.unique())\n",
    "print 'Number of training proteins:', len(training_uniprots)\n",
    "print 'Number of testing proteins:', len(testing_uniprots)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set size: 10245\n",
      "Testing set size: 11533\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/anand/.virtualenvs/tensorflow/local/lib/python2.7/site-packages/pandas/core/indexing.py:288: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self.obj[key] = _infer_fill_value(value)\n",
      "/home/anand/.virtualenvs/tensorflow/local/lib/python2.7/site-packages/pandas/core/indexing.py:465: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self.obj[item] = s\n"
     ]
    }
   ],
   "source": [
    "# Add all prESTs in training uniprots to training set\n",
    "DF_train = DF_prest_filtered[DF_prest_filtered.uniprot_id.isin(training_uniprots)]\n",
    "DF_train.loc[:,'status'] = ['Train']*len(DF_train)\n",
    "print 'Training set size:', len(DF_train)\n",
    "\n",
    "# Repeat for test set\n",
    "DF_test = DF_prest_filtered[DF_prest_filtered.uniprot_id.isin(testing_uniprots)]\n",
    "DF_test.loc[:,'status'] = ['Test']*len(DF_test)\n",
    "print 'Testing set size:', len(DF_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Create machine learning algorithm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Size of final dataset: 21778\n"
     ]
    }
   ],
   "source": [
    "DF_prest_final = pd.concat([DF_train,DF_test],axis=0)\n",
    "print 'Size of final dataset:',len(DF_prest_final)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "aa_features = ['list_comp_A','list_comp_R','list_comp_N',\n",
    "               'list_comp_D','list_comp_C','list_comp_Q','list_comp_E','list_comp_G',\n",
    "               'list_comp_H','list_comp_I','list_comp_L','list_comp_K','list_comp_M',\n",
    "               'list_comp_F','list_comp_P','list_comp_S','list_comp_T','list_comp_W',\n",
    "               'list_comp_Y','list_comp_V',\n",
    "               # Amino acid types\n",
    "               'frac_aliphatic', 'frac_uncharged_polar', 'frac_polar',\n",
    "               'frac_hydrophobic', 'frac_positive', 'frac_sulfur', 'frac_negative', 'frac_amide',\n",
    "               'frac_alcohol']\n",
    "\n",
    "disorder_features = ['disembl_COILS_frac','disembl_REM465_frac','disembl_HOTLOOPS_frac','ronn_avg','ronn_frac',\n",
    "                    'disopred_avg','disopred_pb_avg','disopred_frac','disopred_pb_frac']\n",
    "\n",
    "                    \n",
    "phys_features = ['aa_len',                                        \n",
    "                # Physical properties\n",
    "                 'bio_pI','bio_mW','bio_aromaticity','bio_instability','bio_gravy','abs_avg_charge','abs_charge',\n",
    "                 'avg_charge']\n",
    "                    \n",
    "rna_features = ['AAA','AAC','AAT','AAG','ACA','ACC','ACT','ACG','ATA','ATC','ATT','ATG','AGA','AGC','AGT',\n",
    "                'AGG','CAA','CAC','CAT','CAG','CCA','CCC','CCT','CCG','CTA','CTC','CTT','CTG','CGA','CGC',\n",
    "                'CGT','CGG','TAA','TAC','TAT','TAG','TCA','TCC','TCT','TCG','TTA','TTC','TTT','TTG','TGA',\n",
    "                'TGC','TGT','TGG','GAA','GAC','GAT','GAG','GCA','GCC','GCT','GCG','GTA','GTC','GTT','GTG',\n",
    "                'GGA','GGC','GGT','GGG',\n",
    "                'GC_content','list_nuc_A','list_nuc_C','list_nuc_G','list_nuc_T',\n",
    "                'GC30',\n",
    "                # SD sequences\n",
    "                'sd_like_fwd','sd_like_fwd_frac','sd_like_rev','sd_like_rev_frac','sd_seq_fwd',\n",
    "                'sd_seq_fwd_frac','sd_seq_rev','sd_seq_rev_frac',\n",
    "                # RNA folding\n",
    "                'RNA_folding_energy','RNA_40_energy','tAI']\n",
    "                    \n",
    "ss_features = ['acc20_mean','ss_helix', 'ss_ext', 'ss_c', 'ss8_helix', 'ss8_ext', 'ss8_turn', 'ss8_helix3',\n",
    "               'ss8_pi_helix', 'ss8_bridge', 'ss8_bend', 'ss8_coil', 'acc_frac',\n",
    "               'in_gravy','out_gravy', 'acc_hydrophilic_in','acc_hydrophilic_out',\n",
    "               'acc_hydrophobic_in','acc_hydrophobic_out']\n",
    "\n",
    "list_of_features = aa_features+disorder_features+phys_features+rna_features+ss_features + ['status']\n",
    "\n",
    "DF_prest_features = DF_prest_final[list_of_features]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No. of samples = 21778\n",
      "No. of features = 148\n"
     ]
    }
   ],
   "source": [
    "X = DF_prest_features.values\n",
    "y = DF_prest_final['expressed'].values\n",
    "n_samples, n_features = X.shape\n",
    "print(\"No. of samples = %d\" % n_samples)\n",
    "print(\"No. of features = %d\" % n_features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/anand/.virtualenvs/tensorflow/local/lib/python2.7/site-packages/sklearn/utils/validation.py:429: DataConversionWarning: Data with input dtype object was converted to float64 by StandardScaler.\n",
      "  warnings.warn(msg, _DataConversionWarning)\n"
     ]
    }
   ],
   "source": [
    "X_train = X[X[:,-1]=='Train'][:,:-1]\n",
    "X_test = X[X[:,-1]=='Test'][:,:-1]\n",
    "y_train = y[X[:,-1]=='Train']\n",
    "y_test = y[X[:,-1]=='Test']\n",
    "\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "scl = StandardScaler().fit(X_train)\n",
    "X_train = scl.transform(X_train)\n",
    "X_test = scl.transform(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train Random Forest\n",
    "[Back to Top](#Table-of-Contents)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "clf = RandomForestClassifier(n_estimators=200,oob_score=True,random_state=0,n_jobs=-1)\n",
    "\n",
    "params = {'n_estimators': [200,500,1000],\n",
    "           'max_depth': [6,10,15,20,None],\n",
    "           'min_samples_leaf': [1,2,5],\n",
    "           'min_samples_split': [2,5]\n",
    "        }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initial Training Accuracy: 1.0\n",
      "Initial Testing Accuracy: 0.678487817567\n",
      "Fitting 3 folds for each of 90 candidates, totalling 270 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:   19.3s\n",
      "[Parallel(n_jobs=-1)]: Done 184 tasks      | elapsed:  1.8min\n",
      "[Parallel(n_jobs=-1)]: Done 270 out of 270 | elapsed:  2.7min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final Training Accuracy: 0.90326988775\n",
      "Final Testing Accuracy: 0.678921356109\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "clf.fit(X_train,y_train)\n",
    "print 'Initial Training Accuracy:',clf.score(X_train,y_train)\n",
    "print 'Initial Testing Accuracy:',clf.score(X_test,y_test)\n",
    "\n",
    "n_subsamples = 1000\n",
    "X_small_train, y_small_train = X_train[:n_subsamples], y_train[:n_subsamples]\n",
    "\n",
    "gs = GridSearchCV(clf,\n",
    "                  params,\n",
    "                  cv=3,\n",
    "                  verbose=1,\n",
    "                  scoring = 'roc_auc',\n",
    "                  error_score=0,\n",
    "                  n_jobs=-1)\n",
    "\n",
    "_ = gs.fit(X_small_train, y_small_train)\n",
    "clf.set_params(**gs.best_params_)\n",
    "clf.fit(X_train, y_train)\n",
    "\n",
    "print 'Final Training Accuracy:',clf.score(X_train,y_train)\n",
    "print 'Final Testing Accuracy:',clf.score(X_test,y_test)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train Deep Learning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "tf.logging.set_verbosity(tf.logging.ERROR)\n",
    "TF_DIR = ROOT_DIR+'misc_files/tf_models/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_set = tf.contrib.learn.python.learn.datasets.base.Dataset(data=X_train.astype(np.float32),target=y_train.astype(np.float32))\n",
    "test_set = tf.contrib.learn.python.learn.datasets.base.Dataset(data=X_test.astype(np.float32),target=y_test.astype(np.float32))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Specify important validation metrics\n",
    "validation_metrics = {'accuracy': tf.contrib.metrics.streaming_accuracy,\n",
    "                      'precision': tf.contrib.metrics.streaming_precision,\n",
    "                      'recall': tf.contrib.metrics.streaming_recall}\n",
    "\n",
    "# Specify that all features have real-value data\n",
    "feature_columns = [tf.contrib.layers.real_valued_column(\"\", dimension=147)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Initialize model directory\n",
    "model_dir = TF_DIR+'retrospective/'\n",
    "\n",
    "# Initialize Monitor\n",
    "monitor_params = {'x':test_set.data,\n",
    "                 'y':test_set.target,\n",
    "                 'every_n_steps':50,\n",
    "                 'metrics':validation_metrics,\n",
    "                 'early_stopping_metric':'loss',\n",
    "                 'early_stopping_metric_minimize':True,\n",
    "                 'early_stopping_rounds':200}\n",
    "\n",
    "monitor = tf.contrib.learn.monitors.ValidationMonitor(**monitor_params)\n",
    "\n",
    "clf_params = {'feature_columns':feature_columns,\n",
    "              'hidden_units':[100,200,100],\n",
    "              'n_classes':2,\n",
    "              'activation_fn':tf.nn.relu,\n",
    "              'optimizer':tf.train.RMSPropOptimizer(learning_rate=0.0001),\n",
    "              'dropout':0.5,\n",
    "              'config':tf.contrib.learn.RunConfig(save_checkpoints_secs=1),\n",
    "              'model_dir': model_dir}\n",
    "\n",
    "fit_params = {'x':train_set.data,\n",
    "              'y':train_set.target,\n",
    "              'steps':10000,\n",
    "              'monitors':[monitor]}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.676060\n",
      "AUC: 0.752618\n"
     ]
    }
   ],
   "source": [
    "# Initialize classifier\n",
    "final_nn = tf.contrib.learn.DNNClassifier(**clf_params)\n",
    "\n",
    "# Train classifier\n",
    "# final_nn.fit(**fit_params)\n",
    "\n",
    "# Test classifier\n",
    "scores = final_nn.evaluate(x=test_set.data, y=test_set.target)\n",
    "print('Accuracy: {0:f}'.format(scores['accuracy']))\n",
    "print('AUC: {0:f}'.format(scores['auc']))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Figures"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Visualize spread of prESTs across different proteins using original dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>prest_id</th>\n",
       "      <th>uniprot_id</th>\n",
       "      <th>conc_cf</th>\n",
       "      <th>aa_seq</th>\n",
       "      <th>nt_seq</th>\n",
       "      <th>aa_len</th>\n",
       "      <th>true_nt_seq</th>\n",
       "      <th>AAA</th>\n",
       "      <th>AAT</th>\n",
       "      <th>AAC</th>\n",
       "      <th>...</th>\n",
       "      <th>ronn_results</th>\n",
       "      <th>ronn_frac</th>\n",
       "      <th>disopred_results</th>\n",
       "      <th>disopred_avg</th>\n",
       "      <th>disopred_pb_results</th>\n",
       "      <th>disopred_pb_avg</th>\n",
       "      <th>disopred_frac</th>\n",
       "      <th>disopred_pb_frac</th>\n",
       "      <th>expressed</th>\n",
       "      <th>status</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>140099</td>\n",
       "      <td>G3V537</td>\n",
       "      <td>2.9154</td>\n",
       "      <td>TYYAWKHELLGSGTCPALPPREVLGMEELEKLPEEQVAEEELECSA...</td>\n",
       "      <td>GACAAGCTTGCGGCCGCAACCTACTATGCCTGGAAGCATGAGCTGC...</td>\n",
       "      <td>144</td>\n",
       "      <td>ACCTACTATGCCTGGAAGCATGAGCTGCTGGGCTCTGGCACCTGCC...</td>\n",
       "      <td>0.006944</td>\n",
       "      <td>0.013889</td>\n",
       "      <td>0.013889</td>\n",
       "      <td>...</td>\n",
       "      <td>tyyawkhellGSGTcPALPPREVLGMEELEKLPEEQVAEEELECSa...</td>\n",
       "      <td>0.493056</td>\n",
       "      <td>tyyawkhellgsgtCPALPPREVLGMEELEKLPEEQVAEEELECSA...</td>\n",
       "      <td>0.421597</td>\n",
       "      <td>..............^^^^^^--^^^^^^^^^^^^^^^^^^^-^^^^...</td>\n",
       "      <td>0.352917</td>\n",
       "      <td>0.465278</td>\n",
       "      <td>0.423611</td>\n",
       "      <td>False</td>\n",
       "      <td>Train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>140225</td>\n",
       "      <td>P12724</td>\n",
       "      <td>1.4877</td>\n",
       "      <td>SLHARPPQFTRAQWFAIQHISLNPPRCTIAMRAINNYRWRCKNQNT...</td>\n",
       "      <td>GACAAGCTTGCGGCCGCATCACTCCATGCCAGACCCCCACAGTTTA...</td>\n",
       "      <td>136</td>\n",
       "      <td>TCACTCCATGCCAGACCCCCACAGTTTACGAGGGCTCAGTGGTTTG...</td>\n",
       "      <td>0.007353</td>\n",
       "      <td>0.051471</td>\n",
       "      <td>0.058824</td>\n",
       "      <td>...</td>\n",
       "      <td>SLharppqftraqwfaiqhislnpprctiamrainnyrwrcknqnt...</td>\n",
       "      <td>0.058824</td>\n",
       "      <td>slharppqftraqwfaiqhislnpprctiamrainnyrwrcknqnt...</td>\n",
       "      <td>0.034338</td>\n",
       "      <td>.................................................</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>False</td>\n",
       "      <td>Train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>140500</td>\n",
       "      <td>C9J919</td>\n",
       "      <td>2.2183</td>\n",
       "      <td>MQQPPPQKVRINLQQQPPPLQIKSVPLPTLKMQTTLVPPTVESSPE...</td>\n",
       "      <td>GACAAGCTTGCGGCCGCAATGCAACAGCCTCCACCTCAGAAAGTTC...</td>\n",
       "      <td>140</td>\n",
       "      <td>ATGCAACAGCCTCCACCTCAGAAAGTTCGAATCAATTTACAGCAAC...</td>\n",
       "      <td>0.014286</td>\n",
       "      <td>0.021429</td>\n",
       "      <td>0.021429</td>\n",
       "      <td>...</td>\n",
       "      <td>MQQPPPQKVRINLQQQPPPLQIKSVPLPTLKMQTTLVPPTVESSPE...</td>\n",
       "      <td>0.771429</td>\n",
       "      <td>MQQPPPQKVRINLQQQPpplQIKsvplPtLKMQTTLVPPTVESSPE...</td>\n",
       "      <td>0.489143</td>\n",
       "      <td>^^^^^^^^^^^^^^^^^...^^^....-.----^-------^^^^^...</td>\n",
       "      <td>0.453500</td>\n",
       "      <td>0.585714</td>\n",
       "      <td>0.471429</td>\n",
       "      <td>False</td>\n",
       "      <td>Train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>140587</td>\n",
       "      <td>A0A0D9SFN7</td>\n",
       "      <td>1.1715</td>\n",
       "      <td>LSAVAATKSPIIFIGTGEHIDDFEPFKTQPFISKLLGMGDIEGLID...</td>\n",
       "      <td>GACAAGCTTGCGGCCGCACTCAGTGCAGTCGCTGCCACAAAAAGTC...</td>\n",
       "      <td>113</td>\n",
       "      <td>CTCAGTGCAGTCGCTGCCACAAAAAGTCCGATTATTTTCATTGGTA...</td>\n",
       "      <td>0.061947</td>\n",
       "      <td>0.026549</td>\n",
       "      <td>0.008850</td>\n",
       "      <td>...</td>\n",
       "      <td>lsavaatkspiifigtgehiddfepfktqpfiskllgmgdieglid...</td>\n",
       "      <td>0.070796</td>\n",
       "      <td>Lsavaatkspiifigtgehiddfepfktqpfiskllgmgdieglid...</td>\n",
       "      <td>0.163451</td>\n",
       "      <td>-................................................</td>\n",
       "      <td>0.095575</td>\n",
       "      <td>0.106195</td>\n",
       "      <td>0.097345</td>\n",
       "      <td>False</td>\n",
       "      <td>Train</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>140607</td>\n",
       "      <td>P08311</td>\n",
       "      <td>3.0094</td>\n",
       "      <td>NDIMLLQLSRRVRRNRNVNPVALPRAQEGLRPGTLCTVAGWGRVSM...</td>\n",
       "      <td>CAGAATGACATCATGTTATTGCAGCTGAGCAGAAGAGTCAGACGGA...</td>\n",
       "      <td>147</td>\n",
       "      <td>AATGACATCATGTTATTGCAGCTGAGCAGAAGAGTCAGACGGAATC...</td>\n",
       "      <td>0.006803</td>\n",
       "      <td>0.020408</td>\n",
       "      <td>0.020408</td>\n",
       "      <td>...</td>\n",
       "      <td>ndimllqlsrRVRRNRNVNPVALPRAQEGLRPGtlctvagwgrvsm...</td>\n",
       "      <td>0.210884</td>\n",
       "      <td>ndimllqlsrrvrrnrnvnpvalpraqeglrpgtlctvagwgrvsm...</td>\n",
       "      <td>0.061293</td>\n",
       "      <td>.................................................</td>\n",
       "      <td>0.032381</td>\n",
       "      <td>0.034014</td>\n",
       "      <td>0.034014</td>\n",
       "      <td>False</td>\n",
       "      <td>Train</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 165 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    prest_id  uniprot_id  conc_cf  \\\n",
       "1     140099      G3V537   2.9154   \n",
       "2     140225      P12724   1.4877   \n",
       "10    140500      C9J919   2.2183   \n",
       "11    140587  A0A0D9SFN7   1.1715   \n",
       "14    140607      P08311   3.0094   \n",
       "\n",
       "                                               aa_seq  \\\n",
       "1   TYYAWKHELLGSGTCPALPPREVLGMEELEKLPEEQVAEEELECSA...   \n",
       "2   SLHARPPQFTRAQWFAIQHISLNPPRCTIAMRAINNYRWRCKNQNT...   \n",
       "10  MQQPPPQKVRINLQQQPPPLQIKSVPLPTLKMQTTLVPPTVESSPE...   \n",
       "11  LSAVAATKSPIIFIGTGEHIDDFEPFKTQPFISKLLGMGDIEGLID...   \n",
       "14  NDIMLLQLSRRVRRNRNVNPVALPRAQEGLRPGTLCTVAGWGRVSM...   \n",
       "\n",
       "                                               nt_seq  aa_len  \\\n",
       "1   GACAAGCTTGCGGCCGCAACCTACTATGCCTGGAAGCATGAGCTGC...     144   \n",
       "2   GACAAGCTTGCGGCCGCATCACTCCATGCCAGACCCCCACAGTTTA...     136   \n",
       "10  GACAAGCTTGCGGCCGCAATGCAACAGCCTCCACCTCAGAAAGTTC...     140   \n",
       "11  GACAAGCTTGCGGCCGCACTCAGTGCAGTCGCTGCCACAAAAAGTC...     113   \n",
       "14  CAGAATGACATCATGTTATTGCAGCTGAGCAGAAGAGTCAGACGGA...     147   \n",
       "\n",
       "                                          true_nt_seq       AAA       AAT  \\\n",
       "1   ACCTACTATGCCTGGAAGCATGAGCTGCTGGGCTCTGGCACCTGCC...  0.006944  0.013889   \n",
       "2   TCACTCCATGCCAGACCCCCACAGTTTACGAGGGCTCAGTGGTTTG...  0.007353  0.051471   \n",
       "10  ATGCAACAGCCTCCACCTCAGAAAGTTCGAATCAATTTACAGCAAC...  0.014286  0.021429   \n",
       "11  CTCAGTGCAGTCGCTGCCACAAAAAGTCCGATTATTTTCATTGGTA...  0.061947  0.026549   \n",
       "14  AATGACATCATGTTATTGCAGCTGAGCAGAAGAGTCAGACGGAATC...  0.006803  0.020408   \n",
       "\n",
       "         AAC   ...                                         ronn_results  \\\n",
       "1   0.013889   ...    tyyawkhellGSGTcPALPPREVLGMEELEKLPEEQVAEEELECSa...   \n",
       "2   0.058824   ...    SLharppqftraqwfaiqhislnpprctiamrainnyrwrcknqnt...   \n",
       "10  0.021429   ...    MQQPPPQKVRINLQQQPPPLQIKSVPLPTLKMQTTLVPPTVESSPE...   \n",
       "11  0.008850   ...    lsavaatkspiifigtgehiddfepfktqpfiskllgmgdieglid...   \n",
       "14  0.020408   ...    ndimllqlsrRVRRNRNVNPVALPRAQEGLRPGtlctvagwgrvsm...   \n",
       "\n",
       "    ronn_frac                                   disopred_results  \\\n",
       "1    0.493056  tyyawkhellgsgtCPALPPREVLGMEELEKLPEEQVAEEELECSA...   \n",
       "2    0.058824  slharppqftraqwfaiqhislnpprctiamrainnyrwrcknqnt...   \n",
       "10   0.771429  MQQPPPQKVRINLQQQPpplQIKsvplPtLKMQTTLVPPTVESSPE...   \n",
       "11   0.070796  Lsavaatkspiifigtgehiddfepfktqpfiskllgmgdieglid...   \n",
       "14   0.210884  ndimllqlsrrvrrnrnvnpvalpraqeglrpgtlctvagwgrvsm...   \n",
       "\n",
       "    disopred_avg                                disopred_pb_results  \\\n",
       "1       0.421597  ..............^^^^^^--^^^^^^^^^^^^^^^^^^^-^^^^...   \n",
       "2       0.034338  .................................................   \n",
       "10      0.489143  ^^^^^^^^^^^^^^^^^...^^^....-.----^-------^^^^^...   \n",
       "11      0.163451  -................................................   \n",
       "14      0.061293  .................................................   \n",
       "\n",
       "    disopred_pb_avg  disopred_frac  disopred_pb_frac  expressed  status  \n",
       "1          0.352917       0.465278          0.423611      False   Train  \n",
       "2          0.000000       0.000000          0.000000      False   Train  \n",
       "10         0.453500       0.585714          0.471429      False   Train  \n",
       "11         0.095575       0.106195          0.097345      False   Train  \n",
       "14         0.032381       0.034014          0.034014      False   Train  \n",
       "\n",
       "[5 rows x 165 columns]"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "DF_prest_final.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Create feature dataframe for test group\n",
    "DF_retro = DF_prest_final[DF_prest_final.status == 'Test'][aa_features+disorder_features+phys_features+rna_features+ss_features+['aa_seq','prest_id','uniprot_id','expressed']]\n",
    "DF_retro_features = DF_retro.drop(['aa_seq','prest_id','uniprot_id','expressed'],1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2    3118\n",
      "3    1267\n",
      "4     374\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "pie_var = DF_retro.groupby('uniprot_id').count().groupby('prest_id').count().aa_seq\n",
    "pie_var = pd.Series(np.array(pie_var),index=range(2,len(pie_var)+2))\n",
    "print pie_var"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7f5be943a950>"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZcAAAGqCAYAAADHggl+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f5bc8094b90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pie = (pie_var).plot(kind='pie',fontsize=12,figsize=(5,5))\n",
    "pie.set_ylabel('')\n",
    "pie.set_title('Number of prESTs per uniprot ID')\n",
    "#plt.savefig(FIGURE_DIR+'pie_chart.svg',transparent=True,format='svg')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Get classifier score and true score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "DF_retro.loc[:,'rf_score'] = clf.predict_proba(DF_retro_features)[:,1]\n",
    "DF_retro.loc[:,'nn_score'] = final_nn.predict_proba(test_set.data)[:,1]\n",
    "DF_retro.loc[:,'ens_score'] = DF_retro[['rf_score','nn_score']].mean(axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>prest_id</th>\n",
       "      <th>uniprot_id</th>\n",
       "      <th>expressed</th>\n",
       "      <th>rf_score</th>\n",
       "      <th>nn_score</th>\n",
       "      <th>aa_seq</th>\n",
       "      <th>ens_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3020371</td>\n",
       "      <td>X6RKY7</td>\n",
       "      <td>True</td>\n",
       "      <td>0.530606</td>\n",
       "      <td>0.692124</td>\n",
       "      <td>AEFRCFVKENKLIGISQRDYTQYYDHISKQKEEIRRCIQDFFKKHI...</td>\n",
       "      <td>0.611365</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3020373</td>\n",
       "      <td>X6RKY7</td>\n",
       "      <td>True</td>\n",
       "      <td>0.480333</td>\n",
       "      <td>0.545103</td>\n",
       "      <td>KEHVLHCQFSAWYPFFRGVTIKSVILPLPQNVKDYLLDDGTLVVSG...</td>\n",
       "      <td>0.512718</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2730004</td>\n",
       "      <td>X6RKQ2</td>\n",
       "      <td>False</td>\n",
       "      <td>0.414361</td>\n",
       "      <td>0.426968</td>\n",
       "      <td>TGGAFKGFKGLVVPSGGGRFSGFGSGAGGKPLEGLSNGNNITSAPP...</td>\n",
       "      <td>0.420664</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2730001</td>\n",
       "      <td>X6RKQ2</td>\n",
       "      <td>True</td>\n",
       "      <td>0.477594</td>\n",
       "      <td>0.619637</td>\n",
       "      <td>EGDSGECKGGDEEENDEPPKVVVTEVKEEDAFYSKKCKLFYKKDNE...</td>\n",
       "      <td>0.548615</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2730002</td>\n",
       "      <td>X6RKQ2</td>\n",
       "      <td>True</td>\n",
       "      <td>0.464907</td>\n",
       "      <td>0.429130</td>\n",
       "      <td>GSTKLQQESTFLFHGNKTEDTPDKKMEVASEKKTDPSSLGATSASF...</td>\n",
       "      <td>0.447019</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   prest_id uniprot_id expressed  rf_score  nn_score  \\\n",
       "0   3020371     X6RKY7      True  0.530606  0.692124   \n",
       "1   3020373     X6RKY7      True  0.480333  0.545103   \n",
       "2   2730004     X6RKQ2     False  0.414361  0.426968   \n",
       "3   2730001     X6RKQ2      True  0.477594  0.619637   \n",
       "4   2730002     X6RKQ2      True  0.464907  0.429130   \n",
       "\n",
       "                                              aa_seq  ens_score  \n",
       "0  AEFRCFVKENKLIGISQRDYTQYYDHISKQKEEIRRCIQDFFKKHI...   0.611365  \n",
       "1  KEHVLHCQFSAWYPFFRGVTIKSVILPLPQNVKDYLLDDGTLVVSG...   0.512718  \n",
       "2  TGGAFKGFKGLVVPSGGGRFSGFGSGAGGKPLEGLSNGNNITSAPP...   0.420664  \n",
       "3  EGDSGECKGGDEEENDEPPKVVVTEVKEEDAFYSKKCKLFYKKDNE...   0.548615  \n",
       "4  GSTKLQQESTFLFHGNKTEDTPDKKMEVASEKKTDPSSLGATSASF...   0.447019  "
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "DF_retro_final = DF_retro[['prest_id','uniprot_id','expressed','rf_score','nn_score','aa_seq','ens_score']].sort_values('uniprot_id',ascending=0).reset_index(drop=True)\n",
    "DF_retro_final.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Original Data Statistics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original total number of experiments: 11533\n",
      "Original total number of proteins: 4759\n"
     ]
    }
   ],
   "source": [
    "print 'Original total number of experiments:',len(DF_retro_final)\n",
    "print 'Original total number of proteins:',len(DF_retro_final.uniprot_id.unique())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original Number of passed experiments: 5680\n",
      "Original Pass rate: 49.25%\n",
      "Number of proteins with >1 high expression: 3572\n"
     ]
    }
   ],
   "source": [
    "print 'Original Number of passed experiments:',len(DF_retro_final[DF_retro_final.expressed])\n",
    "print 'Original Pass rate: %.2f%%'%(np.true_divide(len(DF_retro_final[DF_retro_final.expressed]),len(DF_retro_final))*100)\n",
    "print 'Number of proteins with >1 high expression:',len(DF_retro_final[DF_retro_final.expressed].uniprot_id.unique())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True positive: 4334\n"
     ]
    }
   ],
   "source": [
    "print 'True positive:', len(DF_retro_final[(DF_retro_final.ens_score > .5) & (DF_retro_final.expressed)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "False Positive: 2295\n"
     ]
    }
   ],
   "source": [
    "print 'False Positive:',len(DF_retro_final[(DF_retro_final.ens_score > .5) & (DF_retro_final.expressed==False)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True Negative: 3558\n"
     ]
    }
   ],
   "source": [
    "print 'True Negative:',len(DF_retro_final[(DF_retro_final.ens_score < .5) & (DF_retro_final.expressed==False)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "False Negative: 1346\n"
     ]
    }
   ],
   "source": [
    "print 'False Negative:',len(DF_retro_final[(DF_retro_final.ens_score < .5) & (DF_retro_final.expressed)])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Grab only the top predicted prESTs for each protein"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 0\n",
      "Total number of proposed experiments: 4759\n",
      "Total number of expressed proteins: 2947\n",
      "Overall pass rate: 0.619247741122\n",
      "\n",
      "Iteration 1\n",
      "Total number of proposed experiments: 6571\n",
      "Total number of expressed proteins: 3490\n",
      "Overall pass rate: 0.531121594887\n",
      "\n",
      "Iteration 2\n",
      "Total number of proposed experiments: 6987\n",
      "Total number of expressed proteins: 3565\n",
      "Overall pass rate: 0.510233290396\n",
      "\n",
      "Iteration 3\n",
      "Total number of proposed experiments: 7069\n",
      "Total number of expressed proteins: 3572\n",
      "Overall pass rate: 0.505304852171\n",
      "\n",
      "Iteration 4\n",
      "Total number of proposed experiments: 7069\n",
      "Total number of expressed proteins: 3572\n",
      "Overall pass rate: 0.505304852171\n",
      "\n",
      "Percent saved experiments: 38.7063209919 %\n"
     ]
    }
   ],
   "source": [
    "n = 5\n",
    "np.random.seed(0)\n",
    "\n",
    "output_df = pd.DataFrame(columns=['prest_id','uniprot_id','expressed','ens_score','aa_seq'])\n",
    "remaining_df = DF_retro_final.copy()\n",
    "\n",
    "for i in range(n):\n",
    "    print 'Iteration',i\n",
    "    new_output_df = remaining_df.sort_values(['uniprot_id','ens_score'],ascending=[1,0]).drop_duplicates('uniprot_id')\n",
    "    output_df = pd.concat([output_df,new_output_df])\n",
    "   \n",
    "    pred_pos_proteins = set(output_df[output_df.ens_score > 0.5].uniprot_id)\n",
    "    true_pos_proteins = set(output_df[output_df.expressed].uniprot_id)\n",
    "    print 'Total number of proposed experiments:',len(output_df)\n",
    "    print 'Total number of expressed proteins:',len(true_pos_proteins)\n",
    "    print 'Overall pass rate:',np.true_divide(len(true_pos_proteins),len(output_df))\n",
    "\n",
    "    # Prepare for next iteration\n",
    "    remaining_df = remaining_df.drop(new_output_df.index)\n",
    "    remaining_df = remaining_df[remaining_df.uniprot_id.isin(true_pos_proteins)==False]\n",
    "\n",
    "    print\n",
    "print 'Percent saved experiments:',(1 - np.true_divide(len(output_df),len(DF_retro_final)))*100,'%'\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}