{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature comparisons between allowing and excluding paths with duplicate nodes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/dhimmels/anaconda3/lib/python3.5/site-packages/matplotlib/__init__.py:872: UserWarning: axes.color_cycle is deprecated and replaced with axes.prop_cycle; please use the latter.\n",
      "  warnings.warn(self.msg_depr % (key, alt_key))\n"
     ]
    }
   ],
   "source": [
    "import re\n",
    "import functools\n",
    "\n",
    "import pandas\n",
    "import matplotlib.pyplot\n",
    "import seaborn\n",
    "import numpy\n",
    "import sklearn.metrics\n",
    "import qgrid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "seaborn.set_style('whitegrid')\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Read feature and partition data\n",
    "feature_df = pandas.read_table('features.tsv.gz', nrows=None)\n",
    "part_df = pandas.read_table('../data/partition.tsv.gz')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "@functools.lru_cache(maxsize=None)\n",
    "def duplicate_metanodes(metapath):\n",
    "    metanodes = re.split('[a-z<>]+', metapath)\n",
    "    return len(set(metanodes)) < len(metanodes)\n",
    "\n",
    "# Restrict to metapaths that have duplicate metanodes\n",
    "feature_df = feature_df[feature_df.metapath.map(duplicate_metanodes)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dfs = {}\n",
    "for key in 'PC', 'DWPC':\n",
    "    # Create spread dataframe\n",
    "    df = feature_df.pivot_table(values=key, index=['compound_id', 'disease_id', 'metapath'], columns='unique_nodes')\n",
    "    df = df.dropna() # use if only reading a part of `dwpc.tsv.gz`\n",
    "    \n",
    "    # Check that the two methods of duplicate node exclusion produce the same results\n",
    "    assert all(numpy.isclose(df.nested, df.expanded) & numpy.isclose(df.expanded, df.labeled))\n",
    "    \n",
    "    # Remove duplicate columns and rename columns\n",
    "    df = df.drop(['nested', 'expanded'], axis=1)\n",
    "    df = df.rename(columns={'False': key + '_dupl', 'labeled': key + '_nodupl'})\n",
    "    dfs[key] = df\n",
    "\n",
    "# Create a spread dataframe of Path Counts\n",
    "spread_df = pandas.concat(dfs.values(), axis=1).reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>unique_nodes</th>\n",
       "      <th>compound_id</th>\n",
       "      <th>disease_id</th>\n",
       "      <th>metapath</th>\n",
       "      <th>DWPC_dupl</th>\n",
       "      <th>DWPC_nodupl</th>\n",
       "      <th>PC_dupl</th>\n",
       "      <th>PC_nodupl</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>DB00091</td>\n",
       "      <td>DOID:1312</td>\n",
       "      <td>CbG&lt;kdG&lt;kdGaD</td>\n",
       "      <td>0.000601</td>\n",
       "      <td>0.000601</td>\n",
       "      <td>18</td>\n",
       "      <td>18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>DB00091</td>\n",
       "      <td>DOID:1312</td>\n",
       "      <td>CbG&lt;kdG&lt;kdGdD</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "unique_nodes compound_id disease_id       metapath  DWPC_dupl  DWPC_nodupl  \\\n",
       "0                DB00091  DOID:1312  CbG<kdG<kdGaD   0.000601     0.000601   \n",
       "1                DB00091  DOID:1312  CbG<kdG<kdGdD   0.000000     0.000000   \n",
       "\n",
       "unique_nodes  PC_dupl  PC_nodupl  \n",
       "0                  18         18  \n",
       "1                   0          0  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "spread_df.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# PC differences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.79% of path counts were affected by the unique node constraint'"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Filter for differing rows\n",
    "pct_disagree = 100 * (spread_df['PC_nodupl'] != spread_df['PC_dupl']).mean()\n",
    "'{:.02f}% of path counts were affected by the unique node constraint'.format(pct_disagree)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def metapath_summarize(df):\n",
    "    \"\"\"\n",
    "    Summarize a feature dataframe for a single metapath.\n",
    "    \"\"\"\n",
    "    row = pandas.Series()\n",
    "    \n",
    "    # Find instances were path counts are affected by duplicate node exlusion\n",
    "    disagree = df['PC_nodupl'] != df['PC_dupl']\n",
    "    \n",
    "    # Percent of compound-disease pairs (instances) that the unique node constraint has an effect\n",
    "    row['pct_affected'] = 100 * sum(disagree) / len(df)\n",
    "    \n",
    "    # Compare average path counts when allowing and excluding duplicate nodes\n",
    "    row['mean_PC_dupl'] = df['PC_dupl'].mean()\n",
    "    row['mean_PC_nodupl'] = df['PC_nodupl'].mean()\n",
    "    row['mean_pct_PC_increase'] = 100 * row['mean_PC_dupl'] / row['mean_PC_nodupl'] - 100\n",
    "    \n",
    "    # Calculate PC increases caused by allowing duplicate nodes seperately for indications and non-indications\n",
    "    neg_df = df[df.indication == 0]\n",
    "    pos_df = df[df.indication == 1]\n",
    "    row['mean_pct_PC_increase_nonind'] = 100 * neg_df['PC_dupl'].mean() / neg_df['PC_nodupl'].mean() - 100\n",
    "    row['mean_pct_PC_increase_ind'] = 100 * pos_df['PC_dupl'].mean() / pos_df['PC_nodupl'].mean() - 100\n",
    "    \n",
    "    # Calculate AUROCs of DWPCs when allowing and excluding duplicate nodes\n",
    "    row['auroc_dupl'] = sklearn.metrics.roc_auc_score(df['indication'], df['DWPC_dupl'])\n",
    "    row['auroc_nodupl'] = sklearn.metrics.roc_auc_score(df['indication'], df['DWPC_nodupl'])\n",
    "    row['auroc_diff'] = row['auroc_dupl'] - row['auroc_nodupl']\n",
    "    \n",
    "    return row"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1961"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Summarize features by metapath\n",
    "merged_df = part_df.merge(spread_df)\n",
    "metapath_df = merged_df.groupby('metapath').apply(metapath_summarize).fillna(0)\n",
    "metapath_df = metapath_df.sort_values('pct_affected', ascending=False)\n",
    "metapath_df['indication_metaedge'] = metapath_df.index.str.contains(r'CiD|DiC')\n",
    "metapath_df.reset_index().to_csv('metapaths.tsv', sep='\\t', index=False, float_format='%.3g')\n",
    "len(metapath_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "grid_options = {\n",
    "    'forceFitColumns': False,\n",
    "    'defaultColumnWidth': 200,\n",
    "    'editable': False,\n",
    "    'syncColumnCellResize': True,\n",
    "    'rowHeight': 28,\n",
    "}\n",
    "qgrid.show_grid(metapath_df, grid_options=grid_options, precision=2, remote_js=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'21.62% of metapaths had diminished AUROCs when exlcuding duplicate nodes.'"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "'{:.02f}% of metapaths had diminished AUROCs when exlcuding duplicate nodes.'.format(100 * (metapath_df.auroc_diff > 0).mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'100.00% of AUROC-suffering (> 0.005) metapaths contained an indication metaedge.'"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pct_suffer = metapath_df.query('auroc_diff > 0.005').indication_metaedge.mean() * 100\n",
    "'{:.02f}% of AUROC-suffering (> 0.005) metapaths contained an indication metaedge.'.format(pct_suffer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/dhimmels/anaconda3/lib/python3.5/site-packages/matplotlib/__init__.py:892: UserWarning: axes.color_cycle is deprecated and replaced with axes.prop_cycle; please use the latter.\n",
      "  warnings.warn(self.msg_depr % (key, alt_key))\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZUAAAEMCAYAAAAMMiuwAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSMn8fwRhqGynZBIeAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc887d86e48>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot distributions of decline in metapath AUROC from excluding duplicate-node paths\n",
    "ax = seaborn.violinplot('auroc_diff', 'indication_metaedge', data=metapath_df, orient='h',\n",
    "                        scale='width', cut=0, inner='quartile', linewidth=0.4)\n",
    "matplotlib.pyplot.xlabel('Decline in AUROC from excluding paths with duplicate nodes');\n",
    "matplotlib.pyplot.ylabel('Contains an indication metaedge');\n",
    "matplotlib.pyplot.savefig('AUROC-violins.png', dpi=200, bbox_inches='tight', pad_inches=0.03)"
   ]
  }
 ],
 "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": "3.5.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}