imputation-assess.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Assess the quality of L1000 imputation using genetic perturbation consensus signatures"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import pandas\n",
    "import numpy\n",
    "import matplotlib.pyplot\n",
    "import seaborn\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Read genes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>entrez_gene_id</th>\n",
       "      <th>status</th>\n",
       "      <th>symbol</th>\n",
       "      <th>type_of_gene</th>\n",
       "      <th>description</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>100</td>\n",
       "      <td>imputed</td>\n",
       "      <td>ADA</td>\n",
       "      <td>protein-coding</td>\n",
       "      <td>adenosine deaminase</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1000</td>\n",
       "      <td>imputed</td>\n",
       "      <td>CDH2</td>\n",
       "      <td>protein-coding</td>\n",
       "      <td>cadherin 2, type 1, N-cadherin (neuronal)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   entrez_gene_id   status symbol    type_of_gene  \\\n",
       "0             100  imputed    ADA  protein-coding   \n",
       "1            1000  imputed   CDH2  protein-coding   \n",
       "\n",
       "                                 description  \n",
       "0                        adenosine deaminase  \n",
       "1  cadherin 2, type 1, N-cadherin (neuronal)  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gene_df = pandas.read_table('data/consensi/genes.tsv')\n",
    "gene_df.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Read consensus signatures"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_tidied(path):\n",
    "    \"\"\"Read a dataframe of consensus signatures. Convert from matrix format to a tidy dataframe.\"\"\"\n",
    "    df = pandas.read_table(path)\n",
    "    df = pandas.melt(df, id_vars='perturbagen', var_name='entrez_gene_id', value_name='z_score')\n",
    "    for col in ['perturbagen', 'entrez_gene_id']:\n",
    "        df[col] = df[col].astype(int)\n",
    "    df = df.merge(gene_df[['entrez_gene_id', 'status']])\n",
    "    df['z_score_ihs'] = numpy.arcsinh(df['z_score'])\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "kd_df = get_tidied('data/consensi/consensi-knockdown.tsv.bz2')\n",
    "oe_df = get_tidied('data/consensi/consensi-overexpression.tsv.bz2')\n",
    "\n",
    "kd_df['pert_type'] = 'knockdown'\n",
    "oe_df['pert_type'] = 'overexpression'\n",
    "\n",
    "dysreg_df = pandas.concat([kd_df, oe_df])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>perturbagen</th>\n",
       "      <th>entrez_gene_id</th>\n",
       "      <th>z_score</th>\n",
       "      <th>status</th>\n",
       "      <th>z_score_ihs</th>\n",
       "      <th>pert_type</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2</td>\n",
       "      <td>100</td>\n",
       "      <td>-0.047</td>\n",
       "      <td>imputed</td>\n",
       "      <td>-0.046983</td>\n",
       "      <td>knockdown</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>9</td>\n",
       "      <td>100</td>\n",
       "      <td>-1.218</td>\n",
       "      <td>imputed</td>\n",
       "      <td>-1.027446</td>\n",
       "      <td>knockdown</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   perturbagen  entrez_gene_id  z_score   status  z_score_ihs  pert_type\n",
       "0            2             100   -0.047  imputed    -0.046983  knockdown\n",
       "1            9             100   -1.218  imputed    -1.027446  knockdown"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dysreg_df.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Perturbation-with-self z-scores "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "self_df = dysreg_df.query(\"perturbagen == entrez_gene_id\").copy()\n",
    "self_df['direction'] = self_df.z_score.map(lambda x: 'down' if x < 0 else 'up')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "seaborn.set(style='whitegrid')\n",
    "seaborn.set_style({'figure.figsize': (6, 3.7)})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZMAAAD7CAYAAACvzHniAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhLCYhIkQQgiL/T8iVP+FoR0tFwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7a275abeb8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax = seaborn.violinplot(x='z_score_ihs', y='status', data=self_df, inner='quartile', hue='pert_type', split=True)\n",
    "text = ax.set_xlabel(\"IHS-transformed z-score\")\n",
    "text = ax.set_ylabel('')\n",
    "ax.legend_.set_title('')\n",
    "matplotlib.pyplot.savefig('viz/self-targeting-perts.png', bbox_inches='tight', cut=0, dpi=300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>pert_type</th>\n",
       "      <th>knockdown</th>\n",
       "      <th>overexpression</th>\n",
       "      <th>All</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>status</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>imputed</th>\n",
       "      <td>2068</td>\n",
       "      <td>1108</td>\n",
       "      <td>3176</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>measured</th>\n",
       "      <td>960</td>\n",
       "      <td>486</td>\n",
       "      <td>1446</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>3028</td>\n",
       "      <td>1594</td>\n",
       "      <td>4622</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "pert_type  knockdown  overexpression   All\n",
       "status                                    \n",
       "imputed         2068            1108  3176\n",
       "measured         960             486  1446\n",
       "All             3028            1594  4622"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pandas.crosstab(self_df.status, self_df.pert_type, margins=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "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>pert_type</th>\n",
       "      <th>status</th>\n",
       "      <th>percent_up</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>knockdown</td>\n",
       "      <td>imputed</td>\n",
       "      <td>45.647969</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>knockdown</td>\n",
       "      <td>measured</td>\n",
       "      <td>2.916667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>overexpression</td>\n",
       "      <td>imputed</td>\n",
       "      <td>50.722022</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>overexpression</td>\n",
       "      <td>measured</td>\n",
       "      <td>63.991770</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        pert_type    status  percent_up\n",
       "0       knockdown   imputed   45.647969\n",
       "1       knockdown  measured    2.916667\n",
       "2  overexpression   imputed   50.722022\n",
       "3  overexpression  measured   63.991770"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Percent of perts that upregulate their target\n",
    "percent_df = 100 * self_df.groupby(['pert_type', 'status']).apply(lambda df: (df.z_score > 0).mean())\n",
    "percent_df.name = 'percent_up'\n",
    "percent_df = percent_df.reset_index()\n",
    "percent_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Significant genes for knockdowns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "measured    175879\n",
       "imputed     152189\n",
       "Name: status, dtype: int64"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "kd_sig_df = pandas.read_table('data/consensi/signif/dysreg-knockdown.tsv')\n",
    "kd_sig_df.status.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>direction</th>\n",
       "      <th>down</th>\n",
       "      <th>up</th>\n",
       "      <th>All</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>status</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>imputed</th>\n",
       "      <td>16</td>\n",
       "      <td>8</td>\n",
       "      <td>24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>measured</th>\n",
       "      <td>790</td>\n",
       "      <td>1</td>\n",
       "      <td>791</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>806</td>\n",
       "      <td>9</td>\n",
       "      <td>815</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "direction  down  up  All\n",
       "status                  \n",
       "imputed      16   8   24\n",
       "measured    790   1  791\n",
       "All         806   9  815"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "self_df = kd_sig_df.query(\"perturbagen == entrez_gene_id\")\n",
    "pandas.crosstab(self_df.status, self_df.direction, margins=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Significant genes for overexpressions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "measured    24333\n",
       "imputed     16309\n",
       "Name: status, dtype: int64"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "oe_sig_df = pandas.read_table('data/consensi/signif/dysreg-overexpression.tsv')\n",
    "oe_sig_df.status.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>direction</th>\n",
       "      <th>down</th>\n",
       "      <th>up</th>\n",
       "      <th>All</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>status</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>imputed</th>\n",
       "      <td>1</td>\n",
       "      <td>4</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>measured</th>\n",
       "      <td>3</td>\n",
       "      <td>120</td>\n",
       "      <td>123</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>All</th>\n",
       "      <td>4</td>\n",
       "      <td>124</td>\n",
       "      <td>128</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "direction  down   up  All\n",
       "status                   \n",
       "imputed       1    4    5\n",
       "measured      3  120  123\n",
       "All           4  124  128"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "self_df = oe_sig_df.query(\"perturbagen == entrez_gene_id\")\n",
    "pandas.crosstab(self_df.status, self_df.direction, margins=True)"
   ]
  }
 ],
 "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
}