examples/Inferring a 10-node network.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Inferring a 10-node network\n",
    "==============="
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "from Swing import Swing\n",
    "import pandas as pd\n",
    "\n",
    "#optionally, for visualization\n",
    "import networkx as nx\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "pd.set_option('display.max_rows', 10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SWING (Sliding Window Inference for Network Generation) is a network inference tool implemented in python that incorporates elements from multivariate Granger causality and sliding window regression. This tutorial demonstrates how to infer a 10-node network using SWING-RF (RandomForest)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SWING has user-selected parameters that require knowledge of the system of interest. A more detailed discussion about setting these parameters and sensitivity of parameters is available in the Supporting Information of *Finkle et al (2018).*\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 0: examine/pre-process the input data-file."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SWING requires a $(m+1)$ x $(n+1)$ time-series gene expression dataset, where $m$ refers to the number of timepoints associated with each perturbation which populate each row, and $n$ refers to the number of genes which refers to each gene. It is expected that the first row will be the header row specifying the column/gene name labels (hence the +1). The first column should specify the time-point labels.\n",
    "\n",
    "Our *in silico* dataset has 10 different sets of time-series for which a perturbation was applied and removed. Each time-series experiment has 21 timepoints. Gene expression was measured for 10 genes at 50 min intervals for 1000 min."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Time</th>\n",
       "      <th>G1</th>\n",
       "      <th>G2</th>\n",
       "      <th>G3</th>\n",
       "      <th>G4</th>\n",
       "      <th>G5</th>\n",
       "      <th>G6</th>\n",
       "      <th>G7</th>\n",
       "      <th>G8</th>\n",
       "      <th>G9</th>\n",
       "      <th>G10</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0.089284</td>\n",
       "      <td>0.270871</td>\n",
       "      <td>0.839036</td>\n",
       "      <td>0.200885</td>\n",
       "      <td>0.182632</td>\n",
       "      <td>0.356788</td>\n",
       "      <td>0.006343</td>\n",
       "      <td>0.599246</td>\n",
       "      <td>0.019762</td>\n",
       "      <td>0.033317</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>50</td>\n",
       "      <td>0.114581</td>\n",
       "      <td>0.318827</td>\n",
       "      <td>0.693707</td>\n",
       "      <td>0.142432</td>\n",
       "      <td>0.261206</td>\n",
       "      <td>0.198684</td>\n",
       "      <td>0.034523</td>\n",
       "      <td>0.640026</td>\n",
       "      <td>0.007827</td>\n",
       "      <td>0.034452</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>100</td>\n",
       "      <td>0.112723</td>\n",
       "      <td>0.393308</td>\n",
       "      <td>0.738423</td>\n",
       "      <td>0.073780</td>\n",
       "      <td>0.265523</td>\n",
       "      <td>0.057642</td>\n",
       "      <td>0.019089</td>\n",
       "      <td>0.664219</td>\n",
       "      <td>0.024778</td>\n",
       "      <td>0.047530</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>150</td>\n",
       "      <td>0.047108</td>\n",
       "      <td>0.369182</td>\n",
       "      <td>0.758015</td>\n",
       "      <td>0.066540</td>\n",
       "      <td>0.211461</td>\n",
       "      <td>0.015320</td>\n",
       "      <td>0.055761</td>\n",
       "      <td>0.627534</td>\n",
       "      <td>0.020925</td>\n",
       "      <td>0.025798</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>200</td>\n",
       "      <td>0.061564</td>\n",
       "      <td>0.342303</td>\n",
       "      <td>0.787797</td>\n",
       "      <td>0.082469</td>\n",
       "      <td>0.244511</td>\n",
       "      <td>0.004667</td>\n",
       "      <td>0.018079</td>\n",
       "      <td>0.713676</td>\n",
       "      <td>0.031663</td>\n",
       "      <td>0.031388</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Time        G1        G2        G3        G4        G5        G6        G7  \\\n",
       "0     0  0.089284  0.270871  0.839036  0.200885  0.182632  0.356788  0.006343   \n",
       "1    50  0.114581  0.318827  0.693707  0.142432  0.261206  0.198684  0.034523   \n",
       "2   100  0.112723  0.393308  0.738423  0.073780  0.265523  0.057642  0.019089   \n",
       "3   150  0.047108  0.369182  0.758015  0.066540  0.211461  0.015320  0.055761   \n",
       "4   200  0.061564  0.342303  0.787797  0.082469  0.244511  0.004667  0.018079   \n",
       "\n",
       "         G8        G9       G10  \n",
       "0  0.599246  0.019762  0.033317  \n",
       "1  0.640026  0.007827  0.034452  \n",
       "2  0.664219  0.024778  0.047530  \n",
       "3  0.627534  0.020925  0.025798  \n",
       "4  0.713676  0.031663  0.031388  "
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset = pd.read_csv('../data/Ecoli_10_node/Ecoli10-1_dream4_timeseries.tsv', sep='\\t')\n",
    "dataset.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(210, 11)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 210 rows = 10 experiments * 21 timepoints each\n",
    "# 10 gene columns + 1 time column\n",
    "dataset.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 1: initialize a SWING object. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**To generate a SWING object, we will specify the following parameters:**\n",
    "\n",
    "- Inference-related parameters:\n",
    "  - *min_lag* and *max_lag*: *min_lag* and *max_lag* refers to the allowed delay range (also known as $k_{min}$ and $k_{max}$). If *min_lag* = 1 and *max_lag* = 3, we are assuming the responses are delayed by at least one timepoint and at most 3 timepoints. We recommend you set values for *min_lag* and *max_lag* using prior knowledge or delay analysis using a subset of experimentally validated edges.  \n",
    "  - *window_width*: the window size (also known as *w*). This value should be $1<w \\leq T$ where $T$ refers to the number of timepoints in each perturbation. If *w* = $T$, then only 1 window will be created.\n",
    "  - *window_type*: the embedded inference algorithm. Currently the three implemented algorithms are LASSO, RF, and PLSR. Depending on the algorithm you choose, you will also specify additional parameters.\n",
    "- Data-parsing related parameters:\n",
    "  - *gene_start_column*: the position of the first column with numerical expression data. This is 0-indexed, so if your first column denotes the timepoint labels and the second column contains numerical data, the gene_start_column should be 1.\n",
    "  - *time_label*: the label of the time-point labels.\n",
    "  - *separator*: the delimiter to use.\n",
    "  - *gene_end*: the position of the last gene (0-indexed).\n",
    "  - *file_path*: the file_path of the time-series data-file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# parsing-related parameters\n",
    "gene_start_column = 1\n",
    "time_label = \"Time\"\n",
    "separator = \"\\t\"\n",
    "gene_end = None\n",
    "file_path = \"../data/Ecoli_10_node/Ecoli10-1_dream4_timeseries.tsv\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# inference related parameters\n",
    "# k_min and k_max was determined using our cross-correlation analysis detailed in Finkle et al 2018.\n",
    "# Sensitivity analysis of these parameters are also available in Supporting Information of Finkle et al 2018.\n",
    "k_min = 1\n",
    "k_max = 3\n",
    "w = 10\n",
    "method = 'RandomForest'\n",
    "\n",
    "# random-forest specific parameter\n",
    "trees = 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sg = Swing(file_path, gene_start_column, gene_end, time_label, separator, min_lag=k_min, max_lag=k_max, window_width=w, window_type=method)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Time</th>\n",
       "      <th>G1</th>\n",
       "      <th>G2</th>\n",
       "      <th>G3</th>\n",
       "      <th>G4</th>\n",
       "      <th>G5</th>\n",
       "      <th>G6</th>\n",
       "      <th>G7</th>\n",
       "      <th>G8</th>\n",
       "      <th>G9</th>\n",
       "      <th>G10</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>-0.563726</td>\n",
       "      <td>-0.606203</td>\n",
       "      <td>1.228299</td>\n",
       "      <td>-0.558114</td>\n",
       "      <td>-0.922628</td>\n",
       "      <td>0.041356</td>\n",
       "      <td>-0.527995</td>\n",
       "      <td>-0.272521</td>\n",
       "      <td>-0.479344</td>\n",
       "      <td>-0.671869</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>50</td>\n",
       "      <td>0.024698</td>\n",
       "      <td>-0.223974</td>\n",
       "      <td>0.046455</td>\n",
       "      <td>-1.179870</td>\n",
       "      <td>-0.039325</td>\n",
       "      <td>-1.051685</td>\n",
       "      <td>-0.336545</td>\n",
       "      <td>0.221865</td>\n",
       "      <td>-1.159974</td>\n",
       "      <td>-0.630530</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>100</td>\n",
       "      <td>-0.018525</td>\n",
       "      <td>0.369675</td>\n",
       "      <td>0.410093</td>\n",
       "      <td>-1.910127</td>\n",
       "      <td>0.009209</td>\n",
       "      <td>-2.026762</td>\n",
       "      <td>-0.441405</td>\n",
       "      <td>0.515168</td>\n",
       "      <td>-0.193301</td>\n",
       "      <td>-0.154017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>150</td>\n",
       "      <td>-1.544745</td>\n",
       "      <td>0.177383</td>\n",
       "      <td>0.569417</td>\n",
       "      <td>-1.987136</td>\n",
       "      <td>-0.598544</td>\n",
       "      <td>-2.319353</td>\n",
       "      <td>-0.192261</td>\n",
       "      <td>0.070426</td>\n",
       "      <td>-0.412980</td>\n",
       "      <td>-0.945839</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>200</td>\n",
       "      <td>-1.208491</td>\n",
       "      <td>-0.036860</td>\n",
       "      <td>0.811614</td>\n",
       "      <td>-1.817705</td>\n",
       "      <td>-0.227007</td>\n",
       "      <td>-2.393001</td>\n",
       "      <td>-0.448262</td>\n",
       "      <td>1.114764</td>\n",
       "      <td>0.199386</td>\n",
       "      <td>-0.742136</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>205</th>\n",
       "      <td>800</td>\n",
       "      <td>-0.489930</td>\n",
       "      <td>0.422353</td>\n",
       "      <td>0.419013</td>\n",
       "      <td>0.827494</td>\n",
       "      <td>0.415479</td>\n",
       "      <td>0.342655</td>\n",
       "      <td>-0.259271</td>\n",
       "      <td>0.875540</td>\n",
       "      <td>0.374526</td>\n",
       "      <td>-0.546858</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>206</th>\n",
       "      <td>850</td>\n",
       "      <td>-0.685166</td>\n",
       "      <td>0.071311</td>\n",
       "      <td>-0.398351</td>\n",
       "      <td>0.324308</td>\n",
       "      <td>-0.219049</td>\n",
       "      <td>0.350101</td>\n",
       "      <td>-0.373390</td>\n",
       "      <td>-0.345331</td>\n",
       "      <td>-0.173951</td>\n",
       "      <td>-0.349612</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>207</th>\n",
       "      <td>900</td>\n",
       "      <td>-0.381753</td>\n",
       "      <td>-0.259554</td>\n",
       "      <td>0.182480</td>\n",
       "      <td>0.434024</td>\n",
       "      <td>-0.307650</td>\n",
       "      <td>-0.189032</td>\n",
       "      <td>-0.455310</td>\n",
       "      <td>-0.919943</td>\n",
       "      <td>-0.398313</td>\n",
       "      <td>0.568248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>208</th>\n",
       "      <td>950</td>\n",
       "      <td>-0.530694</td>\n",
       "      <td>-1.121127</td>\n",
       "      <td>1.611699</td>\n",
       "      <td>0.949935</td>\n",
       "      <td>-0.401332</td>\n",
       "      <td>0.167607</td>\n",
       "      <td>-0.462697</td>\n",
       "      <td>-0.813640</td>\n",
       "      <td>-0.423046</td>\n",
       "      <td>-0.836947</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>209</th>\n",
       "      <td>1000</td>\n",
       "      <td>-0.136001</td>\n",
       "      <td>-0.587548</td>\n",
       "      <td>0.719714</td>\n",
       "      <td>1.315835</td>\n",
       "      <td>-0.677365</td>\n",
       "      <td>-0.372990</td>\n",
       "      <td>-0.434845</td>\n",
       "      <td>1.739931</td>\n",
       "      <td>-0.558551</td>\n",
       "      <td>-0.964657</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>210 rows × 11 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     Time        G1        G2        G3        G4        G5        G6  \\\n",
       "0       0 -0.563726 -0.606203  1.228299 -0.558114 -0.922628  0.041356   \n",
       "1      50  0.024698 -0.223974  0.046455 -1.179870 -0.039325 -1.051685   \n",
       "2     100 -0.018525  0.369675  0.410093 -1.910127  0.009209 -2.026762   \n",
       "3     150 -1.544745  0.177383  0.569417 -1.987136 -0.598544 -2.319353   \n",
       "4     200 -1.208491 -0.036860  0.811614 -1.817705 -0.227007 -2.393001   \n",
       "..    ...       ...       ...       ...       ...       ...       ...   \n",
       "205   800 -0.489930  0.422353  0.419013  0.827494  0.415479  0.342655   \n",
       "206   850 -0.685166  0.071311 -0.398351  0.324308 -0.219049  0.350101   \n",
       "207   900 -0.381753 -0.259554  0.182480  0.434024 -0.307650 -0.189032   \n",
       "208   950 -0.530694 -1.121127  1.611699  0.949935 -0.401332  0.167607   \n",
       "209  1000 -0.136001 -0.587548  0.719714  1.315835 -0.677365 -0.372990   \n",
       "\n",
       "           G7        G8        G9       G10  \n",
       "0   -0.527995 -0.272521 -0.479344 -0.671869  \n",
       "1   -0.336545  0.221865 -1.159974 -0.630530  \n",
       "2   -0.441405  0.515168 -0.193301 -0.154017  \n",
       "3   -0.192261  0.070426 -0.412980 -0.945839  \n",
       "4   -0.448262  1.114764  0.199386 -0.742136  \n",
       "..        ...       ...       ...       ...  \n",
       "205 -0.259271  0.875540  0.374526 -0.546858  \n",
       "206 -0.373390 -0.345331 -0.173951 -0.349612  \n",
       "207 -0.455310 -0.919943 -0.398313  0.568248  \n",
       "208 -0.462697 -0.813640 -0.423046 -0.836947  \n",
       "209 -0.434845  1.739931 -0.558551 -0.964657  \n",
       "\n",
       "[210 rows x 11 columns]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Optionally, z-score the data. zscore_all_data() normalizes all the data by column (by gene).  \n",
    "sg.zscore_all_data()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 2: Partition data into windows."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Using the parameters we defined earlier (window_width, min_lag, max_lag...), we will partition the data into time-resolved subsets of data or \"windows\" using create_windows(). Our time-series dataset has 21 time-points and 10 perturbations. We specified a window size of 10, so total number of windows = (21 - 10 +1) = 12. \n",
    "\n",
    "SWING generates a number of models using windowed data. These models built with subsets of the data are later aggregated into a single model. Each model has a single response window (consisting of a subset of a time-series) and multiple explanatory windows (consisting of subsets of the explanatory data). \n",
    "\n",
    "Since we specified that min_lag = 1, and max_lag = 3, each response window is expected to have 3 explanatory windows that are lagging by 1 time-point, 2 time-points, and 3 time-points. **The first 3 response windows will not have sufficient explanatory data, and are removed from the window_list.**\n",
    "\n",
    "Also, since $w=10$, it is expected that each window of data, $w_q$, will have 10 timepoints from each experiment."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'time_label': 'Time', 'gene_start': 1, 'gene_end': 11, 'nth_window': 3}\n",
      "{'time_label': 'Time', 'gene_start': 1, 'gene_end': 11, 'nth_window': 4}\n",
      "{'time_label': 'Time', 'gene_start': 1, 'gene_end': 11, 'nth_window': 5}\n",
      "{'time_label': 'Time', 'gene_start': 1, 'gene_end': 11, 'nth_window': 6}\n",
      "{'time_label': 'Time', 'gene_start': 1, 'gene_end': 11, 'nth_window': 7}\n",
      "{'time_label': 'Time', 'gene_start': 1, 'gene_end': 11, 'nth_window': 8}\n",
      "{'time_label': 'Time', 'gene_start': 1, 'gene_end': 11, 'nth_window': 9}\n",
      "{'time_label': 'Time', 'gene_start': 1, 'gene_end': 11, 'nth_window': 10}\n",
      "{'time_label': 'Time', 'gene_start': 1, 'gene_end': 11, 'nth_window': 11}\n"
     ]
    }
   ],
   "source": [
    "sg.create_windows()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Swing.RFRWindow.RandomForestRegressionWindow at 0x110b085f8>"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# window_list denotes only models/response windows with explanatory data\n",
    "sg.window_list[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0, 1, 2]"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sg.window_list[0].earlier_windows"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 3: Initialize/Optimize Parameters for SWING."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For SWING-RF, this step will set the number of trees for each window to its default parameter. For other embedded algorithms such as SWING-PLSR, it will estimate optimal parameters for the inference algorithm of interest in the context of the data. This is discussed later in the tutorial."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<Swing.RFRWindow.RandomForestRegressionWindow at 0x110b085f8>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a408c88>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a4087f0>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x110b084e0>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a408be0>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x110b08198>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a430da0>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a430630>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a42f198>]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sg.optimize_params()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "500"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sg.window_list[0].n_trees"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 4: Estimate edge importance for each window model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Edge importance is estimated for each model using RandomForest and fit_windows(). Each model will output an estimated importance matrix denoting the importance or stability of each edge. For example, in SWING-RandomForest, the embedded inference algorithm is RandomForestRegression from scikit-learn, and feature\\_importances\\_ will be estimated for each edge."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<Swing.RFRWindow.RandomForestRegressionWindow at 0x110b085f8>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a408c88>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a4087f0>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x110b084e0>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a408be0>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x110b08198>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a430da0>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a430630>,\n",
       " <Swing.RFRWindow.RandomForestRegressionWindow at 0x10a42f198>]"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#For the tutorial, we set the number of trees to 100. \n",
    "sg.fit_windows(n_trees=trees, show_progress=False, n_jobs=-1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For each model, the feature importances between the explanatory data (G1-G10 from 3 different time-delayed windows; columns) to response data (G1-G10 response window; rows) are scored. Below is the raw version of the feature matrices we generate. We will compile the raw data into one dataframe in the next step."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>Parent</th>\n",
       "      <th>G1</th>\n",
       "      <th>G2</th>\n",
       "      <th>G3</th>\n",
       "      <th>G4</th>\n",
       "      <th>G5</th>\n",
       "      <th>G6</th>\n",
       "      <th>G7</th>\n",
       "      <th>G8</th>\n",
       "      <th>G9</th>\n",
       "      <th>G10</th>\n",
       "      <th>...</th>\n",
       "      <th>G1</th>\n",
       "      <th>G2</th>\n",
       "      <th>G3</th>\n",
       "      <th>G4</th>\n",
       "      <th>G5</th>\n",
       "      <th>G6</th>\n",
       "      <th>G7</th>\n",
       "      <th>G8</th>\n",
       "      <th>G9</th>\n",
       "      <th>G10</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Child</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>G1</th>\n",
       "      <td>0.057491</td>\n",
       "      <td>0.051730</td>\n",
       "      <td>0.046511</td>\n",
       "      <td>0.024360</td>\n",
       "      <td>0.008360</td>\n",
       "      <td>0.017964</td>\n",
       "      <td>0.019451</td>\n",
       "      <td>0.015411</td>\n",
       "      <td>0.015400</td>\n",
       "      <td>0.013775</td>\n",
       "      <td>...</td>\n",
       "      <td>0.112792</td>\n",
       "      <td>0.123032</td>\n",
       "      <td>0.055853</td>\n",
       "      <td>0.016534</td>\n",
       "      <td>0.016658</td>\n",
       "      <td>0.023360</td>\n",
       "      <td>0.021795</td>\n",
       "      <td>0.014729</td>\n",
       "      <td>0.008505</td>\n",
       "      <td>0.031588</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G2</th>\n",
       "      <td>0.028834</td>\n",
       "      <td>0.057703</td>\n",
       "      <td>0.067596</td>\n",
       "      <td>0.010202</td>\n",
       "      <td>0.004084</td>\n",
       "      <td>0.006058</td>\n",
       "      <td>0.006871</td>\n",
       "      <td>0.009097</td>\n",
       "      <td>0.003322</td>\n",
       "      <td>0.010256</td>\n",
       "      <td>...</td>\n",
       "      <td>0.049620</td>\n",
       "      <td>0.159983</td>\n",
       "      <td>0.147231</td>\n",
       "      <td>0.023026</td>\n",
       "      <td>0.009959</td>\n",
       "      <td>0.005066</td>\n",
       "      <td>0.029393</td>\n",
       "      <td>0.045240</td>\n",
       "      <td>0.010725</td>\n",
       "      <td>0.026582</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G3</th>\n",
       "      <td>0.021176</td>\n",
       "      <td>0.045878</td>\n",
       "      <td>0.063680</td>\n",
       "      <td>0.020412</td>\n",
       "      <td>0.017394</td>\n",
       "      <td>0.012928</td>\n",
       "      <td>0.009733</td>\n",
       "      <td>0.012828</td>\n",
       "      <td>0.012199</td>\n",
       "      <td>0.024435</td>\n",
       "      <td>...</td>\n",
       "      <td>0.023362</td>\n",
       "      <td>0.081976</td>\n",
       "      <td>0.170403</td>\n",
       "      <td>0.028628</td>\n",
       "      <td>0.036218</td>\n",
       "      <td>0.010672</td>\n",
       "      <td>0.013900</td>\n",
       "      <td>0.021272</td>\n",
       "      <td>0.026817</td>\n",
       "      <td>0.023956</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G4</th>\n",
       "      <td>0.014934</td>\n",
       "      <td>0.018285</td>\n",
       "      <td>0.011634</td>\n",
       "      <td>0.110901</td>\n",
       "      <td>0.009717</td>\n",
       "      <td>0.013879</td>\n",
       "      <td>0.012894</td>\n",
       "      <td>0.010235</td>\n",
       "      <td>0.015536</td>\n",
       "      <td>0.009567</td>\n",
       "      <td>...</td>\n",
       "      <td>0.019519</td>\n",
       "      <td>0.030243</td>\n",
       "      <td>0.017707</td>\n",
       "      <td>0.230668</td>\n",
       "      <td>0.015440</td>\n",
       "      <td>0.033015</td>\n",
       "      <td>0.029873</td>\n",
       "      <td>0.024325</td>\n",
       "      <td>0.015402</td>\n",
       "      <td>0.017198</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G5</th>\n",
       "      <td>0.017698</td>\n",
       "      <td>0.018827</td>\n",
       "      <td>0.033501</td>\n",
       "      <td>0.022312</td>\n",
       "      <td>0.082914</td>\n",
       "      <td>0.019614</td>\n",
       "      <td>0.014084</td>\n",
       "      <td>0.016332</td>\n",
       "      <td>0.018230</td>\n",
       "      <td>0.013053</td>\n",
       "      <td>...</td>\n",
       "      <td>0.018324</td>\n",
       "      <td>0.037160</td>\n",
       "      <td>0.062658</td>\n",
       "      <td>0.016059</td>\n",
       "      <td>0.168748</td>\n",
       "      <td>0.042584</td>\n",
       "      <td>0.021392</td>\n",
       "      <td>0.019887</td>\n",
       "      <td>0.024674</td>\n",
       "      <td>0.025721</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G6</th>\n",
       "      <td>0.008725</td>\n",
       "      <td>0.013174</td>\n",
       "      <td>0.019987</td>\n",
       "      <td>0.059764</td>\n",
       "      <td>0.009588</td>\n",
       "      <td>0.128521</td>\n",
       "      <td>0.007883</td>\n",
       "      <td>0.006747</td>\n",
       "      <td>0.007073</td>\n",
       "      <td>0.013980</td>\n",
       "      <td>...</td>\n",
       "      <td>0.016310</td>\n",
       "      <td>0.029650</td>\n",
       "      <td>0.025345</td>\n",
       "      <td>0.037605</td>\n",
       "      <td>0.020518</td>\n",
       "      <td>0.178182</td>\n",
       "      <td>0.013880</td>\n",
       "      <td>0.014942</td>\n",
       "      <td>0.007361</td>\n",
       "      <td>0.017907</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G7</th>\n",
       "      <td>0.032306</td>\n",
       "      <td>0.010526</td>\n",
       "      <td>0.002908</td>\n",
       "      <td>0.018793</td>\n",
       "      <td>0.007122</td>\n",
       "      <td>0.022690</td>\n",
       "      <td>0.143754</td>\n",
       "      <td>0.007218</td>\n",
       "      <td>0.004980</td>\n",
       "      <td>0.004311</td>\n",
       "      <td>...</td>\n",
       "      <td>0.060794</td>\n",
       "      <td>0.048361</td>\n",
       "      <td>0.010496</td>\n",
       "      <td>0.015941</td>\n",
       "      <td>0.005761</td>\n",
       "      <td>0.013621</td>\n",
       "      <td>0.188139</td>\n",
       "      <td>0.009591</td>\n",
       "      <td>0.007005</td>\n",
       "      <td>0.012487</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G8</th>\n",
       "      <td>0.035626</td>\n",
       "      <td>0.033302</td>\n",
       "      <td>0.014237</td>\n",
       "      <td>0.046981</td>\n",
       "      <td>0.016023</td>\n",
       "      <td>0.017924</td>\n",
       "      <td>0.045028</td>\n",
       "      <td>0.014618</td>\n",
       "      <td>0.014955</td>\n",
       "      <td>0.019238</td>\n",
       "      <td>...</td>\n",
       "      <td>0.047666</td>\n",
       "      <td>0.052577</td>\n",
       "      <td>0.050500</td>\n",
       "      <td>0.032443</td>\n",
       "      <td>0.012115</td>\n",
       "      <td>0.015041</td>\n",
       "      <td>0.084575</td>\n",
       "      <td>0.051005</td>\n",
       "      <td>0.033405</td>\n",
       "      <td>0.015968</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G9</th>\n",
       "      <td>0.013409</td>\n",
       "      <td>0.017490</td>\n",
       "      <td>0.033879</td>\n",
       "      <td>0.023737</td>\n",
       "      <td>0.014008</td>\n",
       "      <td>0.011351</td>\n",
       "      <td>0.011467</td>\n",
       "      <td>0.010007</td>\n",
       "      <td>0.077737</td>\n",
       "      <td>0.041083</td>\n",
       "      <td>...</td>\n",
       "      <td>0.017052</td>\n",
       "      <td>0.035045</td>\n",
       "      <td>0.015049</td>\n",
       "      <td>0.016145</td>\n",
       "      <td>0.060643</td>\n",
       "      <td>0.015528</td>\n",
       "      <td>0.014198</td>\n",
       "      <td>0.014241</td>\n",
       "      <td>0.138755</td>\n",
       "      <td>0.054114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>G10</th>\n",
       "      <td>0.008492</td>\n",
       "      <td>0.094848</td>\n",
       "      <td>0.088647</td>\n",
       "      <td>0.020975</td>\n",
       "      <td>0.017107</td>\n",
       "      <td>0.007255</td>\n",
       "      <td>0.009332</td>\n",
       "      <td>0.010186</td>\n",
       "      <td>0.023737</td>\n",
       "      <td>0.034704</td>\n",
       "      <td>...</td>\n",
       "      <td>0.017213</td>\n",
       "      <td>0.082623</td>\n",
       "      <td>0.050579</td>\n",
       "      <td>0.014354</td>\n",
       "      <td>0.016997</td>\n",
       "      <td>0.017515</td>\n",
       "      <td>0.015202</td>\n",
       "      <td>0.006767</td>\n",
       "      <td>0.030957</td>\n",
       "      <td>0.107517</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>10 rows × 30 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "Parent        G1        G2        G3        G4        G5        G6        G7  \\\n",
       "Child                                                                          \n",
       "G1      0.057491  0.051730  0.046511  0.024360  0.008360  0.017964  0.019451   \n",
       "G2      0.028834  0.057703  0.067596  0.010202  0.004084  0.006058  0.006871   \n",
       "G3      0.021176  0.045878  0.063680  0.020412  0.017394  0.012928  0.009733   \n",
       "G4      0.014934  0.018285  0.011634  0.110901  0.009717  0.013879  0.012894   \n",
       "G5      0.017698  0.018827  0.033501  0.022312  0.082914  0.019614  0.014084   \n",
       "G6      0.008725  0.013174  0.019987  0.059764  0.009588  0.128521  0.007883   \n",
       "G7      0.032306  0.010526  0.002908  0.018793  0.007122  0.022690  0.143754   \n",
       "G8      0.035626  0.033302  0.014237  0.046981  0.016023  0.017924  0.045028   \n",
       "G9      0.013409  0.017490  0.033879  0.023737  0.014008  0.011351  0.011467   \n",
       "G10     0.008492  0.094848  0.088647  0.020975  0.017107  0.007255  0.009332   \n",
       "\n",
       "Parent        G8        G9       G10    ...           G1        G2        G3  \\\n",
       "Child                                   ...                                    \n",
       "G1      0.015411  0.015400  0.013775    ...     0.112792  0.123032  0.055853   \n",
       "G2      0.009097  0.003322  0.010256    ...     0.049620  0.159983  0.147231   \n",
       "G3      0.012828  0.012199  0.024435    ...     0.023362  0.081976  0.170403   \n",
       "G4      0.010235  0.015536  0.009567    ...     0.019519  0.030243  0.017707   \n",
       "G5      0.016332  0.018230  0.013053    ...     0.018324  0.037160  0.062658   \n",
       "G6      0.006747  0.007073  0.013980    ...     0.016310  0.029650  0.025345   \n",
       "G7      0.007218  0.004980  0.004311    ...     0.060794  0.048361  0.010496   \n",
       "G8      0.014618  0.014955  0.019238    ...     0.047666  0.052577  0.050500   \n",
       "G9      0.010007  0.077737  0.041083    ...     0.017052  0.035045  0.015049   \n",
       "G10     0.010186  0.023737  0.034704    ...     0.017213  0.082623  0.050579   \n",
       "\n",
       "Parent        G4        G5        G6        G7        G8        G9       G10  \n",
       "Child                                                                         \n",
       "G1      0.016534  0.016658  0.023360  0.021795  0.014729  0.008505  0.031588  \n",
       "G2      0.023026  0.009959  0.005066  0.029393  0.045240  0.010725  0.026582  \n",
       "G3      0.028628  0.036218  0.010672  0.013900  0.021272  0.026817  0.023956  \n",
       "G4      0.230668  0.015440  0.033015  0.029873  0.024325  0.015402  0.017198  \n",
       "G5      0.016059  0.168748  0.042584  0.021392  0.019887  0.024674  0.025721  \n",
       "G6      0.037605  0.020518  0.178182  0.013880  0.014942  0.007361  0.017907  \n",
       "G7      0.015941  0.005761  0.013621  0.188139  0.009591  0.007005  0.012487  \n",
       "G8      0.032443  0.012115  0.015041  0.084575  0.051005  0.033405  0.015968  \n",
       "G9      0.016145  0.060643  0.015528  0.014198  0.014241  0.138755  0.054114  \n",
       "G10     0.014354  0.016997  0.017515  0.015202  0.006767  0.030957  0.107517  \n",
       "\n",
       "[10 rows x 30 columns]"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sg.window_list[0].edge_importance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 5: Compile the all edges into one table"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*compile_edges()* generates an aggregate edge table that summarizes the properties of each edge of each window."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Compiling all model edges...[DONE]\n"
     ]
    }
   ],
   "source": [
    "sg.compile_edges()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Parent</th>\n",
       "      <th>Child</th>\n",
       "      <th>Importance</th>\n",
       "      <th>P_window</th>\n",
       "      <th>C_window</th>\n",
       "      <th>adj_imp</th>\n",
       "      <th>Rank</th>\n",
       "      <th>Edge</th>\n",
       "      <th>Lag</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>G3</td>\n",
       "      <td>G2</td>\n",
       "      <td>0.147231</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>0.147231</td>\n",
       "      <td>9</td>\n",
       "      <td>(G3, G2)</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>G2</td>\n",
       "      <td>G1</td>\n",
       "      <td>0.123032</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>0.123032</td>\n",
       "      <td>14</td>\n",
       "      <td>(G2, G1)</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>G2</td>\n",
       "      <td>G10</td>\n",
       "      <td>0.094848</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0.094848</td>\n",
       "      <td>19</td>\n",
       "      <td>(G2, G10)</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>G3</td>\n",
       "      <td>G2</td>\n",
       "      <td>0.093540</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>0.093540</td>\n",
       "      <td>20</td>\n",
       "      <td>(G3, G2)</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>G3</td>\n",
       "      <td>G10</td>\n",
       "      <td>0.088647</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0.088647</td>\n",
       "      <td>22</td>\n",
       "      <td>(G3, G10)</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2695</th>\n",
       "      <td>G6</td>\n",
       "      <td>G7</td>\n",
       "      <td>0.010959</td>\n",
       "      <td>9</td>\n",
       "      <td>11</td>\n",
       "      <td>0.010959</td>\n",
       "      <td>295</td>\n",
       "      <td>(G6, G7)</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2696</th>\n",
       "      <td>G3</td>\n",
       "      <td>G7</td>\n",
       "      <td>0.010084</td>\n",
       "      <td>10</td>\n",
       "      <td>11</td>\n",
       "      <td>0.010084</td>\n",
       "      <td>296</td>\n",
       "      <td>(G3, G7)</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2697</th>\n",
       "      <td>G3</td>\n",
       "      <td>G7</td>\n",
       "      <td>0.010048</td>\n",
       "      <td>8</td>\n",
       "      <td>11</td>\n",
       "      <td>0.010048</td>\n",
       "      <td>297</td>\n",
       "      <td>(G3, G7)</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2698</th>\n",
       "      <td>G5</td>\n",
       "      <td>G7</td>\n",
       "      <td>0.010044</td>\n",
       "      <td>10</td>\n",
       "      <td>11</td>\n",
       "      <td>0.010044</td>\n",
       "      <td>298</td>\n",
       "      <td>(G5, G7)</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2699</th>\n",
       "      <td>G10</td>\n",
       "      <td>G7</td>\n",
       "      <td>0.009998</td>\n",
       "      <td>9</td>\n",
       "      <td>11</td>\n",
       "      <td>0.009998</td>\n",
       "      <td>299</td>\n",
       "      <td>(G10, G7)</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2430 rows × 9 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     Parent Child  Importance  P_window  C_window   adj_imp  Rank       Edge  \\\n",
       "9        G3    G2    0.147231         2         3  0.147231     9   (G3, G2)   \n",
       "14       G2    G1    0.123032         2         3  0.123032    14   (G2, G1)   \n",
       "19       G2   G10    0.094848         0         3  0.094848    19  (G2, G10)   \n",
       "20       G3    G2    0.093540         1         3  0.093540    20   (G3, G2)   \n",
       "22       G3   G10    0.088647         0         3  0.088647    22  (G3, G10)   \n",
       "...     ...   ...         ...       ...       ...       ...   ...        ...   \n",
       "2695     G6    G7    0.010959         9        11  0.010959   295   (G6, G7)   \n",
       "2696     G3    G7    0.010084        10        11  0.010084   296   (G3, G7)   \n",
       "2697     G3    G7    0.010048         8        11  0.010048   297   (G3, G7)   \n",
       "2698     G5    G7    0.010044        10        11  0.010044   298   (G5, G7)   \n",
       "2699    G10    G7    0.009998         9        11  0.009998   299  (G10, G7)   \n",
       "\n",
       "      Lag  \n",
       "9       1  \n",
       "14      1  \n",
       "19      3  \n",
       "20      2  \n",
       "22      3  \n",
       "...   ...  \n",
       "2695    2  \n",
       "2696    1  \n",
       "2697    3  \n",
       "2698    1  \n",
       "2699    2  \n",
       "\n",
       "[2430 rows x 9 columns]"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sg.full_edge_list"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 6: Get Aggregate Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will generate an aggregate model by averaging the rank of each edge accross all windowed models. There are several options that allow you to specify whether the mean or median rank is selected and this is discussed in the documentations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Lumping edges...0 out of 90...[DONE]\n",
      "Calculating mean edge importance...[DONE]\n"
     ]
    }
   ],
   "source": [
    "sg.make_static_edge_dict(self_edges=False, lag_method='mean_mean')\n",
    "ranked_edges = sg.make_sort_df(sg.edge_dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>regulator-target</th>\n",
       "      <th>mean_importance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>(G3, G2)</td>\n",
       "      <td>0.083092</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>(G2, G1)</td>\n",
       "      <td>0.071084</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>(G2, G10)</td>\n",
       "      <td>0.064172</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>(G2, G3)</td>\n",
       "      <td>0.056506</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>(G3, G10)</td>\n",
       "      <td>0.056406</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>85</th>\n",
       "      <td>(G9, G2)</td>\n",
       "      <td>0.012702</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>86</th>\n",
       "      <td>(G8, G6)</td>\n",
       "      <td>0.012146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>87</th>\n",
       "      <td>(G7, G6)</td>\n",
       "      <td>0.010763</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>88</th>\n",
       "      <td>(G3, G7)</td>\n",
       "      <td>0.010681</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>89</th>\n",
       "      <td>(G5, G7)</td>\n",
       "      <td>0.007999</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>90 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   regulator-target  mean_importance\n",
       "0          (G3, G2)         0.083092\n",
       "1          (G2, G1)         0.071084\n",
       "2         (G2, G10)         0.064172\n",
       "3          (G2, G3)         0.056506\n",
       "4         (G3, G10)         0.056406\n",
       "..              ...              ...\n",
       "85         (G9, G2)         0.012702\n",
       "86         (G8, G6)         0.012146\n",
       "87         (G7, G6)         0.010763\n",
       "88         (G3, G7)         0.010681\n",
       "89         (G5, G7)         0.007999\n",
       "\n",
       "[90 rows x 2 columns]"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ranked_edges"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Optional Step 7: Visualize network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can use networkx to visualize the inferred network."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>regulator-target</th>\n",
       "      <th>mean_importance</th>\n",
       "      <th>Source</th>\n",
       "      <th>Target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>(G3, G2)</td>\n",
       "      <td>0.083092</td>\n",
       "      <td>G3</td>\n",
       "      <td>G2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>(G2, G1)</td>\n",
       "      <td>0.071084</td>\n",
       "      <td>G2</td>\n",
       "      <td>G1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>(G2, G10)</td>\n",
       "      <td>0.064172</td>\n",
       "      <td>G2</td>\n",
       "      <td>G10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>(G2, G3)</td>\n",
       "      <td>0.056506</td>\n",
       "      <td>G2</td>\n",
       "      <td>G3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>(G3, G10)</td>\n",
       "      <td>0.056406</td>\n",
       "      <td>G3</td>\n",
       "      <td>G10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>(G3, G5)</td>\n",
       "      <td>0.042061</td>\n",
       "      <td>G3</td>\n",
       "      <td>G5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>(G2, G7)</td>\n",
       "      <td>0.041773</td>\n",
       "      <td>G2</td>\n",
       "      <td>G7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>(G1, G2)</td>\n",
       "      <td>0.037300</td>\n",
       "      <td>G1</td>\n",
       "      <td>G2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>(G4, G8)</td>\n",
       "      <td>0.037182</td>\n",
       "      <td>G4</td>\n",
       "      <td>G8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>(G2, G5)</td>\n",
       "      <td>0.036159</td>\n",
       "      <td>G2</td>\n",
       "      <td>G5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>16 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   regulator-target  mean_importance Source Target\n",
       "0          (G3, G2)         0.083092     G3     G2\n",
       "1          (G2, G1)         0.071084     G2     G1\n",
       "2         (G2, G10)         0.064172     G2    G10\n",
       "3          (G2, G3)         0.056506     G2     G3\n",
       "4         (G3, G10)         0.056406     G3    G10\n",
       "..              ...              ...    ...    ...\n",
       "11         (G3, G5)         0.042061     G3     G5\n",
       "12         (G2, G7)         0.041773     G2     G7\n",
       "13         (G1, G2)         0.037300     G1     G2\n",
       "14         (G4, G8)         0.037182     G4     G8\n",
       "15         (G2, G5)         0.036159     G2     G5\n",
       "\n",
       "[16 rows x 4 columns]"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ranked_edges[['Source', 'Target']] = ranked_edges['regulator-target'].apply(pd.Series)\n",
    "# set arbitrary cutoff to 20 edges\n",
    "trunc_ranked_edges = ranked_edges.ix[0:15]\n",
    "trunc_ranked_edges"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "G = nx.DiGraph()\n",
    "for index, row in trunc_ranked_edges.iterrows():\n",
    "    G.add_edge(row['Source'], row['Target'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'G1': Text(-25.0788,33.7245,'G1'),\n",
       " 'G10': Text(69.5195,-46.321,'G10'),\n",
       " 'G2': Text(6.20793,18.6342,'G2'),\n",
       " 'G3': Text(27.0781,11.3467,'G3'),\n",
       " 'G4': Text(-52.611,-25.6457,'G4'),\n",
       " 'G5': Text(33.1733,63.5481,'G5'),\n",
       " 'G6': Text(-47.7731,-100,'G6'),\n",
       " 'G7': Text(-37.1908,40.7323,'G7'),\n",
       " 'G8': Text(-56.9153,49.0383,'G8'),\n",
       " 'G9': Text(83.5901,-45.0574,'G9')}"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x112508048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pos=nx.spring_layout(G, scale=100)\n",
    "plt.axis('off')\n",
    "nx.draw_networkx_nodes(G,pos,node_color='r',node_size=500,alpha=0.8)\n",
    "nx.draw_networkx_edges(G,pos,width=1.0,alpha=1)\n",
    "nx.draw_networkx_labels(G,pos,font_size=16)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Optional Step 8: Score Network (if gold standard is available)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The Swing object has a built-in method for scoring models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Scoring model...[DONE]\n",
      "AUROC: 0.7937777777777767\n",
      "AUPR: 0.4378104716073147\n"
     ]
    }
   ],
   "source": [
    "gold_standard_file = \"../data/Ecoli_10_node/Ecoli10-1_goldstandard.tsv\"\n",
    "roc_dict, pr_dict = sg.score(ranked_edges, gold_standard_file=gold_standard_file)\n",
    "print('AUROC:', roc_dict['auroc'][-1])\n",
    "print('AUPR:', pr_dict['aupr'][-1])"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (swing)",
   "language": "python",
   "name": "swing"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}