docs/run_example.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sourcepredict example 1: Gut host species prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import pandas_ml as pdml"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this example, we will use Sourcepredict and Sourcetracker2 applied to the example dataset provided in the Sourcepredict directory.   \n",
    "\n",
    "The [example datasets](https://github.com/maxibor/sourcepredict/blob/master/data/modern_gut_microbiomes_sources.csv) contains the following samples:\n",
    "- *Homo sapiens* gut microbiome ([1](https://doi.org/10.1038/nature11234), [2](https://doi.org/10.1093/gigascience/giz004), [3](https://doi.org/10.1038/s41564-019-0409-6), [4](https://doi.org/10.1016/j.cell.2019.01.001), [5](https://doi.org/10.1038/ncomms7505), [6](http://doi.org/10.1016/j.cub.2015.04.055))\n",
    "- *Canis familiaris* gut microbiome ([1](https://doi.org/10.1186/s40168-018-0450-3))\n",
    "- Soil microbiome ([1](https://doi.org/10.1073/pnas.1215210110), [2](https://www.ncbi.nlm.nih.gov/bioproject/?term=322597), [3](https://dx.doi.org/10.1128%2FAEM.01646-17))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preparing the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnt = pd.read_csv(\"../data/modern_gut_microbiomes_sources.csv\", index_col=0)\n",
    "labels = pd.read_csv(\"../data/modern_gut_microbiomes_labels.csv\", index_col=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This is a TAXID count table containing the samples as columns headers, and the TAXID as row indices"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>SRR1175007</th>\n",
       "      <th>SRR042182</th>\n",
       "      <th>SRR061154</th>\n",
       "      <th>SRR061499</th>\n",
       "      <th>SRR063469</th>\n",
       "      <th>SRR062324</th>\n",
       "      <th>SRR1179037</th>\n",
       "      <th>SRR061236</th>\n",
       "      <th>SRR061456</th>\n",
       "      <th>SRR642021</th>\n",
       "      <th>...</th>\n",
       "      <th>mgm4477903_3</th>\n",
       "      <th>mgm4477807_3</th>\n",
       "      <th>mgm4477874_3</th>\n",
       "      <th>mgm4477904_3</th>\n",
       "      <th>mgm4477804_3</th>\n",
       "      <th>mgm4477873_3</th>\n",
       "      <th>ERR1939166</th>\n",
       "      <th>SRR3578625</th>\n",
       "      <th>ERR1939165</th>\n",
       "      <th>SRR3578645</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>TAXID</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>0</th>\n",
       "      <td>3528337.0</td>\n",
       "      <td>11563613.0</td>\n",
       "      <td>10084261.0</td>\n",
       "      <td>20054993.0</td>\n",
       "      <td>8747525.0</td>\n",
       "      <td>12116517.0</td>\n",
       "      <td>4191329.0</td>\n",
       "      <td>13992760.0</td>\n",
       "      <td>14825759.0</td>\n",
       "      <td>11083673.0</td>\n",
       "      <td>...</td>\n",
       "      <td>6169203.0</td>\n",
       "      <td>8820851.0</td>\n",
       "      <td>5713837.0</td>\n",
       "      <td>10238500.0</td>\n",
       "      <td>5055930.0</td>\n",
       "      <td>10380594.0</td>\n",
       "      <td>13391896.0</td>\n",
       "      <td>1553.0</td>\n",
       "      <td>14802198.0</td>\n",
       "      <td>736.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.0</td>\n",
       "      <td>78.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>127.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>79.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>172.0</td>\n",
       "      <td>...</td>\n",
       "      <td>68.0</td>\n",
       "      <td>247.0</td>\n",
       "      <td>211.0</td>\n",
       "      <td>156.0</td>\n",
       "      <td>147.0</td>\n",
       "      <td>383.0</td>\n",
       "      <td>1353.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1522.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.0</td>\n",
       "      <td>78.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>127.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>79.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>172.0</td>\n",
       "      <td>...</td>\n",
       "      <td>68.0</td>\n",
       "      <td>247.0</td>\n",
       "      <td>211.0</td>\n",
       "      <td>156.0</td>\n",
       "      <td>147.0</td>\n",
       "      <td>383.0</td>\n",
       "      <td>1353.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1522.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.0</td>\n",
       "      <td>129.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>153.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>151.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>165.0</td>\n",
       "      <td>96.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>77.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>65.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.0</td>\n",
       "      <td>160.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>193.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>99.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>55.0</td>\n",
       "      <td>249.0</td>\n",
       "      <td>238.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>263.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>466.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 432 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       SRR1175007   SRR042182   SRR061154   SRR061499  SRR063469   SRR062324  \\\n",
       "TAXID                                                                          \n",
       "0       3528337.0  11563613.0  10084261.0  20054993.0  8747525.0  12116517.0   \n",
       "6             0.0        78.0         0.0       127.0        0.0        79.0   \n",
       "7             0.0        78.0         0.0       127.0        0.0        79.0   \n",
       "9             0.0       129.0         0.0       153.0        0.0       151.0   \n",
       "10            0.0       160.0         0.0       193.0        0.0        99.0   \n",
       "\n",
       "       SRR1179037   SRR061236   SRR061456   SRR642021  ...  mgm4477903_3  \\\n",
       "TAXID                                                  ...                 \n",
       "0       4191329.0  13992760.0  14825759.0  11083673.0  ...     6169203.0   \n",
       "6             0.0         0.0         0.0       172.0  ...          68.0   \n",
       "7             0.0         0.0         0.0       172.0  ...          68.0   \n",
       "9             0.0       165.0        96.0         0.0  ...           0.0   \n",
       "10            0.0        55.0       249.0       238.0  ...           0.0   \n",
       "\n",
       "       mgm4477807_3  mgm4477874_3  mgm4477904_3  mgm4477804_3  mgm4477873_3  \\\n",
       "TAXID                                                                         \n",
       "0         8820851.0     5713837.0    10238500.0     5055930.0    10380594.0   \n",
       "6             247.0         211.0         156.0         147.0         383.0   \n",
       "7             247.0         211.0         156.0         147.0         383.0   \n",
       "9               0.0           0.0           0.0           0.0           0.0   \n",
       "10              0.0           0.0           0.0           0.0           0.0   \n",
       "\n",
       "       ERR1939166  SRR3578625  ERR1939165  SRR3578645  \n",
       "TAXID                                                  \n",
       "0      13391896.0      1553.0  14802198.0       736.0  \n",
       "6          1353.0         0.0      1522.0         0.0  \n",
       "7          1353.0         0.0      1522.0         0.0  \n",
       "9            77.0         0.0        65.0         0.0  \n",
       "10          263.0         0.0       466.0         0.0  \n",
       "\n",
       "[5 rows x 432 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cnt.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The labels file contains the mapping of samples names with their actual origin (sources)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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>labels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SRR1175007</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR042182</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR061154</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR061499</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR063469</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  labels\n",
       "SRR1175007  Homo_sapiens\n",
       "SRR042182   Homo_sapiens\n",
       "SRR061154   Homo_sapiens\n",
       "SRR061499   Homo_sapiens\n",
       "SRR063469   Homo_sapiens"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will divide the source in training (95%) and testing (5%) dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnt_train = cnt.sample(frac=0.95, axis=1)\n",
    "cnt_test = cnt.drop(cnt_train.columns, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We also have to subset the labels file to only the training dataset "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_labels = labels.loc[cnt_train.columns,:]\n",
    "test_labels = labels.loc[cnt_test.columns,:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Sourcepredict"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Last but not least, we must export the files to `csv` to run sourcepredict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnt_train.to_csv(\"gut_species_sources.csv\")\n",
    "cnt_test.to_csv(\"gut_species_sinks.csv\")\n",
    "train_labels.to_csv(\"gut_species_labels.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll now launch sourcepredict with the GMPR normalization method, and the t-SNE embedding, on 6 cores."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Step 1: Checking for unknown proportion\n",
      "  == Sample: SRR1175007 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: SRR1175007\n",
      "\t\t known:98.68%\n",
      "\t\t unknown:1.32%\n",
      "  == Sample: SRR061236 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: SRR061236\n",
      "\t\t known:85.01%\n",
      "\t\t unknown:14.99%\n",
      "  == Sample: SRR063471 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: SRR063471\n",
      "\t\t known:98.4%\n",
      "\t\t unknown:1.6%\n",
      "  == Sample: SRR1930132 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: SRR1930132\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: SRR1930133 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 0.99\n",
      "\t----------------------\n",
      "\t- Sample: SRR1930133\n",
      "\t\t known:98.49%\n",
      "\t\t unknown:1.51%\n",
      "  == Sample: SRR7658586 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: SRR7658586\n",
      "\t\t known:79.65%\n",
      "\t\t unknown:20.35%\n",
      "  == Sample: SRR7658645 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 0.99\n",
      "\t----------------------\n",
      "\t- Sample: SRR7658645\n",
      "\t\t known:26.88%\n",
      "\t\t unknown:73.12%\n",
      "  == Sample: SRR7658584 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 0.98\n",
      "\t----------------------\n",
      "\t- Sample: SRR7658584\n",
      "\t\t known:85.78%\n",
      "\t\t unknown:14.22%\n",
      "  == Sample: SRR7658607 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 0.99\n",
      "\t----------------------\n",
      "\t- Sample: SRR7658607\n",
      "\t\t known:98.74%\n",
      "\t\t unknown:1.26%\n",
      "  == Sample: SRR7658597 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 0.98\n",
      "\t----------------------\n",
      "\t- Sample: SRR7658597\n",
      "\t\t known:99.1%\n",
      "\t\t unknown:0.9%\n",
      "  == Sample: SRR5898944 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: SRR5898944\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: ERR1914439 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: ERR1914439\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: ERR1915140 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: ERR1915140\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: ERR1914041 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: ERR1914041\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: ERR1915022 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: ERR1915022\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: ERR1915826 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: ERR1915826\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: ERR1913400 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: ERR1913400\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: ERR1915765 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: ERR1915765\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: ERR1915225 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: ERR1915225\n",
      "\t\t known:98.48%\n",
      "\t\t unknown:1.52%\n",
      "  == Sample: mgm4477874_3 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 1.0\n",
      "\t----------------------\n",
      "\t- Sample: mgm4477874_3\n",
      "\t\t known:72.37%\n",
      "\t\t unknown:27.63%\n",
      "  == Sample: ERR1939166 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 0.97\n",
      "\t----------------------\n",
      "\t- Sample: ERR1939166\n",
      "\t\t known:47.44%\n",
      "\t\t unknown:52.56%\n",
      "  == Sample: ERR1939165 ==\n",
      "\tAdding unknown\n",
      "\tNormalizing (GMPR)\n",
      "\tComputing Bray-Curtis distance\n",
      "\tPerforming MDS embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tTraining KNN classifier on 6 cores...\n",
      "\t-> Testing Accuracy: 0.97\n",
      "\t----------------------\n",
      "\t- Sample: ERR1939165\n",
      "\t\t known:55.27%\n",
      "\t\t unknown:44.73%\n",
      "Step 2: Checking for source proportion\n",
      "\tComputing weighted_unifrac distance on species rank\n",
      "\tTSNE embedding in 2 dimensions\n",
      "\tKNN machine learning\n",
      "\tPerforming 5 fold cross validation on 6 cores...\n",
      "\tTrained KNN classifier with 10 neighbors\n",
      "\t-> Testing Accuracy: 0.99\n",
      "\t----------------------\n",
      "\t- Sample: SRR1175007\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR061236\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR063471\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR1930132\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR1930133\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR7658586\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR7658645\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR7658584\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR7658607\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR7658597\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: SRR5898944\n",
      "\t\t Canis_familiaris:1.81%\n",
      "\t\t Homo_sapiens:96.72%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: ERR1914439\n",
      "\t\t Canis_familiaris:94.25%\n",
      "\t\t Homo_sapiens:4.28%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: ERR1915140\n",
      "\t\t Canis_familiaris:94.25%\n",
      "\t\t Homo_sapiens:4.28%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: ERR1914041\n",
      "\t\t Canis_familiaris:94.25%\n",
      "\t\t Homo_sapiens:4.28%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: ERR1915022\n",
      "\t\t Canis_familiaris:94.25%\n",
      "\t\t Homo_sapiens:4.28%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: ERR1915826\n",
      "\t\t Canis_familiaris:94.25%\n",
      "\t\t Homo_sapiens:4.28%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: ERR1913400\n",
      "\t\t Canis_familiaris:94.25%\n",
      "\t\t Homo_sapiens:4.28%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: ERR1915765\n",
      "\t\t Canis_familiaris:94.25%\n",
      "\t\t Homo_sapiens:4.28%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: ERR1915225\n",
      "\t\t Canis_familiaris:94.25%\n",
      "\t\t Homo_sapiens:4.28%\n",
      "\t\t Soil:1.47%\n",
      "\t- Sample: mgm4477874_3\n",
      "\t\t Canis_familiaris:2.51%\n",
      "\t\t Homo_sapiens:5.95%\n",
      "\t\t Soil:91.53%\n",
      "\t- Sample: ERR1939166\n",
      "\t\t Canis_familiaris:2.51%\n",
      "\t\t Homo_sapiens:5.95%\n",
      "\t\t Soil:91.53%\n",
      "\t- Sample: ERR1939165\n",
      "\t\t Canis_familiaris:2.51%\n",
      "\t\t Homo_sapiens:5.95%\n",
      "\t\t Soil:91.53%\n",
      "Sourcepredict result written to gut_species_sinks.sourcepredict.csv\n",
      "Embedding coordinates written to example_embedding.csv\n",
      "\u001b[0mCPU times: user 3.64 s, sys: 828 ms, total: 4.47 s\n",
      "Wall time: 4min 2s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "!sourcepredict -s gut_species_sources.csv \\\n",
    "               -l gut_species_labels.csv \\\n",
    "               -n GMPR \\\n",
    "               -m TSNE \\\n",
    "               -e example_embedding.csv \\\n",
    "               -t 6 gut_species_sinks.csv"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Two files were generated by Sourcepredict:\n",
    "- `gut_species_sinks.sourcepredict.csv` which contains the proportions of each source  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "sourcepred = pd.read_csv(\"gut_species_sinks.sourcepredict.csv\", index_col=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "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>SRR1175007</th>\n",
       "      <th>SRR061236</th>\n",
       "      <th>SRR063471</th>\n",
       "      <th>SRR1930132</th>\n",
       "      <th>SRR1930133</th>\n",
       "      <th>SRR7658586</th>\n",
       "      <th>SRR7658645</th>\n",
       "      <th>SRR7658584</th>\n",
       "      <th>SRR7658607</th>\n",
       "      <th>SRR7658597</th>\n",
       "      <th>...</th>\n",
       "      <th>ERR1915140</th>\n",
       "      <th>ERR1914041</th>\n",
       "      <th>ERR1915022</th>\n",
       "      <th>ERR1915826</th>\n",
       "      <th>ERR1913400</th>\n",
       "      <th>ERR1915765</th>\n",
       "      <th>ERR1915225</th>\n",
       "      <th>mgm4477874_3</th>\n",
       "      <th>ERR1939166</th>\n",
       "      <th>ERR1939165</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Canis_familiaris</th>\n",
       "      <td>0.017814</td>\n",
       "      <td>0.015347</td>\n",
       "      <td>0.017765</td>\n",
       "      <td>0.017779</td>\n",
       "      <td>0.017781</td>\n",
       "      <td>0.014379</td>\n",
       "      <td>0.004853</td>\n",
       "      <td>0.015486</td>\n",
       "      <td>0.017825</td>\n",
       "      <td>0.017891</td>\n",
       "      <td>...</td>\n",
       "      <td>0.928216</td>\n",
       "      <td>0.928216</td>\n",
       "      <td>0.928216</td>\n",
       "      <td>0.928216</td>\n",
       "      <td>0.928216</td>\n",
       "      <td>0.928216</td>\n",
       "      <td>0.928216</td>\n",
       "      <td>0.018200</td>\n",
       "      <td>0.011931</td>\n",
       "      <td>0.013898</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Homo_sapiens</th>\n",
       "      <td>0.954443</td>\n",
       "      <td>0.822254</td>\n",
       "      <td>0.951788</td>\n",
       "      <td>0.952581</td>\n",
       "      <td>0.952644</td>\n",
       "      <td>0.770401</td>\n",
       "      <td>0.260037</td>\n",
       "      <td>0.829699</td>\n",
       "      <td>0.955022</td>\n",
       "      <td>0.958548</td>\n",
       "      <td>...</td>\n",
       "      <td>0.042128</td>\n",
       "      <td>0.042128</td>\n",
       "      <td>0.042128</td>\n",
       "      <td>0.042128</td>\n",
       "      <td>0.042128</td>\n",
       "      <td>0.042128</td>\n",
       "      <td>0.042128</td>\n",
       "      <td>0.043078</td>\n",
       "      <td>0.028241</td>\n",
       "      <td>0.032896</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Soil</th>\n",
       "      <td>0.014516</td>\n",
       "      <td>0.012506</td>\n",
       "      <td>0.014476</td>\n",
       "      <td>0.014488</td>\n",
       "      <td>0.014489</td>\n",
       "      <td>0.011717</td>\n",
       "      <td>0.003955</td>\n",
       "      <td>0.012619</td>\n",
       "      <td>0.014525</td>\n",
       "      <td>0.014579</td>\n",
       "      <td>...</td>\n",
       "      <td>0.014504</td>\n",
       "      <td>0.014504</td>\n",
       "      <td>0.014504</td>\n",
       "      <td>0.014504</td>\n",
       "      <td>0.014504</td>\n",
       "      <td>0.014504</td>\n",
       "      <td>0.014504</td>\n",
       "      <td>0.662431</td>\n",
       "      <td>0.434275</td>\n",
       "      <td>0.505859</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>unknown</th>\n",
       "      <td>0.013226</td>\n",
       "      <td>0.149893</td>\n",
       "      <td>0.015971</td>\n",
       "      <td>0.015152</td>\n",
       "      <td>0.015086</td>\n",
       "      <td>0.203503</td>\n",
       "      <td>0.731155</td>\n",
       "      <td>0.142196</td>\n",
       "      <td>0.012627</td>\n",
       "      <td>0.008983</td>\n",
       "      <td>...</td>\n",
       "      <td>0.015152</td>\n",
       "      <td>0.015152</td>\n",
       "      <td>0.015152</td>\n",
       "      <td>0.015152</td>\n",
       "      <td>0.015152</td>\n",
       "      <td>0.015152</td>\n",
       "      <td>0.015152</td>\n",
       "      <td>0.276292</td>\n",
       "      <td>0.525554</td>\n",
       "      <td>0.447347</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>4 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                  SRR1175007  SRR061236  SRR063471  SRR1930132  SRR1930133  \\\n",
       "Canis_familiaris    0.017814   0.015347   0.017765    0.017779    0.017781   \n",
       "Homo_sapiens        0.954443   0.822254   0.951788    0.952581    0.952644   \n",
       "Soil                0.014516   0.012506   0.014476    0.014488    0.014489   \n",
       "unknown             0.013226   0.149893   0.015971    0.015152    0.015086   \n",
       "\n",
       "                  SRR7658586  SRR7658645  SRR7658584  SRR7658607  SRR7658597  \\\n",
       "Canis_familiaris    0.014379    0.004853    0.015486    0.017825    0.017891   \n",
       "Homo_sapiens        0.770401    0.260037    0.829699    0.955022    0.958548   \n",
       "Soil                0.011717    0.003955    0.012619    0.014525    0.014579   \n",
       "unknown             0.203503    0.731155    0.142196    0.012627    0.008983   \n",
       "\n",
       "                  ...  ERR1915140  ERR1914041  ERR1915022  ERR1915826  \\\n",
       "Canis_familiaris  ...    0.928216    0.928216    0.928216    0.928216   \n",
       "Homo_sapiens      ...    0.042128    0.042128    0.042128    0.042128   \n",
       "Soil              ...    0.014504    0.014504    0.014504    0.014504   \n",
       "unknown           ...    0.015152    0.015152    0.015152    0.015152   \n",
       "\n",
       "                  ERR1913400  ERR1915765  ERR1915225  mgm4477874_3  \\\n",
       "Canis_familiaris    0.928216    0.928216    0.928216      0.018200   \n",
       "Homo_sapiens        0.042128    0.042128    0.042128      0.043078   \n",
       "Soil                0.014504    0.014504    0.014504      0.662431   \n",
       "unknown             0.015152    0.015152    0.015152      0.276292   \n",
       "\n",
       "                  ERR1939166  ERR1939165  \n",
       "Canis_familiaris    0.011931    0.013898  \n",
       "Homo_sapiens        0.028241    0.032896  \n",
       "Soil                0.434275    0.505859  \n",
       "unknown             0.525554    0.447347  \n",
       "\n",
       "[4 rows x 22 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sourcepred"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's check which organism was predicted for each samples, and compare it with the true source"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/borry/miniconda3/lib/python3.6/site-packages/pandas/core/indexing.py:1494: FutureWarning: \n",
      "Passing list-likes to .loc or [] with any missing label will raise\n",
      "KeyError in the future, you can use .reindex() as an alternative.\n",
      "\n",
      "See the documentation here:\n",
      "https://pandas.pydata.org/pandas-docs/stable/indexing.html#deprecate-loc-reindex-listlike\n",
      "  return self._getitem_tuple(key)\n"
     ]
    }
   ],
   "source": [
    "comparison = sourcepred.idxmax().to_frame(name=\"prediction\").merge(test_labels, left_index=True, right_index=True)\n",
    "cm = pdml.ConfusionMatrix(y_true=comparison['labels'],y_pred=comparison['prediction'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's look at the confusion matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "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>Predicted</th>\n",
       "      <th>Canis_familiaris</th>\n",
       "      <th>Homo_sapiens</th>\n",
       "      <th>Soil</th>\n",
       "      <th>unknown</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Actual</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Canis_familiaris</th>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Homo_sapiens</th>\n",
       "      <td>0</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Soil</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>unknown</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Predicted         Canis_familiaris  Homo_sapiens  Soil  unknown\n",
       "Actual                                                         \n",
       "Canis_familiaris                 8             0     0        0\n",
       "Homo_sapiens                     0            10     0        1\n",
       "Soil                             0             0     2        1\n",
       "unknown                          0             0     0        0"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cm.to_dataframe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, let's compute the accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.91"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "round(cm.stats()['overall']['Accuracy'],2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "91% of the sink samples were correctly predicted ! "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- The second file generated by sourcepredict is `example_embedding.csv` which contains the embedding coordinates of all samples (sources and sinks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>PC1</th>\n",
       "      <th>PC2</th>\n",
       "      <th>labels</th>\n",
       "      <th>name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SRR1761672</th>\n",
       "      <td>-12.706208</td>\n",
       "      <td>-7.860738</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>SRR1761672</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR061456</th>\n",
       "      <td>-4.520492</td>\n",
       "      <td>24.795073</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>SRR061456</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR1761718</th>\n",
       "      <td>-20.427488</td>\n",
       "      <td>-6.425568</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>SRR1761718</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR7658589</th>\n",
       "      <td>-23.176891</td>\n",
       "      <td>-2.985772</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>SRR7658589</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1914932</th>\n",
       "      <td>28.669333</td>\n",
       "      <td>12.863045</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>ERR1914932</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  PC1        PC2            labels        name\n",
       "SRR1761672 -12.706208  -7.860738      Homo_sapiens  SRR1761672\n",
       "SRR061456   -4.520492  24.795073      Homo_sapiens   SRR061456\n",
       "SRR1761718 -20.427488  -6.425568      Homo_sapiens  SRR1761718\n",
       "SRR7658589 -23.176891  -2.985772      Homo_sapiens  SRR7658589\n",
       "ERR1914932  28.669333  12.863045  Canis_familiaris  ERR1914932"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "embed = pd.read_csv(\"example_embedding.csv\", index_col=0)\n",
    "embed.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can plot this embedding, using for example, [plotnine](https://plotnine.readthedocs.io/en/stable/), which implements the grammar of graphics in Python"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "from plotnine import *\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<ggplot: (-9223372029842878797)>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ggplot(data = embed, mapping = aes(x=\"PC1\",y=\"PC2\", color=\"labels\")) + geom_point() + theme_classic()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see on this plot where the sink samples were embedded"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Sourcetracker2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\"[SourceTracker](https://www.ncbi.nlm.nih.gov/pubmed/21765408) is designed to predict the source of microbial communities in a set of input samples\" and is generally consired as the gold standard method to do so.\n",
    "The version 2 is a rewrite of the original Sourcetracker in Python."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We'll reuse the same training and test files, but we need to reformat them a bit for sourcetracker:\n",
    "- In sourcetracker, the source (training) and sink (file) TAXIDs count table is a single file\n",
    "- The metadata file is slightly different"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "cnt.to_csv(\"gut_species_taxid.csv\", sep=\"\\t\", index_label=\"TAXID\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_labels['SourceSink'] = ['sink']*test_labels.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_labels['SourceSink'] = ['source']*train_labels.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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>SourceSink</th>\n",
       "      <th>Env</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SRR1761672</th>\n",
       "      <td>source</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR061456</th>\n",
       "      <td>source</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR1761718</th>\n",
       "      <td>source</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR7658589</th>\n",
       "      <td>source</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1914932</th>\n",
       "      <td>source</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           SourceSink               Env\n",
       "SRR1761672     source      Homo_sapiens\n",
       "SRR061456      source      Homo_sapiens\n",
       "SRR1761718     source      Homo_sapiens\n",
       "SRR7658589     source      Homo_sapiens\n",
       "ERR1914932     source  Canis_familiaris"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metadata = train_labels.append(test_labels).rename(columns = {\"labels\":\"Env\"})[['SourceSink','Env']]\n",
    "metadata.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "metadata.to_csv(\"st_gut_species_metadata.csv\", sep=\"\\t\", index_label='#SampleID')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, we need to convert the TAXIDs count table to biom format"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "!biom convert -i gut_species_taxid.csv -o gut_species_taxid.biom --table-type=\"Taxon table\" --to-json"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Soucetracker launch command: `sourcetracker2 gibbs -i gut_species_taxid.biom -m st_gut_species_metadata.csv -o gut_species --jobs 6`  \n",
    "(Sourcetracker2 was run on a Linux remote server because of issues running it on MacOS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "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>Canis_familiaris</th>\n",
       "      <th>Homo_sapiens</th>\n",
       "      <th>Soil</th>\n",
       "      <th>Unknown</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>#SampleID</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SRR1175007</th>\n",
       "      <td>0.0170</td>\n",
       "      <td>0.9609</td>\n",
       "      <td>0.0063</td>\n",
       "      <td>0.0158</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR061236</th>\n",
       "      <td>0.0358</td>\n",
       "      <td>0.9365</td>\n",
       "      <td>0.0074</td>\n",
       "      <td>0.0203</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR063471</th>\n",
       "      <td>0.0121</td>\n",
       "      <td>0.9724</td>\n",
       "      <td>0.0032</td>\n",
       "      <td>0.0123</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR1930132</th>\n",
       "      <td>0.1466</td>\n",
       "      <td>0.3761</td>\n",
       "      <td>0.4477</td>\n",
       "      <td>0.0296</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR1930133</th>\n",
       "      <td>0.1182</td>\n",
       "      <td>0.5082</td>\n",
       "      <td>0.3507</td>\n",
       "      <td>0.0229</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            Canis_familiaris  Homo_sapiens    Soil  Unknown\n",
       "#SampleID                                                  \n",
       "SRR1175007            0.0170        0.9609  0.0063   0.0158\n",
       "SRR061236             0.0358        0.9365  0.0074   0.0203\n",
       "SRR063471             0.0121        0.9724  0.0032   0.0123\n",
       "SRR1930132            0.1466        0.3761  0.4477   0.0296\n",
       "SRR1930133            0.1182        0.5082  0.3507   0.0229"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "st_pred = pd.read_csv(\"gut_species/mixing_proportions.txt\", sep = \"\\t\", index_col=0)\n",
    "st_pred.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "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>prediction</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>#SampleID</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SRR1175007</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR061236</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR063471</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR1930132</th>\n",
       "      <td>Soil</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR1930133</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              prediction\n",
       "#SampleID               \n",
       "SRR1175007  Homo_sapiens\n",
       "SRR061236   Homo_sapiens\n",
       "SRR063471   Homo_sapiens\n",
       "SRR1930132          Soil\n",
       "SRR1930133  Homo_sapiens"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "st_comparison = st_pred.idxmax(axis=1).to_frame(name=\"prediction\")\n",
    "st_comparison.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's compare the SourceTracker prediction with the true source"
   ]
  },
  {
   "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></th>\n",
       "      <th>prediction</th>\n",
       "      <th>labels</th>\n",
       "      <th>SourceSink</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>SRR1175007</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR061236</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR063471</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR1930132</th>\n",
       "      <td>Soil</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR1930133</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR7658586</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR7658645</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR7658584</th>\n",
       "      <td>Soil</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR7658607</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR7658597</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SRR5898944</th>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>Homo_sapiens</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1914439</th>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1915140</th>\n",
       "      <td>Soil</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1914041</th>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1915022</th>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1915826</th>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1913400</th>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1915765</th>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1915225</th>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>Canis_familiaris</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mgm4477874_3</th>\n",
       "      <td>Soil</td>\n",
       "      <td>Soil</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1939166</th>\n",
       "      <td>Soil</td>\n",
       "      <td>Soil</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ERR1939165</th>\n",
       "      <td>Soil</td>\n",
       "      <td>Soil</td>\n",
       "      <td>sink</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    prediction            labels SourceSink\n",
       "SRR1175007        Homo_sapiens      Homo_sapiens       sink\n",
       "SRR061236         Homo_sapiens      Homo_sapiens       sink\n",
       "SRR063471         Homo_sapiens      Homo_sapiens       sink\n",
       "SRR1930132                Soil      Homo_sapiens       sink\n",
       "SRR1930133        Homo_sapiens      Homo_sapiens       sink\n",
       "SRR7658586        Homo_sapiens      Homo_sapiens       sink\n",
       "SRR7658645        Homo_sapiens      Homo_sapiens       sink\n",
       "SRR7658584                Soil      Homo_sapiens       sink\n",
       "SRR7658607        Homo_sapiens      Homo_sapiens       sink\n",
       "SRR7658597        Homo_sapiens      Homo_sapiens       sink\n",
       "SRR5898944        Homo_sapiens      Homo_sapiens       sink\n",
       "ERR1914439    Canis_familiaris  Canis_familiaris       sink\n",
       "ERR1915140                Soil  Canis_familiaris       sink\n",
       "ERR1914041    Canis_familiaris  Canis_familiaris       sink\n",
       "ERR1915022    Canis_familiaris  Canis_familiaris       sink\n",
       "ERR1915826    Canis_familiaris  Canis_familiaris       sink\n",
       "ERR1913400    Canis_familiaris  Canis_familiaris       sink\n",
       "ERR1915765    Canis_familiaris  Canis_familiaris       sink\n",
       "ERR1915225    Canis_familiaris  Canis_familiaris       sink\n",
       "mgm4477874_3              Soil              Soil       sink\n",
       "ERR1939166                Soil              Soil       sink\n",
       "ERR1939165                Soil              Soil       sink"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "comparison2 = st_comparison.merge(test_labels, left_index=True, right_index=True)\n",
    "comparison2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Computing the accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "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>Predicted</th>\n",
       "      <th>Canis_familiaris</th>\n",
       "      <th>Homo_sapiens</th>\n",
       "      <th>Soil</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Actual</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Canis_familiaris</th>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Homo_sapiens</th>\n",
       "      <td>0</td>\n",
       "      <td>9</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Soil</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Predicted         Canis_familiaris  Homo_sapiens  Soil\n",
       "Actual                                                \n",
       "Canis_familiaris                 7             0     1\n",
       "Homo_sapiens                     0             9     2\n",
       "Soil                             0             0     3"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cm2 = pdml.ConfusionMatrix(y_true=comparison2[\"labels\"], y_pred=comparison2[\"prediction\"])\n",
    "cm2.to_dataframe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.86"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "acc2 = round(cm2.stats()['overall']['Accuracy'],2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here, Sourcetracker only managed to predict 86% of the sink samples origin correctly"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Conclusion\n",
    "On this dataset, we've seen that Sourcepredict performs similary or even better than Sourcetracker on predicting accurately the source species"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:root] *",
   "language": "python",
   "name": "conda-root-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}