notebooks/CandidateClassification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import sys\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import GridSearchCV, train_test_split, cross_val_score, ShuffleSplit, KFold\n",
    "from sklearn import metrics\n",
    "import seaborn as sns\n",
    "import pickle\n",
    "pd.options.display.float_format = '{:,.3f}'.format\n",
    "\n",
    "import matplotlib\n",
    "font = {'family' : 'Arial', 'size'   : 12}\n",
    "matplotlib.rc('font', **font)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_samples(vectors):\n",
    "    if vectors == 'pfam-dst':\n",
    "        return pd.read_csv('../data/training/positive/mibig_bgcs_all.{}.csv'.format(vectors)).set_index('contig_id')\n",
    "    return pd.read_pickle(\"../data/training/positive/mibig_bgcs_all.{}.pkl\".format(vectors))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_sample_responses(vectors, responses):\n",
    "    samples = get_samples(vectors)\n",
    "    responses = responses.loc[samples.index].dropna()\n",
    "    samples = samples.loc[responses.index]\n",
    "    return samples, responses"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Candidate classification: Multilabel and multiclass method evaluation\n",
    "\n",
    "## Problem specification\n",
    "\n",
    "Predict biosynthetic classes of BGC candidates.\n",
    "\n",
    "### Multilabel approach\n",
    "\n",
    "Assign a set of classes (labels) to each BGC candidate\n",
    "\n",
    "Model output is a binary vector:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Alkaloid</th>\n",
       "      <th>NRP</th>\n",
       "      <th>Other</th>\n",
       "      <th>Polyketide</th>\n",
       "      <th>RiPP</th>\n",
       "      <th>Saccharide</th>\n",
       "      <th>Terpene</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>contig_id</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>BGC0000576.1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0000534.1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0000175.1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0000299.1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0001343.1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Alkaloid  NRP  Other  Polyketide  RiPP  Saccharide  Terpene\n",
       "contig_id                                                                \n",
       "BGC0000576.1         0    0      0           0     1           0        0\n",
       "BGC0000534.1         0    0      0           0     1           0        0\n",
       "BGC0000175.1         0    0      0           1     0           0        0\n",
       "BGC0000299.1         0    1      0           0     0           0        0\n",
       "BGC0001343.1         0    0      0           1     0           0        0"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels = pd.read_csv(\"../data/training/positive/mibig_bgcs_all.classes.csv\").set_index('contig_id')\n",
    "labels = labels[~labels['?'].astype(np.bool)]\n",
    "del labels['?']\n",
    "del labels['Nucleoside']\n",
    "labels.sample(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Input data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Domain set\n",
    "\n",
    "Binary vector of all pfam IDs found in MIBiG dataset, 1 = pfam is present in sequence, 0 = pfam is not present"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>pfam_id</th>\n",
       "      <th>PF00004</th>\n",
       "      <th>PF00005</th>\n",
       "      <th>PF00006</th>\n",
       "      <th>PF00009</th>\n",
       "      <th>PF00010</th>\n",
       "      <th>PF00011</th>\n",
       "      <th>PF00012</th>\n",
       "      <th>PF00013</th>\n",
       "      <th>PF00015</th>\n",
       "      <th>PF00022</th>\n",
       "      <th>...</th>\n",
       "      <th>PF17387</th>\n",
       "      <th>PF17389</th>\n",
       "      <th>PF17390</th>\n",
       "      <th>PF17391</th>\n",
       "      <th>PF17392</th>\n",
       "      <th>PF17414</th>\n",
       "      <th>PF17425</th>\n",
       "      <th>PF17508</th>\n",
       "      <th>PF17584</th>\n",
       "      <th>PF17608</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>contig_id</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>BGC0000001.1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0000002.1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0000003.1</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3 rows × 3294 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "pfam_id       PF00004  PF00005  PF00006  PF00009  PF00010  PF00011  PF00012  \\\n",
       "contig_id                                                                     \n",
       "BGC0000001.1        0        1        0        0        0        0        0   \n",
       "BGC0000002.1        0        1        0        0        0        0        0   \n",
       "BGC0000003.1        0        0        0        0        0        0        0   \n",
       "\n",
       "pfam_id       PF00013  PF00015  PF00022   ...     PF17387  PF17389  PF17390  \\\n",
       "contig_id                                 ...                                 \n",
       "BGC0000001.1        0        0        0   ...           0        0        0   \n",
       "BGC0000002.1        0        0        0   ...           0        0        0   \n",
       "BGC0000003.1        0        0        0   ...           0        0        0   \n",
       "\n",
       "pfam_id       PF17391  PF17392  PF17414  PF17425  PF17508  PF17584  PF17608  \n",
       "contig_id                                                                    \n",
       "BGC0000001.1        0        0        0        0        0        0        0  \n",
       "BGC0000002.1        0        0        0        0        0        0        0  \n",
       "BGC0000003.1        0        0        0        0        0        0        0  \n",
       "\n",
       "[3 rows x 3294 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "get_samples('domain_set').head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "all_ml_samples, all_ml_responses = get_sample_responses('domain_set', labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# AntiSMASH classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>contig_id</th>\n",
       "      <th>nucl_start</th>\n",
       "      <th>nucl_end</th>\n",
       "      <th>candidate_id</th>\n",
       "      <th>classes</th>\n",
       "      <th>detection_rules</th>\n",
       "      <th>genes</th>\n",
       "      <th>subclusters</th>\n",
       "      <th>NRPSs/PKSs</th>\n",
       "      <th>signature_genes</th>\n",
       "      <th>RiPPs</th>\n",
       "      <th>predicted_structure</th>\n",
       "      <th>monomers</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>BGC0000001.1</td>\n",
       "      <td>0</td>\n",
       "      <td>59320</td>\n",
       "      <td>BGC0000001.1:0-59320</td>\n",
       "      <td>t1pks</td>\n",
       "      <td>t1pks: ((PKS_KS &amp;amp; PKS_AT) or (ene_KS &amp;amp;...</td>\n",
       "      <td>AEK75490.1;AEK75491.1;AEK75492.1;AEK75493.1;AE...</td>\n",
       "      <td>nan</td>\n",
       "      <td>NaN</td>\n",
       "      <td>AEK75493;AEK75497;AEK75499;AEK75502;AEK75503;A...</td>\n",
       "      <td>-</td>\n",
       "      <td>nan</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>BGC0000002.1</td>\n",
       "      <td>0</td>\n",
       "      <td>141542</td>\n",
       "      <td>BGC0000002.1:0-141542</td>\n",
       "      <td>oligosaccharide;t1pks</td>\n",
       "      <td>oligosaccharide: (minimum(3,[Glycos_transf_1,G...</td>\n",
       "      <td>AHH99917.1;AHH99918.1;AHH99919.1;AHH99920.1;AH...</td>\n",
       "      <td>nan</td>\n",
       "      <td>NaN</td>\n",
       "      <td>AHH99918;AHH99919;AHH99920;AHH99921;AHH99922;A...</td>\n",
       "      <td>-</td>\n",
       "      <td>nan</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>BGC0000003.1</td>\n",
       "      <td>0</td>\n",
       "      <td>37168</td>\n",
       "      <td>BGC0000003.1:0-37168</td>\n",
       "      <td>t1pks</td>\n",
       "      <td>t1pks: ((PKS_KS &amp;amp; PKS_AT) or (ene_KS &amp;amp;...</td>\n",
       "      <td>BAD97694.1;BAD97695.1;BAD97696.1;BAD97697.1;BA...</td>\n",
       "      <td>nan</td>\n",
       "      <td>BAD97694;BAD97699</td>\n",
       "      <td>BAD97694;BAD97696;BAD97699</td>\n",
       "      <td>-</td>\n",
       "      <td>nan</td>\n",
       "      <td>(pk)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>BGC0000004.1</td>\n",
       "      <td>0</td>\n",
       "      <td>45060</td>\n",
       "      <td>BGC0000004.1:0-45060</td>\n",
       "      <td>t1pks</td>\n",
       "      <td>t1pks: ((PKS_KS &amp;amp; PKS_AT) or (ene_KS &amp;amp;...</td>\n",
       "      <td>BAE71313.1;BAE71314.1;BAE71315.1;BAE71316.1;BA...</td>\n",
       "      <td>nan</td>\n",
       "      <td>BAE71314;BAE71315;BAE71316;BAE71317</td>\n",
       "      <td>BAE71314;BAE71315;BAE71316;BAE71317</td>\n",
       "      <td>-</td>\n",
       "      <td>nan</td>\n",
       "      <td>(pk) + (mal)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>BGC0000006.1</td>\n",
       "      <td>0</td>\n",
       "      <td>43732</td>\n",
       "      <td>BGC0000006.1:0-43732</td>\n",
       "      <td>t1pks</td>\n",
       "      <td>t1pks: ((PKS_KS &amp;amp; PKS_AT) or (ene_KS &amp;amp;...</td>\n",
       "      <td>AAS90092.1;AAS90093.1;AAS90110.1;AAS90111.1;AA...</td>\n",
       "      <td>nan</td>\n",
       "      <td>AAS90093;AAS90110;AAS90111;AAS90095</td>\n",
       "      <td>AAS90093;AAS90110;AAS90111;AAS90095</td>\n",
       "      <td>-</td>\n",
       "      <td>nan</td>\n",
       "      <td>(pk) + (mal)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      contig_id  nucl_start  nucl_end           candidate_id  \\\n",
       "0  BGC0000001.1           0     59320   BGC0000001.1:0-59320   \n",
       "1  BGC0000002.1           0    141542  BGC0000002.1:0-141542   \n",
       "2  BGC0000003.1           0     37168   BGC0000003.1:0-37168   \n",
       "3  BGC0000004.1           0     45060   BGC0000004.1:0-45060   \n",
       "4  BGC0000006.1           0     43732   BGC0000006.1:0-43732   \n",
       "\n",
       "                 classes                                    detection_rules  \\\n",
       "0                  t1pks  t1pks: ((PKS_KS &amp; PKS_AT) or (ene_KS &amp;...   \n",
       "1  oligosaccharide;t1pks  oligosaccharide: (minimum(3,[Glycos_transf_1,G...   \n",
       "2                  t1pks  t1pks: ((PKS_KS &amp; PKS_AT) or (ene_KS &amp;...   \n",
       "3                  t1pks  t1pks: ((PKS_KS &amp; PKS_AT) or (ene_KS &amp;...   \n",
       "4                  t1pks  t1pks: ((PKS_KS &amp; PKS_AT) or (ene_KS &amp;...   \n",
       "\n",
       "                                               genes  subclusters  \\\n",
       "0  AEK75490.1;AEK75491.1;AEK75492.1;AEK75493.1;AE...          nan   \n",
       "1  AHH99917.1;AHH99918.1;AHH99919.1;AHH99920.1;AH...          nan   \n",
       "2  BAD97694.1;BAD97695.1;BAD97696.1;BAD97697.1;BA...          nan   \n",
       "3  BAE71313.1;BAE71314.1;BAE71315.1;BAE71316.1;BA...          nan   \n",
       "4  AAS90092.1;AAS90093.1;AAS90110.1;AAS90111.1;AA...          nan   \n",
       "\n",
       "                            NRPSs/PKSs  \\\n",
       "0                                  NaN   \n",
       "1                                  NaN   \n",
       "2                    BAD97694;BAD97699   \n",
       "3  BAE71314;BAE71315;BAE71316;BAE71317   \n",
       "4  AAS90093;AAS90110;AAS90111;AAS90095   \n",
       "\n",
       "                                     signature_genes RiPPs  \\\n",
       "0  AEK75493;AEK75497;AEK75499;AEK75502;AEK75503;A...     -   \n",
       "1  AHH99918;AHH99919;AHH99920;AHH99921;AHH99922;A...     -   \n",
       "2                         BAD97694;BAD97696;BAD97699     -   \n",
       "3                BAE71314;BAE71315;BAE71316;BAE71317     -   \n",
       "4                AAS90093;AAS90110;AAS90111;AAS90095     -   \n",
       "\n",
       "   predicted_structure      monomers  \n",
       "0                  nan           NaN  \n",
       "1                  nan           NaN  \n",
       "2                  nan          (pk)  \n",
       "3                  nan  (pk) + (mal)  \n",
       "4                  nan  (pk) + (mal)  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "antismash = pd.read_csv('../data/mibig/antismash/classification.csv')\n",
    "antismash.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Alkaloid, NRP, Other, Polyketide, RiPP, Saccharide, Terpene\n",
    "\n",
    "AS_SUPERCLASSES = {\n",
    "    'nrps': 'NRP',\n",
    "    't1pks': 'Polyketide',\n",
    "    't2pks': 'Polyketide',\n",
    "    'terpene': 'Terpene',\n",
    "    'lantipeptide': 'RiPP',\n",
    "    'otherks': 'Polyketide',\n",
    "    'transatpks': 'Polyketide',\n",
    "    'indole': 'Alkaloid',\n",
    "    'bacteriocin': 'RiPP',\n",
    "    'amglyccycl': 'Saccharide',\n",
    "    't3pks': 'Polyketide',\n",
    "    'other': 'Other',\n",
    "    'thiopeptide': 'RiPP',\n",
    "    'oligosaccharide': 'Saccharide',\n",
    "    'lassopeptide': 'RiPP',\n",
    "    'nucleoside': 'Other',\n",
    "    'cyanobactin': 'RiPP',\n",
    "    'siderophore': None, # can be NRP and non-NRP (Other)\n",
    "    'arylpolyene': 'Polyketide',\n",
    "    'melanin': 'Other',\n",
    "    'butyrolactone': 'Other',\n",
    "    'blactam': None, # can be NRP and non-NRP (Other)\n",
    "    'ectoine': 'Other',\n",
    "    'ladderane': None, # not in JSON \n",
    "    'aminocoumarin': 'Other',\n",
    "    'phenazine': 'Other',\n",
    "    'PUFA': 'Polyketide',\n",
    "    'phosphonate': 'Other',\n",
    "    'sactipeptide': 'RiPP',\n",
    "    'linaridin': 'RiPP',\n",
    "    'bottromycin': 'RiPP',\n",
    "    'microviridin': 'RiPP',\n",
    "    'furan': 'Polyketide',\n",
    "    'hserlactone': None, # can be NRP, but is not a subclass\n",
    "    'head_to_tail': 'RiPP', \n",
    "    'glycocin': 'RiPP',\n",
    "    'proteusin': 'RiPP',\n",
    "    'microcin': 'RiPP',\n",
    "    'pbde': 'Other',\n",
    "    'phosphoglycolipid': 'Other',\n",
    "    'fused': None, # not in JSON\n",
    "    'ppysks': 'Other',\n",
    "    'resorcinol': 'Other'\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "contig_id\n",
       "BGC0000001.1                     [t1pks]\n",
       "BGC0000002.1    [oligosaccharide, t1pks]\n",
       "BGC0000003.1                     [t1pks]\n",
       "BGC0000004.1                     [t1pks]\n",
       "BGC0000006.1                     [t1pks]\n",
       "Name: classes, dtype: object"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "as_classes_sublists = antismash.groupby('contig_id')['classes'].apply(lambda lists: [c for classes in lists for c in classes.split(';')])\n",
    "as_classes_sublists.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "contig_id\n",
       "BGC0000001.1                [Polyketide]\n",
       "BGC0000002.1    [Saccharide, Polyketide]\n",
       "BGC0000003.1                [Polyketide]\n",
       "BGC0000004.1                [Polyketide]\n",
       "BGC0000006.1                [Polyketide]\n",
       "BGC0000007.1                [Polyketide]\n",
       "BGC0000008.1                [Polyketide]\n",
       "BGC0000009.1                [Polyketide]\n",
       "BGC0000011.1                [Polyketide]\n",
       "BGC0000012.1                [Polyketide]\n",
       "Name: classes, dtype: object"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "as_superclasses_sublists = as_classes_sublists.apply(lambda classes: [AS_SUPERCLASSES[c] for c in classes])\n",
    "as_superclasses_sublists[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BGCs with unknown superclass: 34\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "contig_id\n",
       "BGC0000182.1           [None, Polyketide, Polyketide]\n",
       "BGC0000213.1                       [Polyketide, None]\n",
       "BGC0000283.1                       [Saccharide, None]\n",
       "BGC0000288.1                              [None, NRP]\n",
       "BGC0000295.1                                   [None]\n",
       "BGC0000299.1                              [None, NRP]\n",
       "BGC0000314.1                              [None, NRP]\n",
       "BGC0000329.1                              [None, NRP]\n",
       "BGC0000429.1                  [NRP, Polyketide, None]\n",
       "BGC0000841.1                                   [None]\n",
       "BGC0000842.1                                   [None]\n",
       "BGC0000843.1                                   [None]\n",
       "BGC0000845.1                                   [None]\n",
       "BGC0000846.1                                   [None]\n",
       "BGC0000847.1                                   [None]\n",
       "BGC0000939.1                                   [None]\n",
       "BGC0000940.1                                   [None]\n",
       "BGC0000941.1                                   [None]\n",
       "BGC0000942.1                                   [None]\n",
       "BGC0000943.1                                   [None]\n",
       "BGC0000944.1                                   [None]\n",
       "BGC0000946.1                                   [None]\n",
       "BGC0000947.1                                   [None]\n",
       "BGC0000949.1                              [None, NRP]\n",
       "BGC0000961.1                  [None, Polyketide, NRP]\n",
       "BGC0001072.1    [Polyketide, Other, Polyketide, None]\n",
       "BGC0001148.1            [RiPP, Polyketide, None, NRP]\n",
       "BGC0001151.1                                   [None]\n",
       "BGC0001191.1                                   [None]\n",
       "BGC0001297.1                              [NRP, None]\n",
       "BGC0001308.1                       [Polyketide, None]\n",
       "BGC0001343.1                                   [None]\n",
       "BGC0001369.1                                   [None]\n",
       "BGC0001408.1                                   [None]\n",
       "Name: classes, dtype: object"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ambiguous_bgcs = as_superclasses_sublists[as_superclasses_sublists.apply(lambda classes: any(c is None for c in classes))]\n",
    "print('BGCs with unknown superclass:', len(ambiguous_bgcs))\n",
    "ambiguous_bgcs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1501"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "selected_bgc_ids = np.setdiff1d(all_ml_samples.index.values, ambiguous_bgcs.index.values)\n",
    "len(selected_bgc_ids)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Filter samples and responses by selected BGC IDs\n",
    "ml_samples = all_ml_samples.loc[selected_bgc_ids]\n",
    "ml_responses = all_ml_responses.loc[selected_bgc_ids]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1501, 7)\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Alkaloid</th>\n",
       "      <th>NRP</th>\n",
       "      <th>Other</th>\n",
       "      <th>Polyketide</th>\n",
       "      <th>RiPP</th>\n",
       "      <th>Saccharide</th>\n",
       "      <th>Terpene</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>contig_id</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>BGC0000001.1</th>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0000002.1</th>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0000003.1</th>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0000004.1</th>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BGC0000005.1</th>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Alkaloid   NRP  Other  Polyketide  RiPP  Saccharide  Terpene\n",
       "contig_id                                                                 \n",
       "BGC0000001.1     0.000 0.000  0.000       1.000 0.000       0.000    0.000\n",
       "BGC0000002.1     0.000 0.000  0.000       1.000 0.000       1.000    0.000\n",
       "BGC0000003.1     0.000 0.000  0.000       1.000 0.000       0.000    0.000\n",
       "BGC0000004.1     0.000 0.000  0.000       1.000 0.000       0.000    0.000\n",
       "BGC0000005.1     0.000 0.000  0.000       0.000 0.000       0.000    0.000"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Convert to binary lists\n",
    "as_responses_list = list(as_superclasses_sublists.apply(lambda classes: [int(c in classes) for c in ml_responses.columns]))\n",
    "# Create dataframe with columns = classes, rows = bgcs\n",
    "as_responses = pd.DataFrame(as_responses_list, index=as_superclasses_sublists.index, columns=ml_responses.columns)\n",
    "# Select BGCs based on IDs in ml_responses\n",
    "as_responses = as_responses.reindex(ml_responses.index).fillna(0)\n",
    "print(as_responses.shape)\n",
    "as_responses.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AntiSMASH predicted no class for 356 BGCs\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Index(['BGC0000005.1', 'BGC0000010.1', 'BGC0000015.2', 'BGC0000016.1',\n",
       "       'BGC0000019.1', 'BGC0000025.1', 'BGC0000088.1', 'BGC0000095.1',\n",
       "       'BGC0000095.11', 'BGC0000095.14',\n",
       "       ...\n",
       "       'BGC0001398.7', 'BGC0001398.8', 'BGC0001401.1', 'BGC0001401.2',\n",
       "       'BGC0001401.3', 'BGC0001401.4', 'BGC0001401.5', 'BGC0001410.1',\n",
       "       'BGC0001411.1', 'BGC0001412.1'],\n",
       "      dtype='object', name='contig_id', length=356)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "as_no_classes = as_responses[as_responses.sum(axis=1) == 0]\n",
    "print('AntiSMASH predicted no class for {} BGCs'.format(len(as_no_classes)))\n",
    "as_no_classes.index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Classes of BGCs not predicted by AntiSMASH:\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Alkaloid      11.000\n",
       "NRP           26.000\n",
       "Other         63.000\n",
       "Polyketide   103.000\n",
       "RiPP          40.000\n",
       "Saccharide    98.000\n",
       "Terpene       40.000\n",
       "dtype: float64"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('Classes of BGCs not predicted by AntiSMASH:')\n",
    "ml_responses.loc[as_no_classes.index].sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>AUC</th>\n",
       "      <th>Precision</th>\n",
       "      <th>Recall</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Alkaloid</th>\n",
       "      <td>0.785</td>\n",
       "      <td>0.434</td>\n",
       "      <td>0.590</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>NRP</th>\n",
       "      <td>0.915</td>\n",
       "      <td>0.939</td>\n",
       "      <td>0.852</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Other</th>\n",
       "      <td>0.671</td>\n",
       "      <td>0.594</td>\n",
       "      <td>0.370</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Polyketide</th>\n",
       "      <td>0.870</td>\n",
       "      <td>0.901</td>\n",
       "      <td>0.806</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>RiPP</th>\n",
       "      <td>0.897</td>\n",
       "      <td>0.958</td>\n",
       "      <td>0.799</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Saccharide</th>\n",
       "      <td>0.607</td>\n",
       "      <td>0.769</td>\n",
       "      <td>0.223</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Terpene</th>\n",
       "      <td>0.744</td>\n",
       "      <td>0.908</td>\n",
       "      <td>0.492</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             AUC  Precision  Recall\n",
       "Alkaloid   0.785      0.434   0.590\n",
       "NRP        0.915      0.939   0.852\n",
       "Other      0.671      0.594   0.370\n",
       "Polyketide 0.870      0.901   0.806\n",
       "RiPP       0.897      0.958   0.799\n",
       "Saccharide 0.607      0.769   0.223\n",
       "Terpene    0.744      0.908   0.492"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "as_scores = pd.DataFrame({\n",
    "    'AUC': metrics.roc_auc_score(ml_responses, as_responses, average=None), \n",
    "    'Precision': metrics.precision_score(ml_responses, as_responses, average=None), \n",
    "    'Recall': metrics.recall_score(ml_responses, as_responses, average=None)\n",
    "}, index=ml_responses.columns)\n",
    "as_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "AUC         0.784\n",
       "Precision   0.786\n",
       "Recall      0.590\n",
       "dtype: float64"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "as_scores.mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Random Forest: Multilabel classification results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Shuffle split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def create_model():\n",
    "    return RandomForestClassifier(n_estimators=100, random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    BGC0000001\n",
       "1    BGC0000002\n",
       "2    BGC0000003\n",
       "3    BGC0000004\n",
       "4    BGC0000005\n",
       "dtype: object"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "major_bgc_ids = pd.Series(selected_bgc_ids).apply(lambda contig_id: contig_id.split('.')[0])\n",
    "major_bgc_ids.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "BGC0000001    BGC0000001.1\n",
       "BGC0000002    BGC0000002.1\n",
       "BGC0000003    BGC0000003.1\n",
       "BGC0000004    BGC0000004.1\n",
       "BGC0000005    BGC0000005.1\n",
       "dtype: object"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "major_to_minor_ids = pd.Series(selected_bgc_ids, index=major_bgc_ids)\n",
    "major_to_minor_ids.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1355"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unique_major_ids = major_bgc_ids.unique()\n",
    "len(unique_major_ids)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train: 1199 Test: 302\n",
      "Train: 1194 Test: 307\n",
      "Train: 1220 Test: 281\n",
      "Train: 1192 Test: 309\n",
      "Train: 1199 Test: 302\n",
      "Accuracy (exact match) 0.6934747298208637\n",
      "AUC         0.792\n",
      "Precision   0.871\n",
      "Recall      0.607\n",
      "dtype: float64\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>AUC</th>\n",
       "      <th>Precision</th>\n",
       "      <th>Recall</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Class</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Alkaloid</th>\n",
       "      <td>0.607</td>\n",
       "      <td>0.733</td>\n",
       "      <td>0.216</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>NRP</th>\n",
       "      <td>0.907</td>\n",
       "      <td>0.904</td>\n",
       "      <td>0.850</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Other</th>\n",
       "      <td>0.583</td>\n",
       "      <td>0.876</td>\n",
       "      <td>0.171</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Polyketide</th>\n",
       "      <td>0.903</td>\n",
       "      <td>0.876</td>\n",
       "      <td>0.898</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>RiPP</th>\n",
       "      <td>0.907</td>\n",
       "      <td>0.935</td>\n",
       "      <td>0.823</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Saccharide</th>\n",
       "      <td>0.811</td>\n",
       "      <td>0.906</td>\n",
       "      <td>0.631</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Terpene</th>\n",
       "      <td>0.824</td>\n",
       "      <td>0.867</td>\n",
       "      <td>0.658</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             AUC  Precision  Recall\n",
       "Class                              \n",
       "Alkaloid   0.607      0.733   0.216\n",
       "NRP        0.907      0.904   0.850\n",
       "Other      0.583      0.876   0.171\n",
       "Polyketide 0.903      0.876   0.898\n",
       "RiPP       0.907      0.935   0.823\n",
       "Saccharide 0.811      0.906   0.631\n",
       "Terpene    0.824      0.867   0.658"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "splitter = KFold(n_splits=5, shuffle=True, random_state=0)\n",
    "\n",
    "scores = []\n",
    "accuracies = []\n",
    "mc_y_test = []\n",
    "mc_pred = []\n",
    "for id_train_idx, id_test_idx in splitter.split(unique_major_ids):\n",
    "    train_major_ids, test_major_ids = unique_major_ids[id_train_idx], unique_major_ids[id_test_idx]\n",
    "    train_minor_ids, test_minor_ids = major_to_minor_ids.loc[train_major_ids], major_to_minor_ids.loc[test_major_ids]\n",
    "    X_train, X_test = ml_samples.loc[train_minor_ids].values, ml_samples.loc[test_minor_ids].values\n",
    "    y_train, y_test = ml_responses.loc[train_minor_ids].values, ml_responses.loc[test_minor_ids].values\n",
    "    print('Train:', len(X_train), 'Test:', len(X_test))\n",
    "    split_model = create_model()\n",
    "    # fitting the model\n",
    "    split_model.fit(X_train, y_train)\n",
    "    # predict the response\n",
    "    pred = split_model.predict(X_test)\n",
    "    mc_y_test += [';'.join(ml_responses.columns[resp.astype(np.bool)]) for resp in y_test]\n",
    "    mc_pred += [';'.join(ml_responses.columns[resp.astype(np.bool)]) for resp in pred]\n",
    "    aucs = metrics.roc_auc_score(y_test, pred, average=None)\n",
    "    precisions = metrics.precision_score(y_test, pred, average=None)\n",
    "    recalls = metrics.recall_score(y_test, pred, average=None)\n",
    "    accuracy = metrics.accuracy_score(y_test, pred)\n",
    "    accuracies.append(accuracy)\n",
    "    scores += [{'Precision': p, 'Recall': r, 'AUC': a, 'Class': c} for p,r,a,c in zip(precisions, recalls, aucs, ml_responses.columns)]\n",
    "    \n",
    "scores = pd.DataFrame(scores).groupby('Class').mean()\n",
    "\n",
    "print('Accuracy (exact match)', np.mean(accuracies))\n",
    "print(scores.mean())\n",
    "scores\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th colspan=\"3\" halign=\"left\">RandomForest</th>\n",
       "      <th colspan=\"3\" halign=\"left\">AntiSMASH</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th># samples</th>\n",
       "      <th>AUC</th>\n",
       "      <th>Precision</th>\n",
       "      <th>Recall</th>\n",
       "      <th>AUC</th>\n",
       "      <th>Precision</th>\n",
       "      <th>Recall</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Class</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>Polyketide</th>\n",
       "      <td>644.000</td>\n",
       "      <td>0.903</td>\n",
       "      <td>0.876</td>\n",
       "      <td>0.898</td>\n",
       "      <td>0.870</td>\n",
       "      <td>0.901</td>\n",
       "      <td>0.806</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>NRP</th>\n",
       "      <td>433.000</td>\n",
       "      <td>0.907</td>\n",
       "      <td>0.904</td>\n",
       "      <td>0.850</td>\n",
       "      <td>0.915</td>\n",
       "      <td>0.939</td>\n",
       "      <td>0.852</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>RiPP</th>\n",
       "      <td>199.000</td>\n",
       "      <td>0.907</td>\n",
       "      <td>0.935</td>\n",
       "      <td>0.823</td>\n",
       "      <td>0.897</td>\n",
       "      <td>0.958</td>\n",
       "      <td>0.799</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Saccharide</th>\n",
       "      <td>179.000</td>\n",
       "      <td>0.811</td>\n",
       "      <td>0.906</td>\n",
       "      <td>0.631</td>\n",
       "      <td>0.607</td>\n",
       "      <td>0.769</td>\n",
       "      <td>0.223</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Other</th>\n",
       "      <td>154.000</td>\n",
       "      <td>0.583</td>\n",
       "      <td>0.876</td>\n",
       "      <td>0.171</td>\n",
       "      <td>0.671</td>\n",
       "      <td>0.594</td>\n",
       "      <td>0.370</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Terpene</th>\n",
       "      <td>120.000</td>\n",
       "      <td>0.824</td>\n",
       "      <td>0.867</td>\n",
       "      <td>0.658</td>\n",
       "      <td>0.744</td>\n",
       "      <td>0.908</td>\n",
       "      <td>0.492</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Alkaloid</th>\n",
       "      <td>39.000</td>\n",
       "      <td>0.607</td>\n",
       "      <td>0.733</td>\n",
       "      <td>0.216</td>\n",
       "      <td>0.785</td>\n",
       "      <td>0.434</td>\n",
       "      <td>0.590</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Average</th>\n",
       "      <td>252.571</td>\n",
       "      <td>0.792</td>\n",
       "      <td>0.871</td>\n",
       "      <td>0.607</td>\n",
       "      <td>0.784</td>\n",
       "      <td>0.786</td>\n",
       "      <td>0.590</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     RandomForest                  AntiSMASH                 \n",
       "           # samples          AUC Precision Recall       AUC Precision Recall\n",
       "Class                                                                        \n",
       "Polyketide   644.000        0.903     0.876  0.898     0.870     0.901  0.806\n",
       "NRP          433.000        0.907     0.904  0.850     0.915     0.939  0.852\n",
       "RiPP         199.000        0.907     0.935  0.823     0.897     0.958  0.799\n",
       "Saccharide   179.000        0.811     0.906  0.631     0.607     0.769  0.223\n",
       "Other        154.000        0.583     0.876  0.171     0.671     0.594  0.370\n",
       "Terpene      120.000        0.824     0.867  0.658     0.744     0.908  0.492\n",
       "Alkaloid      39.000        0.607     0.733  0.216     0.785     0.434  0.590\n",
       "Average      252.571        0.792     0.871  0.607     0.784     0.786  0.590"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "merged_scores = pd.concat([scores, as_scores[scores.columns]], axis=1)\n",
    "merged_scores.columns = pd.MultiIndex.from_product([['RandomForest','AntiSMASH'], scores.columns.values])\n",
    "merged_scores.insert(0, ('', '# samples'), ml_responses.sum().astype(np.int))\n",
    "merged_scores = merged_scores.sort_values(by=('', '# samples'), ascending=False)\n",
    "merged_scores = merged_scores.append(pd.Series(merged_scores.mean(), name='Average'))\n",
    "merged_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x112e5a4e0>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x112d944e0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ax = merged_scores[[('RandomForest','AUC'),('AntiSMASH','AUC')]].plot.barh()\n",
    "ax.legend(bbox_to_anchor=(1,1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import itertools\n",
    "\n",
    "def plot_confusion_matrix(cm, classes,\n",
    "                          normalize=False,\n",
    "                          title='Confusion matrix',\n",
    "                          cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting `normalize=True`.\n",
    "    \"\"\"\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "        print(\"Normalized confusion matrix\")\n",
    "    else:\n",
    "        print('Confusion matrix, without normalization')\n",
    "\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=90)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    fmt = '.2f' if normalize else 'd'\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, format(cm[i, j], fmt),\n",
    "                 horizontalalignment=\"center\",\n",
    "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')\n",
    "    return plt.gcf()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix, without normalization\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11274a2b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "mc_all_labels = np.unique(np.concatenate([mc_y_test, mc_pred]))\n",
    "matrix = metrics.confusion_matrix(mc_y_test, mc_pred, labels=mc_all_labels)\n",
    "matrix = (matrix / 10).astype(np.int)\n",
    "fig, ax = plt.subplots(1, 1, figsize=(14, 10))\n",
    "fig = plot_confusion_matrix(matrix, mc_all_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "fig.savefig('../data/figures/paper/supplementary/S8.pdf', bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Final model training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=1, min_samples_split=2,\n",
       "            min_weight_fraction_leaf=0.0, n_estimators=100, n_jobs=1,\n",
       "            oob_score=False, random_state=0, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = RandomForestClassifier(n_estimators=100, random_state=0)\n",
    "model.fit(all_ml_samples.values, all_ml_responses.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# TODO: save list of features (or feature transformer) along with model\n",
    "#with open('../data/models/trained/class-domain_set-rf.pickle', 'wb') as f:\n",
    "#    pickle.dump(model, f)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def plot_importance(model, label='Global', **kwargs):\n",
    "    importance = pd.DataFrame({'importance': model.feature_importances_, 'pfam_id': ml_samples.columns.values})\n",
    "    importance = importance.set_index('pfam_id')['importance'].sort_values()\n",
    "    ax = importance[-10:].plot.barh(**kwargs)\n",
    "    ax.set_xlabel('{} feature importance'.format(label))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x114024d30>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Per-class feature importance\n",
    "\n",
    "fig, axes = plt.subplots(2, 4, figsize=(14, 6))\n",
    "\n",
    "for i, cls in enumerate(ml_responses.columns):\n",
    "    class_model = RandomForestClassifier(n_estimators=100, random_state=0)\n",
    "    class_model.fit(all_ml_samples.values, all_ml_responses[cls].values)\n",
    "    plot_importance(class_model, label=cls, ax=axes[i // 4][i % 4])\n",
    "\n",
    "plot_importance(model, ax=axes[-1][-1], color='red')\n",
    "plt.tight_layout()\n",
    "fig.savefig('../data/figures/paper/supplementary/S9.pdf', bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Candidate class prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def read_vectors_to_predict(path, train_columns):\n",
    "    vectors = pd.read_pickle(path)\n",
    "    for col in set(train_columns).difference(vectors.columns):\n",
    "        vectors[col] = 0\n",
    "    return vectors[train_columns.values]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model_path = '128lstm-100pfamdim-8pfamiter-posweighted-neg-10k-fpr2'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>pfam_id</th>\n",
       "      <th>PF00004</th>\n",
       "      <th>PF00005</th>\n",
       "      <th>PF00006</th>\n",
       "      <th>PF00009</th>\n",
       "      <th>PF00010</th>\n",
       "      <th>PF00011</th>\n",
       "      <th>PF00012</th>\n",
       "      <th>PF00013</th>\n",
       "      <th>PF00015</th>\n",
       "      <th>PF00022</th>\n",
       "      <th>...</th>\n",
       "      <th>PF17387</th>\n",
       "      <th>PF17389</th>\n",
       "      <th>PF17390</th>\n",
       "      <th>PF17391</th>\n",
       "      <th>PF17392</th>\n",
       "      <th>PF17414</th>\n",
       "      <th>PF17425</th>\n",
       "      <th>PF17508</th>\n",
       "      <th>PF17584</th>\n",
       "      <th>PF17608</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>contig_id</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>000749207dc6a8d25a7b67282a2fe375</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>00244b39ef89b1f18ccf4da018657383</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0024a1c619f0dba6e8420c60b6c815e7</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>002d6748dad4145b1f09ca9dc93ead44</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>003041bd79286a337bdd94c332198c33</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 3294 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "pfam_id                           PF00004  PF00005  PF00006  PF00009  PF00010  \\\n",
       "contig_id                                                                       \n",
       "000749207dc6a8d25a7b67282a2fe375        0        0        0        0    0.000   \n",
       "00244b39ef89b1f18ccf4da018657383        0        0        0        0    0.000   \n",
       "0024a1c619f0dba6e8420c60b6c815e7        0        0        0        0    0.000   \n",
       "002d6748dad4145b1f09ca9dc93ead44        0        0        0        0    0.000   \n",
       "003041bd79286a337bdd94c332198c33        0        0        0        0    0.000   \n",
       "\n",
       "pfam_id                           PF00011  PF00012  PF00013  PF00015  PF00022  \\\n",
       "contig_id                                                                       \n",
       "000749207dc6a8d25a7b67282a2fe375        0        0        0        0    0.000   \n",
       "00244b39ef89b1f18ccf4da018657383        0        0        0        0    0.000   \n",
       "0024a1c619f0dba6e8420c60b6c815e7        0        0        0        0    0.000   \n",
       "002d6748dad4145b1f09ca9dc93ead44        0        0        0        0    0.000   \n",
       "003041bd79286a337bdd94c332198c33        0        0        0        0    0.000   \n",
       "\n",
       "pfam_id                            ...     PF17387  PF17389  PF17390  PF17391  \\\n",
       "contig_id                          ...                                          \n",
       "000749207dc6a8d25a7b67282a2fe375   ...           0        0        0    0.000   \n",
       "00244b39ef89b1f18ccf4da018657383   ...           0        0        0    0.000   \n",
       "0024a1c619f0dba6e8420c60b6c815e7   ...           0        0        0    0.000   \n",
       "002d6748dad4145b1f09ca9dc93ead44   ...           0        0        0    0.000   \n",
       "003041bd79286a337bdd94c332198c33   ...           0        0        0    0.000   \n",
       "\n",
       "pfam_id                           PF17392  PF17414  PF17425  PF17508  PF17584  \\\n",
       "contig_id                                                                       \n",
       "000749207dc6a8d25a7b67282a2fe375    0.000    0.000        0        0    0.000   \n",
       "00244b39ef89b1f18ccf4da018657383    0.000    0.000        0        0    0.000   \n",
       "0024a1c619f0dba6e8420c60b6c815e7    0.000    0.000        0        0    0.000   \n",
       "002d6748dad4145b1f09ca9dc93ead44    0.000    0.000        0        0    0.000   \n",
       "003041bd79286a337bdd94c332198c33    0.000    0.000        0        0    0.000   \n",
       "\n",
       "pfam_id                           PF17608  \n",
       "contig_id                                  \n",
       "000749207dc6a8d25a7b67282a2fe375    0.000  \n",
       "00244b39ef89b1f18ccf4da018657383    0.000  \n",
       "0024a1c619f0dba6e8420c60b6c815e7    0.000  \n",
       "002d6748dad4145b1f09ca9dc93ead44    0.000  \n",
       "003041bd79286a337bdd94c332198c33    0.000  \n",
       "\n",
       "[5 rows x 3294 columns]"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "candidates = read_vectors_to_predict(\"../data/bacteria/candidates/{}/domain_set.pkl\".format(model_path), ml_samples.columns)\n",
    "candidates.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Alkaloid</th>\n",
       "      <th>NRP</th>\n",
       "      <th>Other</th>\n",
       "      <th>Polyketide</th>\n",
       "      <th>RiPP</th>\n",
       "      <th>Saccharide</th>\n",
       "      <th>Terpene</th>\n",
       "      <th>classes</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>contig_id</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>000749207dc6a8d25a7b67282a2fe375</th>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.060</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.940</td>\n",
       "      <td>0.000</td>\n",
       "      <td>Saccharide</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>00244b39ef89b1f18ccf4da018657383</th>\n",
       "      <td>0.010</td>\n",
       "      <td>0.020</td>\n",
       "      <td>0.180</td>\n",
       "      <td>0.250</td>\n",
       "      <td>0.020</td>\n",
       "      <td>0.609</td>\n",
       "      <td>0.010</td>\n",
       "      <td>Saccharide</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0024a1c619f0dba6e8420c60b6c815e7</th>\n",
       "      <td>0.000</td>\n",
       "      <td>0.920</td>\n",
       "      <td>0.020</td>\n",
       "      <td>0.030</td>\n",
       "      <td>0.030</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.000</td>\n",
       "      <td>NRP</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>002d6748dad4145b1f09ca9dc93ead44</th>\n",
       "      <td>0.000</td>\n",
       "      <td>0.040</td>\n",
       "      <td>0.050</td>\n",
       "      <td>0.076</td>\n",
       "      <td>0.044</td>\n",
       "      <td>0.780</td>\n",
       "      <td>0.010</td>\n",
       "      <td>Saccharide</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>003041bd79286a337bdd94c332198c33</th>\n",
       "      <td>0.010</td>\n",
       "      <td>0.000</td>\n",
       "      <td>0.030</td>\n",
       "      <td>0.070</td>\n",
       "      <td>0.010</td>\n",
       "      <td>0.870</td>\n",
       "      <td>0.010</td>\n",
       "      <td>Saccharide</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                  Alkaloid   NRP  Other  Polyketide  RiPP  \\\n",
       "contig_id                                                                   \n",
       "000749207dc6a8d25a7b67282a2fe375     0.000 0.000  0.060       0.000 0.000   \n",
       "00244b39ef89b1f18ccf4da018657383     0.010 0.020  0.180       0.250 0.020   \n",
       "0024a1c619f0dba6e8420c60b6c815e7     0.000 0.920  0.020       0.030 0.030   \n",
       "002d6748dad4145b1f09ca9dc93ead44     0.000 0.040  0.050       0.076 0.044   \n",
       "003041bd79286a337bdd94c332198c33     0.010 0.000  0.030       0.070 0.010   \n",
       "\n",
       "                                  Saccharide  Terpene     classes  \n",
       "contig_id                                                          \n",
       "000749207dc6a8d25a7b67282a2fe375       0.940    0.000  Saccharide  \n",
       "00244b39ef89b1f18ccf4da018657383       0.609    0.010  Saccharide  \n",
       "0024a1c619f0dba6e8420c60b6c815e7       0.000    0.000         NRP  \n",
       "002d6748dad4145b1f09ca9dc93ead44       0.780    0.010  Saccharide  \n",
       "003041bd79286a337bdd94c332198c33       0.870    0.010  Saccharide  "
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "columns = ml_responses.columns\n",
    "results = np.array([r[:,1] for r in model.predict_proba(candidates.values)]).transpose()\n",
    "predictions = pd.DataFrame(results, index=candidates.index, columns=columns)\n",
    "predictions['classes'] = [';'.join(columns[x >= 0.5]) for x in results]\n",
    "predictions.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "predictions.to_csv('../data/bacteria/candidates/{}/classes/class-domain_set-rf/classes.csv'.format(model_path))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}