FeatureImportances.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.metrics import pairwise_distances_argmin_min\n",
    "from sklearn.inspection import permutation_importance\n",
    "from sklearn.cluster import KMeans\n",
    "\n",
    "from descriptor_calculator import * # Module for molecular fingerprint & descriptor calculations\n",
    "from training import * # Module for model training and cross validation\n",
    "from rdkit.Chem import Draw # Visualize representative test set structures"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>ID_Number</th>\n",
       "      <th>Canonical_Smiles</th>\n",
       "      <th>Round_Number</th>\n",
       "      <th>Max_LCAP</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>4897720</td>\n",
       "      <td>CCCCn1ccnc1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>18461035</td>\n",
       "      <td>CC(=O)c1ocnc1</td>\n",
       "      <td>0</td>\n",
       "      <td>72</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1762470</td>\n",
       "      <td>Fc1ccc(cc1)c2ocnc2</td>\n",
       "      <td>0</td>\n",
       "      <td>87</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>16916940</td>\n",
       "      <td>Brc1ccc(cc1)c2cncs2</td>\n",
       "      <td>0</td>\n",
       "      <td>78</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1040111</td>\n",
       "      <td>Fc1cc(ccc1[C@@H]2CCc3cncn23)C#N</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   ID_Number                 Canonical_Smiles  Round_Number  Max_LCAP\n",
       "0    4897720                      CCCCn1ccnc1             0         0\n",
       "1   18461035                    CC(=O)c1ocnc1             0        72\n",
       "2    1762470               Fc1ccc(cc1)c2ocnc2             0        87\n",
       "3   16916940              Brc1ccc(cc1)c2cncs2             0        78\n",
       "4    1040111  Fc1cc(ccc1[C@@H]2CCc3cncn23)C#N             0         0"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load and clean raw data\n",
    "df_raw = pd.read_csv('./data/data.csv')\n",
    "\n",
    "# Filter out rows without Max_LCAP values\n",
    "df = df_raw[df_raw['Max_LCAP'].notna()]\n",
    "df = df.reset_index(drop=True)\n",
    "df['Round_Number'] = df['Round_Number'].apply(lambda x: int(str(x)[-1]))\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are 3747 imported MIX features from internal platform.\n"
     ]
    },
    {
     "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>ID_Number</th>\n",
       "      <th>Canonical_Smiles</th>\n",
       "      <th>Round_Number</th>\n",
       "      <th>Max_LCAP</th>\n",
       "      <th>BrX1sp3BrX1sp305</th>\n",
       "      <th>CX1sp3BrX1sp303</th>\n",
       "      <th>CX1sp3BrX1sp304</th>\n",
       "      <th>CX1sp3BrX1sp305</th>\n",
       "      <th>CX1sp3BrX1sp306</th>\n",
       "      <th>CX1sp3BrX1sp307</th>\n",
       "      <th>...</th>\n",
       "      <th>X4sp3ClX1sp311</th>\n",
       "      <th>X4sp3NX2sp205</th>\n",
       "      <th>X4sp3NX2sp206</th>\n",
       "      <th>X4sp3NX2sp207</th>\n",
       "      <th>X4sp3NX2sp209</th>\n",
       "      <th>X4sp3NX2sp211</th>\n",
       "      <th>X4sp3NX3sp305</th>\n",
       "      <th>X4sp3OX1sp211</th>\n",
       "      <th>X4sp3OX2sp303</th>\n",
       "      <th>X4sp3SX2sp304</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>4897720</td>\n",
       "      <td>CCCCn1ccnc1</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>1</th>\n",
       "      <td>18461035</td>\n",
       "      <td>CC(=O)c1ocnc1</td>\n",
       "      <td>0</td>\n",
       "      <td>72</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>2</th>\n",
       "      <td>1762470</td>\n",
       "      <td>Fc1ccc(cc1)c2ocnc2</td>\n",
       "      <td>0</td>\n",
       "      <td>87</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>3</th>\n",
       "      <td>16916940</td>\n",
       "      <td>Brc1ccc(cc1)c2cncs2</td>\n",
       "      <td>0</td>\n",
       "      <td>78</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>4</th>\n",
       "      <td>1040111</td>\n",
       "      <td>Fc1cc(ccc1[C@@H]2CCc3cncn23)C#N</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>5 rows × 3751 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   ID_Number                 Canonical_Smiles  Round_Number  Max_LCAP  \\\n",
       "0    4897720                      CCCCn1ccnc1             0         0   \n",
       "1   18461035                    CC(=O)c1ocnc1             0        72   \n",
       "2    1762470               Fc1ccc(cc1)c2ocnc2             0        87   \n",
       "3   16916940              Brc1ccc(cc1)c2cncs2             0        78   \n",
       "4    1040111  Fc1cc(ccc1[C@@H]2CCc3cncn23)C#N             0         0   \n",
       "\n",
       "   BrX1sp3BrX1sp305  CX1sp3BrX1sp303  CX1sp3BrX1sp304  CX1sp3BrX1sp305  \\\n",
       "0                 0                0                0                0   \n",
       "1                 0                0                0                0   \n",
       "2                 0                0                0                0   \n",
       "3                 0                0                0                0   \n",
       "4                 0                0                0                0   \n",
       "\n",
       "   CX1sp3BrX1sp306  CX1sp3BrX1sp307  ...  X4sp3ClX1sp311  X4sp3NX2sp205  \\\n",
       "0                0                0  ...               0              0   \n",
       "1                0                0  ...               0              0   \n",
       "2                0                0  ...               0              0   \n",
       "3                0                0  ...               0              0   \n",
       "4                0                0  ...               0              0   \n",
       "\n",
       "   X4sp3NX2sp206  X4sp3NX2sp207  X4sp3NX2sp209  X4sp3NX2sp211  X4sp3NX3sp305  \\\n",
       "0              0              0              0              0              0   \n",
       "1              0              0              0              0              0   \n",
       "2              0              0              0              0              0   \n",
       "3              0              0              0              0              0   \n",
       "4              0              0              0              0              0   \n",
       "\n",
       "   X4sp3OX1sp211  X4sp3OX2sp303  X4sp3SX2sp304  \n",
       "0              0              0              0  \n",
       "1              0              0              0  \n",
       "2              0              0              0  \n",
       "3              0              0              0  \n",
       "4              0              0              0  \n",
       "\n",
       "[5 rows x 3751 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load descriptors calculated using internal platform\n",
    "mix_desc_df = pd.read_csv('./data/mix_des.csv')\n",
    "mix_desc_df = mix_desc_df.dropna(axis=1)\n",
    "mix_features = mix_desc_df.columns.tolist()[1:]\n",
    "print(f'There are {len(mix_features)} imported MIX features from internal platform.')\n",
    "\n",
    "# Merge the mix descriptors with the main dataframe on 'ID_Number'\n",
    "df = df.merge(mix_desc_df, on='ID_Number')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are 12 imported DFT features.\n"
     ]
    }
   ],
   "source": [
    "# Load descriptors calculated from DFT-optimized geometries\n",
    "df_dft = pd.read_csv('./data/data_dft.csv')\n",
    "\n",
    "# Define DFT features\n",
    "dft_features = [\n",
    "    'N (eV)', 'HOMO (eV)', 'LUMO (eV)', 'Gap (eV)', \n",
    "    'Omega', 'dqG', 'Bmin', 'Bmax', \n",
    "    'L', 'PMI_X', 'PMI_Y', 'PMI_Z'\n",
    "]\n",
    "\n",
    "print(f'There are {len(dft_features)} imported DFT features.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Calculate molecular fingerprints and descriptors\n",
    "rdkit_desc = calc_rdkit_desc(df)  # Compute RDKit descriptors\n",
    "rdkit_fp = calc_rdkit_fp(df, radius=2, nBits=2048)  # Compute molecular fingerprints\n",
    "\n",
    "# Combine descriptors and fingerprints into a single dataframe\n",
    "desc_fp = rdkit_desc.join(rdkit_fp)\n",
    "\n",
    "# Optionally save combined dataframe of descriptors and fingerprints to a CSV file\n",
    "#desc_fp.to_csv('./data/rdkit_desc_fp.csv', index=False)\n",
    "\n",
    "rdkit_features = desc_fp.columns.tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>Canonical_Smiles</th>\n",
       "      <th>Round_Number</th>\n",
       "      <th>Max_LCAP</th>\n",
       "      <th>ID_Number</th>\n",
       "      <th>BrX1sp3BrX1sp305</th>\n",
       "      <th>CX1sp3BrX1sp303</th>\n",
       "      <th>CX1sp3BrX1sp304</th>\n",
       "      <th>CX1sp3BrX1sp305</th>\n",
       "      <th>CX1sp3BrX1sp306</th>\n",
       "      <th>CX1sp3BrX1sp307</th>\n",
       "      <th>...</th>\n",
       "      <th>LUMO (eV)</th>\n",
       "      <th>Gap (eV)</th>\n",
       "      <th>Omega</th>\n",
       "      <th>dqG</th>\n",
       "      <th>Bmin</th>\n",
       "      <th>Bmax</th>\n",
       "      <th>L</th>\n",
       "      <th>PMI_X</th>\n",
       "      <th>PMI_Y</th>\n",
       "      <th>PMI_Z</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>CCCCn1ccnc1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4897720</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>1.55</td>\n",
       "      <td>9.23</td>\n",
       "      <td>0.018667</td>\n",
       "      <td>66.200171</td>\n",
       "      <td>1.71</td>\n",
       "      <td>7.91</td>\n",
       "      <td>5.10</td>\n",
       "      <td>15.11284</td>\n",
       "      <td>119.05900</td>\n",
       "      <td>132.05374</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>CC(=O)c1ocnc1</td>\n",
       "      <td>0</td>\n",
       "      <td>72</td>\n",
       "      <td>18461035</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>-0.79</td>\n",
       "      <td>7.93</td>\n",
       "      <td>0.052357</td>\n",
       "      <td>52.809485</td>\n",
       "      <td>1.74</td>\n",
       "      <td>4.66</td>\n",
       "      <td>7.15</td>\n",
       "      <td>17.18341</td>\n",
       "      <td>52.33765</td>\n",
       "      <td>69.00428</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Fc1ccc(cc1)c2ocnc2</td>\n",
       "      <td>0</td>\n",
       "      <td>87</td>\n",
       "      <td>1762470</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>-0.34</td>\n",
       "      <td>7.07</td>\n",
       "      <td>0.039040</td>\n",
       "      <td>56.561304</td>\n",
       "      <td>1.65</td>\n",
       "      <td>6.65</td>\n",
       "      <td>8.40</td>\n",
       "      <td>23.60169</td>\n",
       "      <td>178.26557</td>\n",
       "      <td>201.86727</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Brc1ccc(cc1)c2cncs2</td>\n",
       "      <td>0</td>\n",
       "      <td>78</td>\n",
       "      <td>16916940</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>-0.88</td>\n",
       "      <td>6.60</td>\n",
       "      <td>0.048713</td>\n",
       "      <td>55.462471</td>\n",
       "      <td>1.71</td>\n",
       "      <td>5.40</td>\n",
       "      <td>10.80</td>\n",
       "      <td>29.47762</td>\n",
       "      <td>386.11696</td>\n",
       "      <td>415.59459</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Fc1cc(ccc1[C@@H]2CCc3cncn23)C#N</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1040111</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>-0.95</td>\n",
       "      <td>6.63</td>\n",
       "      <td>0.050420</td>\n",
       "      <td>63.966738</td>\n",
       "      <td>1.78</td>\n",
       "      <td>7.98</td>\n",
       "      <td>5.95</td>\n",
       "      <td>87.74469</td>\n",
       "      <td>306.39359</td>\n",
       "      <td>323.60486</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 6019 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                  Canonical_Smiles  Round_Number  Max_LCAP  ID_Number  \\\n",
       "0                      CCCCn1ccnc1             0         0    4897720   \n",
       "1                    CC(=O)c1ocnc1             0        72   18461035   \n",
       "2               Fc1ccc(cc1)c2ocnc2             0        87    1762470   \n",
       "3              Brc1ccc(cc1)c2cncs2             0        78   16916940   \n",
       "4  Fc1cc(ccc1[C@@H]2CCc3cncn23)C#N             0         0    1040111   \n",
       "\n",
       "   BrX1sp3BrX1sp305  CX1sp3BrX1sp303  CX1sp3BrX1sp304  CX1sp3BrX1sp305  \\\n",
       "0                 0                0                0                0   \n",
       "1                 0                0                0                0   \n",
       "2                 0                0                0                0   \n",
       "3                 0                0                0                0   \n",
       "4                 0                0                0                0   \n",
       "\n",
       "   CX1sp3BrX1sp306  CX1sp3BrX1sp307  ...  LUMO (eV)  Gap (eV)     Omega  \\\n",
       "0                0                0  ...       1.55      9.23  0.018667   \n",
       "1                0                0  ...      -0.79      7.93  0.052357   \n",
       "2                0                0  ...      -0.34      7.07  0.039040   \n",
       "3                0                0  ...      -0.88      6.60  0.048713   \n",
       "4                0                0  ...      -0.95      6.63  0.050420   \n",
       "\n",
       "         dqG  Bmin  Bmax      L     PMI_X      PMI_Y      PMI_Z  \n",
       "0  66.200171  1.71  7.91   5.10  15.11284  119.05900  132.05374  \n",
       "1  52.809485  1.74  4.66   7.15  17.18341   52.33765   69.00428  \n",
       "2  56.561304  1.65  6.65   8.40  23.60169  178.26557  201.86727  \n",
       "3  55.462471  1.71  5.40  10.80  29.47762  386.11696  415.59459  \n",
       "4  63.966738  1.78  7.98   5.95  87.74469  306.39359  323.60486  \n",
       "\n",
       "[5 rows x 6019 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Combine all features into a single dataframe\n",
    "df_clean = pd.concat(\n",
    "    [\n",
    "        df[['Canonical_Smiles', 'Round_Number', 'Max_LCAP', 'ID_Number'] + mix_features], \n",
    "        desc_fp\n",
    "    ], \n",
    "    axis=1\n",
    ")\n",
    "\n",
    "# Merge on 'SMILES' to include DFT features\n",
    "df_clean = df_clean.merge(df_dft[dft_features + ['Canonical_Smiles']], on='Canonical_Smiles')\n",
    "\n",
    "# Optionally save all features to a CSV file\n",
    "#df_clean.to_csv('all_features.csv', index=False)\n",
    "\n",
    "df_clean.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Calculate permutation feature importance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Defining a representative test set using k-means clustering for PFI analysis "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# For representative test set generation, define the round for clustering\n",
    "r = 7 # Use all data for clustering to choose a representative test set for PFI calculation\n",
    "df_to_cluster = df_clean[df_clean.Round_Number <= r]\n",
    "\n",
    "# Determine the number of clusters\n",
    "n_clusters = int(0.2 * len(df_to_cluster))\n",
    "\n",
    "# Prepare features for K-means clustering\n",
    "X = df_to_cluster[rdkit_features]\n",
    "\n",
    "# Perform K-means clustering\n",
    "k_means = KMeans(n_clusters=n_clusters, init='random', n_init=10, random_state=0)\n",
    "kmeans_y_pred = k_means.fit_predict(X)\n",
    "\n",
    "df_all_clusters = pd.DataFrame({\n",
    "    'smiles': df_to_cluster.smiles,\n",
    "    'KMeansCluster': kmeans_y_pred\n",
    "})\n",
    "\n",
    "# Select ligands closest to the centroid of each cluster\n",
    "selected_ligands = []\n",
    "for n in range(n_clusters):\n",
    "    # Find all members that belong to cluster n\n",
    "    members = df_all_clusters[df_all_clusters['KMeansCluster'] == n].index.tolist()\n",
    "    \n",
    "    # Find the closest member to the center of cluster n\n",
    "    closest_ind = pairwise_distances_argmin_min(k_means.cluster_centers_, X.iloc[members])[0][n]\n",
    "    selected_ligands.append(members[closest_ind])\n",
    "\n",
    "# Mark selected ligands as centroids\n",
    "df_all_clusters['Centroid'] = 'No'\n",
    "for i in selected_ligands:\n",
    "    df_all_clusters.at[i, 'Centroid'] = 'Yes'\n",
    "\n",
    "# Fill missing values in the 'Centroid' column\n",
    "df_all_clusters['Centroid'].fillna('No', inplace=True)\n",
    "\n",
    "# Create training and testing DataFrames\n",
    "train_df = df_all_clusters[df_all_clusters['Centroid'] == 'No'].merge(df_clean, on='smiles')\n",
    "kmeans_test_df = df_all_clusters[df_all_clusters['Centroid'] == 'Yes'].merge(df_clean, on='smiles')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total run time is 2.80 seconds\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Define features and model for PFI\n",
    "features = rdkit_features\n",
    "rf = RandomForestRegressor()\n",
    "\n",
    "# Train model and evaluate performance\n",
    "trained_rf, performance, preds = model_training(\n",
    "    data=train_df, \n",
    "    model=rf, \n",
    "    split_method=None, # Use k-means representative test set instead\n",
    "    descriptors=features, \n",
    "    test_df=kmeans_test_df\n",
    ")\n",
    "\n",
    "# Prepare training and testing data\n",
    "X_test, y_test = kmeans_test_df[features], kmeans_test_df['Max_LCAP']\n",
    "X_train, y_train = train_df[features], train_df['Max_LCAP']\n",
    "\n",
    "# Calculate permutation importance for test set\n",
    "result_test = permutation_importance(\n",
    "    trained_rf, X_test, y_test, \n",
    "    n_repeats=5, \n",
    "    random_state=42, \n",
    "    n_jobs=2, \n",
    "    scoring='neg_mean_absolute_error'\n",
    ")\n",
    "\n",
    "# Sort feature importances for test set\n",
    "sorted_importances_idx_test = result_test.importances_mean.argsort()\n",
    "importances_test = pd.DataFrame(\n",
    "    result_test.importances[sorted_importances_idx_test].T,\n",
    "    columns=X_test.columns[sorted_importances_idx_test],\n",
    ")\n",
    "\n",
    "# Calculate permutation importance for training data\n",
    "result_train = permutation_importance(\n",
    "    trained_rf, X_train, y_train, \n",
    "    n_repeats=5, \n",
    "    random_state=42, \n",
    "    n_jobs=2, \n",
    "    scoring='neg_mean_absolute_error'\n",
    ")\n",
    "\n",
    "# Sort feature importances for training set\n",
    "sorted_importances_idx_train = result_train.importances_mean.argsort()\n",
    "importances_train = pd.DataFrame(\n",
    "    result_train.importances[sorted_importances_idx_train].T,\n",
    "    columns=X_train.columns[sorted_importances_idx_train],\n",
    ")\n",
    "\n",
    "# Uncomment to save the importances to CSV\n",
    "# importances_test.to_csv('./data/PFI_results/mix_features_PFI_test.csv', index=False)\n",
    "# importances_train.to_csv('./data/PFI_results/mix_features_PFI_train.csv', index=False)\n",
    "\n",
    "# Calculate and sort the average importances\n",
    "pfi_test = importances_test.mean(axis=0).sort_values(ascending=False)[:10]\n",
    "pfi_train = importances_train.mean(axis=0).sort_values(ascending=False)[:10]\n",
    "\n",
    "# Plotting\n",
    "f, axs = plt.subplots(1, 2, figsize=(15, 5))\n",
    "sns.barplot(y=pfi_test.index, x=pfi_test, ax=axs[0])\n",
    "axs[0].set_title(\"Permutation Importances (test set)\")\n",
    "axs[0].axvline(x=0, color=\"k\", linestyle=\"--\")\n",
    "axs[0].set_xlabel(\"Decrease in accuracy score\")\n",
    "axs[0].figure.tight_layout()\n",
    "\n",
    "sns.barplot(y=pfi_train.index, x=pfi_train, ax=axs[1])\n",
    "axs[1].set_title(\"Permutation Importances (train set)\")\n",
    "axs[1].axvline(x=0, color=\"k\", linestyle=\"--\")\n",
    "axs[1].set_xlabel(\"Decrease in accuracy score\")\n",
    "axs[1].figure.tight_layout()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Using Round 7 compounds as test set for PFI analysis "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total run time is 6.58 seconds\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Define features and model for PFI\n",
    "features = rdkit_features\n",
    "rf = RandomForestRegressor()\n",
    "\n",
    "# Train model using the provided function\n",
    "trained_rf, performance, preds = model_training(\n",
    "    data=train_df,\n",
    "    model=rf,\n",
    "    split_method='round', # Use latest round of compounds as test set\n",
    "    descriptors=features,\n",
    "    test_df=None # Change to kmeans_test_df representative set for splitting data using k-means clustering \n",
    ")\n",
    "\n",
    "\n",
    "# Prepare test and training data\n",
    "X_test, y_test = kmeans_test_df[features], kmeans_test_df['Max_LCAP']\n",
    "X_train, y_train = train_df[features], train_df['Max_LCAP']\n",
    "\n",
    "# Calculate permutation importance for test data\n",
    "result_test = permutation_importance(\n",
    "    trained_rf, \n",
    "    X_test, \n",
    "    y_test, \n",
    "    n_repeats=5, \n",
    "    random_state=42, \n",
    "    n_jobs=2, \n",
    "    scoring='neg_mean_absolute_error'\n",
    ")\n",
    "\n",
    "# Prepare sorted importances dataframe for test data\n",
    "sorted_importances_idx_test = result_test.importances_mean.argsort()\n",
    "importances_test = pd.DataFrame(\n",
    "    result_test.importances[sorted_importances_idx_test].T,\n",
    "    columns=X_test.columns[sorted_importances_idx_test]\n",
    ")\n",
    "\n",
    "# Calculate permutation importance for training data\n",
    "result_train = permutation_importance(\n",
    "    trained_rf, \n",
    "    X_train, \n",
    "    y_train, \n",
    "    n_repeats=5, \n",
    "    random_state=42, \n",
    "    n_jobs=2, \n",
    "    scoring='neg_mean_absolute_error'\n",
    ")\n",
    "\n",
    "# Prepare sorted importances dataframe for training data\n",
    "sorted_importances_idx_train = result_train.importances_mean.argsort()\n",
    "importances_train = pd.DataFrame(\n",
    "    result_train.importances[sorted_importances_idx_train].T,\n",
    "    columns=X_train.columns[sorted_importances_idx_train]\n",
    ")\n",
    "\n",
    "# Optionally save results to CSV (uncomment if needed)\n",
    "# importances_test.to_csv('./data/PFI_results/mix_features_PFI_test.csv', index=False)\n",
    "# importances_train.to_csv('./data/PFI_results/mix_features_PFI_train.csv', index=False)\n",
    "\n",
    "# Calculate and sort average permutation importances\n",
    "pfi_test = importances_test.mean(axis=0).sort_values(ascending=False)[:10]\n",
    "pfi_train = importances_train.mean(axis=0).sort_values(ascending=False)[:10]\n",
    "\n",
    "# Create bar plots for test and train permutation importances\n",
    "f, axs = plt.subplots(1, 2, figsize=(15, 5))\n",
    "\n",
    "# Plot for Test Set\n",
    "sns.barplot(y=pfi_test.index, x=pfi_test, ax=axs[0])\n",
    "axs[0].set_title(\"Permutation Importances (Test Set)\")\n",
    "axs[0].axvline(x=0, color=\"k\", linestyle=\"--\")\n",
    "axs[0].set_xlabel(\"Decrease in Accuracy Score\")\n",
    "\n",
    "# Plot for Train Set\n",
    "sns.barplot(y=pfi_train.index, x=pfi_train, ax=axs[1])\n",
    "axs[1].set_title(\"Permutation Importances (Train Set)\")\n",
    "axs[1].axvline(x=0, color=\"k\", linestyle=\"--\")\n",
    "axs[1].set_xlabel(\"Decrease in Accuracy Score\")\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Calculate percentage change in MAE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Calculate permutation feature importances as a percentage of Mean Absolute Error (MAE)\n",
    "mae_value = performance['mae']\n",
    "\n",
    "# For test set\n",
    "pfi_test = 100 * importances_test.mean(axis=0) / mae_value\n",
    "pfi_test = pfi_test.sort_values(ascending=False)[:15]  # Top 15 features\n",
    "\n",
    "# For training set\n",
    "pfi_train = 100 * importances_train.mean(axis=0) / mae_value\n",
    "pfi_train = pfi_train.sort_values(ascending=False)[:15]  # Top 15 features\n",
    "\n",
    "# Create bar plots for test and train permutation importances\n",
    "f, axs = plt.subplots(1, 2, figsize=(15, 5))\n",
    "\n",
    "# Plot for test Set\n",
    "sns.barplot(y=pfi_test.index, x=pfi_test, ax=axs[0], color='steelblue')\n",
    "axs[0].set_title(\"Permutation Feature Importances (Test Set)\")\n",
    "axs[0].axvline(x=0, color=\"k\", linestyle=\"--\")\n",
    "axs[0].set_xlabel(\"Increase in Mean Absolute Error (%)\")\n",
    "axs[0].figure.tight_layout()\n",
    "\n",
    "# Plot for train Set\n",
    "sns.barplot(y=pfi_train.index, x=pfi_train, ax=axs[1], color='steelblue')\n",
    "axs[1].set_title(\"Permutation Feature Importances (Train Set)\")\n",
    "axs[1].axvline(x=0, color=\"k\", linestyle=\"--\")\n",
    "axs[1].set_xlabel(\"Increase in Mean Absolute Error (%)\")\n",
    "axs[1].figure.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Viewing important fingerprint bits "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.PngImagePlugin.PngImageFile image mode=RGB size=150x150 at 0x2AAADAB06B70>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Specify molecular fingerprint bit of interest\n",
    "feature = 'bit1233'\n",
    "bit_no = int(feature[3:])  # Extract bit number\n",
    "\n",
    "# Retrieve SMILES of first molecular entity that has specified fingerprint bit set\n",
    "smi = df_clean[df_clean[feature] == 1.0]['smiles'].tolist()[0]\n",
    "\n",
    "# Convert SMILES string to an RDKit molecule object\n",
    "mol = Chem.MolFromSmiles(smi)\n",
    "\n",
    "# Generate molecular fingerprint and associated bit info\n",
    "bit_info = {}\n",
    "fp = AllChem.GetMorganFingerprintAsBitVect(mol, radius=3, bitInfo=bit_info)\n",
    "Draw.DrawMorganBit(mol, bit_no, bit_info)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "CHML1",
   "language": "python",
   "name": "chml1"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}