comparison/undersampling.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "63d72d8a-8598-49fc-9db6-1eec8ec56686",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading nx_dataset...\n",
      "nx_dataset columns: ['smiles', 'pred(Caco-2_LogPapp)', 'pred(CYP3A4_pIC50)', 'pred(CYP2C9_pIC50)', 'pred(CYP2D6_pIC50)', 'pred(CYP3A4_pIC50)_log', 'pred(CYP2C9_pIC50)_log', 'pred(CYP2D6_pIC50)_log', 'CYP3A4_class', 'CYP2C9_class', 'CYP2D6_class']\n",
      "Converting SMILES to canonical form in nx_dataset...\n",
      "\n",
      "Loading admetica_dataset...\n",
      "admetica_dataset columns: ['smiles', 'Activity']\n",
      "Converting SMILES to canonical form in admetica_dataset...\n",
      "\n",
      "Initial nx_dataset shape: (273638, 11)\n",
      "nx_dataset shape after removing overlapping SMILES with admetica_dataset: (273402, 11)\n",
      "\n",
      "Balancing dataset by 'CYP3A4_class'...\n",
      "Balanced class distribution:\n",
      "CYP3A4_class\n",
      "0    22618\n",
      "1    22618\n",
      "Name: count, dtype: int64\n",
      "\n",
      "Keeping only 'smiles' and 'CYP3A4_class' columns in balanced_data...\n",
      "\n",
      "Renaming 'Activity' column to 'CYP3A4_class' in admetica_dataset...\n",
      "\n",
      "Combining balanced_data and admetica_dataset...\n",
      "\n",
      "Combined dataset saved to /Users/aleksashka/admetica/ADMET/metabolism/cyp3a4-inhibitor/balanced_admetica_3A4.csv\n",
      "Final combined dataset shape: (57781, 2)\n",
      "Loading nx_dataset...\n",
      "nx_dataset columns: ['smiles', 'pred(Caco-2_LogPapp)', 'pred(CYP3A4_pIC50)', 'pred(CYP2C9_pIC50)', 'pred(CYP2D6_pIC50)', 'pred(CYP3A4_pIC50)_log', 'pred(CYP2C9_pIC50)_log', 'pred(CYP2D6_pIC50)_log', 'CYP3A4_class', 'CYP2C9_class', 'CYP2D6_class']\n",
      "Converting SMILES to canonical form in nx_dataset...\n",
      "\n",
      "Loading admetica_dataset...\n",
      "admetica_dataset columns: ['smiles', 'Activity']\n",
      "Converting SMILES to canonical form in admetica_dataset...\n",
      "\n",
      "Initial nx_dataset shape: (273638, 11)\n",
      "nx_dataset shape after removing overlapping SMILES with admetica_dataset: (273382, 11)\n",
      "\n",
      "Balancing dataset by 'CYP2C9_class'...\n",
      "Balanced class distribution:\n",
      "CYP2C9_class\n",
      "0    3975\n",
      "1    3975\n",
      "Name: count, dtype: int64\n",
      "\n",
      "Keeping only 'smiles' and 'CYP2C9_class' columns in balanced_data...\n",
      "\n",
      "Renaming 'Activity' column to 'CYP2C9_class' in admetica_dataset...\n",
      "\n",
      "Combining balanced_data and admetica_dataset...\n",
      "\n",
      "Combined dataset saved to /Users/aleksashka/admetica/ADMET/metabolism/cyp2c9-inhibitor/balanced_admetica_2C9.csv\n",
      "Final combined dataset shape: (20299, 2)\n"
     ]
    }
   ],
   "source": [
    "# Import necessary libraries\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "from rdkit import Chem\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Define the canonical SMILES conversion function\n",
    "def canon_smiles(smiles):\n",
    "    \"\"\"Convert a SMILES string into its canonical form using RDKit.\"\"\"\n",
    "    try:\n",
    "        mol = Chem.MolFromSmiles(smiles)\n",
    "        if mol:\n",
    "            return Chem.MolToSmiles(mol, canonical=True)\n",
    "        else:\n",
    "            return float('nan')\n",
    "    except:\n",
    "        return float('nan')\n",
    "\n",
    "cyp_types = ['3A4','2C9']\n",
    "for cyp in cyp_types:\n",
    "        # Step 1: Load and preprocess nx_dataset\n",
    "        print(\"Loading nx_dataset...\")\n",
    "        current_directory = os.getcwd()\n",
    "        admetica_folder = os.path.abspath(os.path.join(current_directory, '..'))\n",
    "        nx_path = os.path.join(current_directory, 'novartis/novartis_data.csv')\n",
    "        nx_dataset = pd.read_csv(nx_path)\n",
    "        print(f\"nx_dataset columns: {nx_dataset.columns.tolist()}\")\n",
    "        \n",
    "        # Apply the canonical SMILES conversion to nx_dataset\n",
    "        print(\"Converting SMILES to canonical form in nx_dataset...\")\n",
    "        nx_dataset['smiles'] = nx_dataset['smiles'].apply(canon_smiles)\n",
    "        \n",
    "        # Step 2: Load and preprocess admetica_dataset\n",
    "        print(\"\\nLoading admetica_dataset...\")\n",
    "        admetica_dataset = pd.read_csv(f'{admetica_folder}/ADMET/metabolism/cyp{cyp.lower()}-inhibitor/cyp{cyp.lower()}-inhibitor_curated.csv')\n",
    "        print(f\"admetica_dataset columns: {admetica_dataset.columns.tolist()}\")\n",
    "        \n",
    "        # Apply the canonical SMILES conversion to admetica_dataset\n",
    "        print(\"Converting SMILES to canonical form in admetica_dataset...\")\n",
    "        admetica_dataset['smiles'] = admetica_dataset['smiles'].apply(canon_smiles)\n",
    "        \n",
    "        # Step 3: Remove rows from nx_dataset where 'smiles' exist in admetica_dataset\n",
    "        print(f\"\\nInitial nx_dataset shape: {nx_dataset.shape}\")\n",
    "        nx_dataset = nx_dataset[~nx_dataset['smiles'].isin(admetica_dataset['smiles'])]\n",
    "        print(f\"nx_dataset shape after removing overlapping SMILES with admetica_dataset: {nx_dataset.shape}\")\n",
    "        # Step 4: Create a balanced dataset based on 'CYP2C9_class'\n",
    "        print(f\"\\nBalancing dataset by 'CYP{cyp}_class'...\")\n",
    "        \n",
    "        # Split data into class 0 and class 1\n",
    "        class_0 = nx_dataset[nx_dataset[f'CYP{cyp}_class'] == 0]\n",
    "        class_1 = nx_dataset[nx_dataset[f'CYP{cyp}_class'] == 1]\n",
    "        \n",
    "        # Randomly sample instances from class 0 to match class 1\n",
    "        class_0_sample = class_0.sample(n=len(class_1), random_state=42)\n",
    "        \n",
    "        # Create a balanced dataset\n",
    "        balanced_data = pd.concat([class_0_sample, class_1])\n",
    "        \n",
    "        # Check class distribution\n",
    "        print(\"Balanced class distribution:\")\n",
    "        print(balanced_data[f'CYP{cyp}_class'].value_counts())\n",
    "        \n",
    "        # Step 5: Keep only relevant columns\n",
    "        print(f\"\\nKeeping only 'smiles' and 'CYP{cyp}_class' columns in balanced_data...\")\n",
    "        balanced_data = balanced_data[['smiles', f'CYP{cyp}_class']]\n",
    "        \n",
    "        # Step 6: Clean and rename columns in admetica_dataset\n",
    "        admetica_dataset = admetica_dataset.drop_duplicates(subset='smiles')\n",
    "        print(f\"\\nRenaming 'Activity' column to 'CYP{cyp}_class' in admetica_dataset...\")\n",
    "        admetica_dataset = admetica_dataset.rename(columns={'Activity': f'CYP{cyp}_class'})\n",
    "        \n",
    "        # Step 7: Concatenate the two datasets vertically\n",
    "        print(\"\\nCombining balanced_data and admetica_dataset...\")\n",
    "        combined_data = pd.concat([balanced_data, admetica_dataset], ignore_index=True)\n",
    "        \n",
    "        # Step 8: Save the combined dataset to CSV\n",
    "        output_file = f'{admetica_folder}/ADMET/metabolism/cyp{cyp.lower()}-inhibitor/balanced_admetica_{cyp}.csv'\n",
    "        combined_data.to_csv(output_file, index=False)\n",
    "        print(f\"\\nCombined dataset saved to {output_file}\")\n",
    "        print(f\"Final combined dataset shape: {combined_data.shape}\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.12.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}