bo_demo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Obtaining BO results from different molecule representation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step 1: Load Datasets\n",
    "In the present demonstration, we will use a model dataset *LIPO* correlating molecular structure with lipophilicity, a typical quantitative-structure property relationship (QSPR) modelling task.\n",
    "\n",
    "Hersey, A. ChEMBL Deposited Data Set - AZ dataset; 2015. https://doi.org/10.6019/chembl3301361"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set size: (420, 198)\n",
      "Candidate set size: (3780, 198)\n",
      "Best Observed in Training Set: 1.8142925537876984\n",
      "Best Observed in Full Set: 1.9299001765826398\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "from data_helper import gen_data_feat,load_lipo_feat\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.decomposition import PCA\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "featurizer_name = 'rdkit'\n",
    "# Here we adapt 2/98 split for training and candidate set\n",
    "partition_ratio = 0.1\n",
    "\n",
    "# Load from pre-featurized data\n",
    "X, y = load_lipo_feat(filename='data/lipo_{}.csv'.format(featurizer_name))\n",
    "\n",
    "# Split data into start training and candidate sets\n",
    "mol_track = np.arange(X.shape[0])\n",
    "\n",
    "# Split data into start training and candidate sets\n",
    "X_train, X_candidate, y_train, y_candidate, mol_track_train, mol_track_candidate = train_test_split(\n",
    "    X, y, mol_track,\n",
    "    test_size=1-partition_ratio,\n",
    "    random_state=1, #set random state for reproducibility, but vary in each trial\n",
    "    shuffle=True\n",
    ")\n",
    "\n",
    "# Standardize input data if needed\n",
    "scaler = StandardScaler()\n",
    "X_train = scaler.fit_transform(X_train)\n",
    "X_candidate = scaler.transform(X_candidate)\n",
    "\n",
    "# Apply PCA to reduce dimensionality (optional)\n",
    "# pca = PCA(n_components=50)\n",
    "# X_train = pca.fit_transform(X_train)\n",
    "# X_cadidate = pca.transform(X_cadidate)\n",
    "\n",
    "print('Training set size:', X_train.shape)\n",
    "print('Candidate set size:', X_candidate.shape)\n",
    "\n",
    "print('Best Observed in Training Set:', y_train.max())\n",
    "print('Best Observed in Full Set:', y.max())\n",
    "\n",
    "plt.hist(y_train, bins=20, alpha=0.5, label='Training Set')\n",
    "plt.hist(y_candidate, bins=20, alpha=0.5, label='Candidate Set')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Perform 'fair' partitioning through removing higher values from training set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set size: (374, 198)\n",
      "Candidate set size: (3826, 198)\n",
      "Best Observed in Training Set: 1.1867083157580154\n",
      "Best Observed in Full Set: 1.9299001765826398\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from data_helper import fair_train_test_split\n",
    "X_train, X_candidate, y_train, y_candidate, mol_track_train, mol_track_candidate = fair_train_test_split(\n",
    "    X, y, mol_track,\n",
    "    test_size=1-partition_ratio,\n",
    "    random_state=1, #set random state for reproducibility, but vary in each trial\n",
    "    shuffle=True,\n",
    "    threshold=1.2\n",
    ")\n",
    "\n",
    "# Standardize input data if needed\n",
    "scaler = StandardScaler()\n",
    "X_train = scaler.fit_transform(X_train)\n",
    "X_candidate = scaler.transform(X_candidate)\n",
    "\n",
    "print('Training set size:', X_train.shape)\n",
    "print('Candidate set size:', X_candidate.shape)\n",
    "\n",
    "print('Best Observed in Training Set:', y_train.max())\n",
    "print('Best Observed in Full Set:', y.max())\n",
    "\n",
    "plt.hist(y_train, bins=20, alpha=0.5, label='Training Set')\n",
    "plt.hist(y_candidate, bins=20, alpha=0.5, label='Candidate Set')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Step2: Prepare for Surrogate"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We here first import surrogates to do a preliminary modelling test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import r2_score\n",
    "\n",
    "from surrogates import Surrogate\n",
    "from surrogates import RandomForestSurrogate\n",
    "from surrogates import GPTanimotoSurrogate\n",
    "from surrogates import GPRQSurrogate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define surrogate model.\n",
    "test_surrogate = GPRQSurrogate()\n",
    "test_surrogate.load_data(train_x=X_train, train_y=y_train)\n",
    "\n",
    "# Fit surrogate model.\n",
    "test_surrogate.fit()\n",
    "\n",
    "# Get means and uncertainties from surrogate model.\n",
    "means, uncertainties = test_surrogate.predict_means_and_stddevs(X_candidate)\n",
    "print(f'Test shape: {X_candidate.shape}')\n",
    "print(f'Mean shape: {means.shape}')\n",
    "print(f'Uncertainty shape: {uncertainties.shape}')\n",
    "\n",
    "# Report results of model fit.\n",
    "print(f'R^2 Score on test set: {r2_score(y_candidate, means)}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Step3: Setup BO"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We first run on a single trial of BO, the iteration will be 10 rounds."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from surrogates import acqf_EI\n",
    "######################\n",
    "###Define setting#####\n",
    "######################\n",
    "\n",
    "num_iter = 20\n",
    "num_trial = 1\n",
    "\n",
    "\n",
    "###################################\n",
    "#####Train Surrogates##############\n",
    "###################################\n",
    "\n",
    "my_surrogate = GPRQSurrogate()\n",
    "my_surrogate.load_data(train_x=X_train, train_y=y_train)\n",
    "best_observed = y_train.max()\n",
    "\n",
    "#initialize the containers of new points and best observed values\n",
    "X_new_candidates , y_new_candidates= [],[]\n",
    "current_bests = []\n",
    "\n",
    "for iter in range(1,num_iter+1):\n",
    "\n",
    "    # Fit surrogate model.\n",
    "    my_surrogate.fit()\n",
    "\n",
    "    ######################################################################\n",
    "    #####Eval element in candidate set and max Acquisition function#######\n",
    "    ######################################################################\n",
    "\n",
    "    means, uncertainties = my_surrogate.predict_means_and_stddevs(X_candidate)\n",
    "        \n",
    "    # Calculate the Expected Improvement\n",
    "    ei = acqf_EI(means, uncertainties, best_observed)\n",
    "\n",
    "    # Find the index with the highest Expected Improvement\n",
    "    new_index = np.argmax(ei)\n",
    "    new_x = X_candidate[new_index]\n",
    "    new_y = y_candidate[new_index]\n",
    "\n",
    "    # Add the new point to the training set\n",
    "    my_surrogate.add_data(new_x, new_y)\n",
    "\n",
    "    # Remove the new point from the candidate set\n",
    "    X_candidate = np.delete(X_candidate, new_index, axis=0)\n",
    "    y_candidate = np.delete(y_candidate, new_index)\n",
    "\n",
    "    # Update the best observed value\n",
    "    if new_y > best_observed:\n",
    "        best_observed = new_y\n",
    "\n",
    "    # Record the new point and best observed value at this iteration\n",
    "    X_new_candidates , y_new_candidates = np.append(X_new_candidates, new_x), np.append(y_new_candidates, new_y)\n",
    "    current_bests = np.append(current_bests, best_observed)\n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Step4: Visualize Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Calculate Regret\n",
    "\n",
    "regret = np.max(y) - current_bests\n",
    "\n",
    "plt.plot(range(1, num_iter+1), regret, label='Best Observed')"
   ]
  }
 ],
 "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": "3.9.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}