passive_learning.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 141,
   "id": "71356a8e",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "import csv\n",
    "import tensorflow as tf\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from keras.preprocessing.text import Tokenizer \n",
    "from keras_preprocessing.sequence import pad_sequences \n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "from imblearn.under_sampling import RandomUnderSampler\n",
    "\n",
    "from rdkit import Chem\n",
    "from rdkit.Chem import RDKFingerprint\n",
    "from rdkit.Chem import rdMolDescriptors\n",
    "\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.metrics import accuracy_score"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c4e58deb",
   "metadata": {},
   "source": [
    "## EXTRACT DATA\n",
    "CSV file to be present in local machine.<br>\n",
    "FIle name is data.csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "1124863d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of entries processed:  29537\n"
     ]
    }
   ],
   "source": [
    "data = []\n",
    "with open('data.csv', 'r') as csvfile: \n",
    "    reader = csv.reader(csvfile, skipinitialspace=True)\n",
    "    data.append(tuple(next(reader)))\n",
    "    for ID, SMILE, B, RES, is_active in reader:\n",
    "        data.append((ID, SMILE, B, RES, is_active))\n",
    "\n",
    "# remove header\n",
    "data = data[1:]\n",
    "maxlen = len(data)\n",
    "print('Number of entries processed: ', maxlen)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "8839e596",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get only the SMILES data\n",
    "SMILES = []\n",
    "for i in data:\n",
    "    SMILES.append(i[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "7f38a1fc",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['c1(c2c([nH]c1)ccc(c2)OCc1ccccc1)CN(C)C ',\n",
       " 'c12[nH]c3c(c1CCN=C2C)ccc(c3)O.Cl.O ']"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SMILES[0:2]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4e6f4bc",
   "metadata": {},
   "source": [
    "## ONE-HOT ENCODING OF SMILES\n",
    "This will help in vectrorizing smiles data.<br>\n",
    "This section will use the tokenizer function that is provided in tensorflow, and we will use keras as the front-end to access the function.<br>\n",
    "Credit: https://www.kaggle.com/code/art3mis/220221-getting-started-with-smiles <br>\n",
    "Inspiration: https://towardsdatascience.com/basic-molecular-representation-for-machine-learning-b6be52e9ff76"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "9f57adb5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Encoded string: [ 0  0  0 ...  4  2 10]\n"
     ]
    }
   ],
   "source": [
    "# Create a tokenizer for vectorization. \n",
    "tokenizer = Tokenizer(char_level=True, lower=False)\n",
    "tokenizer.fit_on_texts(SMILES)\n",
    "# Encoding text_to_sequences\n",
    "sequences = tokenizer.texts_to_sequences(SMILES)\n",
    "pad_data = pad_sequences(sequences, maxlen=maxlen)\n",
    "print ('Encoded string:',pad_data[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "458484f0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Decoded string: ['c 1 ( c 2 c ( [ n H ] c 1 ) c c c ( c 2 ) O C c 1 c c c c c 1 ) C N ( C ) C  ']\n"
     ]
    }
   ],
   "source": [
    "#sequences_to_text decodes\n",
    "print ('Decoded string:',tokenizer.sequences_to_texts(pad_data[0:1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "edb3eed2",
   "metadata": {},
   "outputs": [],
   "source": [
    "def decodeOH2SMILES(array, tokenizer):\n",
    "    # Converts one-hot encoded sequences back to SMILES strings.\n",
    "    output = tokenizer.sequences_to_texts(array)\n",
    "    #Modify this if you change the start or end token.\n",
    "    disallowed_characters = 'E! '\n",
    "    for character in disallowed_characters:\n",
    "        output = [i.replace(character,'') for i in output]\n",
    "    return output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "2a4292b4",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['c1(c2c([nH]c1)ccc(c2)OCc1ccccc1)CN(C)C', 'c12[nH]c3c(c1CCN=C2C)ccc(c3)O.Cl.O']\n"
     ]
    }
   ],
   "source": [
    "decoded_smiles = decodeOH2SMILES(pad_data[0:2],tokenizer)\n",
    "print(decoded_smiles)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5db38628",
   "metadata": {},
   "source": [
    "## MOLECULAR FINGERPRINT ENCODING OF SMILES\n",
    "This will help in converting smiles data so that it can be used for applying ML techniques. This is more often used (from surveying)<br>\n",
    "Credit: https://towardsdatascience.com/basic-molecular-representation-for-machine-learning-b6be52e9ff76"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "id": "032911ee",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<rdkit.Chem.rdchem.Mol at 0x7f7b489b12a0>"
      ]
     },
     "execution_count": 109,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Drawing molecules\n",
    "mol = Chem.MolFromSmiles(SMILES[0])\n",
    "mol"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "id": "66bad49e",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[20:09:52] WARNING: not removing hydrogen atom without neighbors\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[1, 0, 0, ..., 0, 0, 1],\n",
       "       [1, 0, 0, ..., 0, 0, 1],\n",
       "       [0, 0, 1, ..., 0, 0, 1],\n",
       "       ...,\n",
       "       [1, 0, 0, ..., 0, 0, 0],\n",
       "       [1, 0, 0, ..., 0, 1, 0],\n",
       "       [1, 0, 0, ..., 0, 0, 0]])"
      ]
     },
     "execution_count": 112,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# We will be using RDK fingerprint to vectorize all our SMILES structures\n",
    "fingerprints = []\n",
    "for i in range(len(SMILES)):\n",
    "    mol = Chem.MolFromSmiles(SMILES[i])\n",
    "    fingerprint_rdk = np.array(RDKFingerprint(mol))\n",
    "    fingerprints.append(fingerprint_rdk)\n",
    "fingerprints = np.array(fingerprints)\n",
    "fingerprints"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "id": "59d025aa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of features = 2048\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of features =\",fingerprints.shape[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "98ff12f0",
   "metadata": {},
   "source": [
    "## VISUALIZATION OF DATA\n",
    "Data is obtained from the following paper - https://doi.org/10.1371/journal.pcbi.1010613 <br>\n",
    "In this paper (and a previous paper of theirs), they conducted a series of HTS to obtain the Average B-score and residual values for a series of small molecules to see their potency against Burkholderia cenocepacia.<br>\n",
    "Quoting the paper,<br>\n",
    "\n",
    "\"The dataset used in the ML approach consisted of 29,537 compounds with residual growth (RG) values and average B-scores. The RG measures the ratio of bacterial growth in the presence and absence of the compounds. The B-score measures relative potency that adjusts the RG for any screening artifacts resulting from well position (row and column) in the assay plate during the HTS. The B-score is inversely proportional to compound potency, where negative B-scores indicate greater growth inhibitory activity of the compounds. To binarize the compounds, the previously established average B-score threshold of -17.5 was chosen [https://doi.org/10.1371/journal.pone.0128587]. Overall, 256 small molecules were classified as growth inhibitory.\"\n",
    "\n",
    "Staying true to the essence of the paper, we will be setting a threshold of -17.5 for B-score in our data as well, i.e. below this threshold, the compound is labeled active, and otherwise inactive."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "07e003d7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of active compounds = 256\n"
     ]
    }
   ],
   "source": [
    "# Checking data for number of active compounds\n",
    "# idx - ID, SMILE, B, RES, is_active\n",
    "count = 0\n",
    "for i in range(len(data)):\n",
    "    if float(data[i][2]) <= -17.5:\n",
    "        if float(data[i][4]) == 1:\n",
    "            count += 1\n",
    "print(\"Number of active compounds =\",count)\n",
    "# This follows what is mentioned in the paper."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "id": "488ee69f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get the B-score values\n",
    "Bscores = []\n",
    "active = []\n",
    "is_active = []\n",
    "for i in range(len(data)):\n",
    "    Bscores.append(float(data[i][2]))\n",
    "    is_active.append(float(data[i][4]))\n",
    "    if float(data[i][4]) == 0:\n",
    "        active.append(\"Inactive\")\n",
    "    else:\n",
    "         active.append(\"Active\")   \n",
    "compound_idx = np.arange(len(Bscores))\n",
    "is_active = np.array(is_active)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "f3a3ab9c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Text(0.5, 0, 'Compound Index'), Text(0, 0.5, 'B-Score')]"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualizing data\n",
    "# Plot B-scores\n",
    "fig, ax = plt.subplots()\n",
    "# plt.scatter(compound_idx[0:100], Bscores[0:100], c =\"blue\")\n",
    "sns.set(style='whitegrid')\n",
    "sns.scatterplot(x=compound_idx,\n",
    "                    y=Bscores,\n",
    "                    hue=active, marker='+')\n",
    "# ax.set_ylim(-500, 100)\n",
    "ax.set(xlabel='Compound Index', ylabel='B-Score')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5f9083d0",
   "metadata": {},
   "source": [
    "## SPLITTING OF DATA\n",
    "This is an important section because our data is heavily unbalanced. Therefore, we do random undersampling of the major class.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "id": "c8f7884a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Includes undersampling of major label to make classes balanced.\n",
    "def train_test_split(fingerprints, active, split_ratio=0.8):\n",
    "    rus = RandomUnderSampler() # to make it reproducible -- random_state=rng\n",
    "    X_resampled, y_resampled = rus.fit_resample(fingerprints, active)\n",
    "    \n",
    "    # split data\n",
    "    # get random permutations\n",
    "    rand_idx = np.random.permutation(len(X_resampled))\n",
    "    train_size = round(len(X_resampled)*split_ratio)\n",
    "    trainX = X_resampled[rand_idx[0:train_size]]\n",
    "    testX = X_resampled[rand_idx[train_size:len(X_resampled)]]\n",
    "    trainY = y_resampled[rand_idx[0:train_size]]\n",
    "    testY = y_resampled[rand_idx[train_size:len(X_resampled)]]\n",
    "    \n",
    "    return trainX, trainY, testX, testY"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "id": "7f99282e",
   "metadata": {},
   "outputs": [],
   "source": [
    "trainX, trainY, testX, testY = train_test_split(fingerprints, is_active)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "id": "7924d0f9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of active compounds in:\n",
      "Train Set =  211\n",
      "Test Set =  45\n",
      "Number of inactive compounds in:\n",
      "Train Set =  199\n",
      "Test Set =  57\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of active compounds in:\")\n",
    "print(\"Train Set = \", sum(trainY == 1))\n",
    "print(\"Test Set = \", sum(testY == 1))\n",
    "\n",
    "print(\"Number of inactive compounds in:\")\n",
    "print(\"Train Set = \", sum(trainY == 0))\n",
    "print(\"Test Set = \", sum(testY == 0))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ace9e5d3",
   "metadata": {},
   "source": [
    "## IMPLEMENTING LOGISTIC REGRESSION ON MOLECULAR FINGERPRINTS\n",
    "Logistic regression will be our base learner for this project. Below is the passive learning implementation for the same."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 181,
   "id": "99db39c2",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LogR():\n",
    "    def __init__(self, x, y):\n",
    "        self.trainX = x\n",
    "        self.trainY = y.reshape(len(y),)\n",
    "        self.model = LogisticRegression(random_state=0, max_iter=500)\n",
    "    def fit(self):\n",
    "        self.model.fit(self.trainX, self.trainY)\n",
    "    def predict(self, x):\n",
    "        return self.model.predict(x)\n",
    "#     def predict_prob(self, x):\n",
    "#         return self.model.predict_proba(x)\n",
    "#     def predict_logprob(self, x):\n",
    "#         return self.model.predict_log_proba(x)\n",
    "    def generate_acc(self, pred, true):\n",
    "        return accuracy_score(true, pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "id": "442e106b",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = LogR(trainX, trainY)\n",
    "model.fit()\n",
    "y_pred = model.predict(testX)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "id": "2e60aa70",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6274509803921569"
      ]
     },
     "execution_count": 183,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.generate_acc(y_pred, testY)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "id": "96a6680f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Run multiple simulations since we are sampling the dataset for train-test-split\n",
    "acc = []\n",
    "for simulations in range(100):\n",
    "    trainX, trainY, testX, testY = train_test_split(fingerprints, is_active)\n",
    "    model = LogR(trainX, trainY)\n",
    "    model.fit()\n",
    "    y_pred = model.predict(testX)\n",
    "    acc.append(model.generate_acc(y_pred, testY))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "id": "2bdb356b",
   "metadata": {},
   "outputs": [],
   "source": [
    "avg_acc = np.mean(acc)\n",
    "acc_std = np.std(acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 188,
   "id": "e8f70df2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The average accuracy over all simulations = 0.7148039215686275\n",
      "The standard deviation over all simulations = 0.040401079349573495\n"
     ]
    }
   ],
   "source": [
    "print(\"The average accuracy over all simulations =\", avg_acc)\n",
    "print(\"The standard deviation over all simulations =\", acc_std)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7dbe2416",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:IDL2]",
   "language": "python",
   "name": "conda-env-IDL2-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}