iSIM_simulated_fps.ipynb

{"cells":[{"cell_type":"markdown","metadata":{},"source":["# **iSIM**\n","\n","***\n","Miranda-Quintana Group, Department of Chemistry, University of Florida\n","\n","***\n","Please cite the original iSIM paper:\n","\n","***\n","This notebook contains an example on how to generate simulated datasets of binary and real-value fingerprints and calculates iSIM values for them."]},{"cell_type":"code","execution_count":1,"metadata":{},"outputs":[],"source":["from isim_comp import calculate_isim\n","from isim_real import calculate_isim_real, pairwise_average_real\n","from isim_utils import pairwise_average\n","import numpy as np\n","import random"]},{"cell_type":"markdown","metadata":{},"source":["## Binary fingerprints\n","Please select the number of datasets, the desired range of elements in the set and number of bits in the fingerprint. Beware that for comparison the quadratic scaling way of fingerprints is computed, if large number of sets is simulated, the computing time will be large."]},{"cell_type":"code","execution_count":3,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["R2: 0.9999997890555186\n","MAE: 4.077067238988768e-05\n","RMSE: 0.00013070820419047868\n"]}],"source":["# Number of datasets to simulate\n","n_datasets = 100 #---> in the original paper, we used 10000 datasets\n","\n","# Similarity index to use\n","n_ary = 'JT'\n","\n","# Initialize lists to store iSIM and pairwise values\n","isim_values = []\n","pair_values = []\n","\n","for rep in range(n_datasets):\n","\n","    # Generate random binary fingerprints\n","    fp_total = random.randint(10, 100) #---> in the original paper, we used 100-1000 fingerprints\n","    fp_size = random.randint(166, 2049) #---> in the original paper, we used fingerprints of 166-2048 bits\n","    bias = np.random.uniform(0.01,1) #---> the bias is to cover the whole range of possible similarity values\n","    total_fingerprints = np.random.choice([int(0),int(1)], size=(fp_total, fp_size), p = [bias, 1 - bias])\n","\n","    # Append values \n","    isim_values.append(calculate_isim(total_fingerprints, n_ary = n_ary))\n","    pair_values.append(pairwise_average(total_fingerprints, n_ary = n_ary))\n","\n","\n","# Print results of the comparison of the iSIM and pairwise values\n","print('R2:', np.corrcoef(isim_values, pair_values)[0,1]**2)\n","print('MAE:', np.mean(np.abs(np.array(isim_values) - np.array(pair_values))))\n","print('RMSE:', np.sqrt(np.mean((np.array(isim_values) - np.array(pair_values))**2)))\n"]},{"cell_type":"markdown","metadata":{},"source":["## Real-value fingerprints"]},{"cell_type":"code","execution_count":4,"metadata":{},"outputs":[{"name":"stdout","output_type":"stream","text":["R2: 0.9997910836814017\n","MAE: 2.2956587628044645e-05\n","RMSE: 3.48212293794904e-05\n"]}],"source":["# Number of datasets to simulate\n","n_datasets = 100 #---> in the original paper, we used 10000 datasets\n","\n","# Similarity index to use\n","n_ary = 'JT'\n","\n","# Initialize lists to store iSIM and pairwise values\n","isim_values = []\n","pair_values = []\n","\n","# Generate random real-number fingerprints and compute isim and pairwise values\n","for rep in range(n_datasets):\n","\n","    # Random number of fingerprints and size\n","    fp_total = random.randint(10, 100) # ---> in the original paper, we used 100-1000 fingerprints\n","    fp_size = random.randint(166, 2049) # ---> in the original paper, we used fingerprints of 166-2048 bits\n","\n","    # Random matrix generation\n","    total_fingerprints = np.random.random((fp_total, fp_size))\n","\n","    # Append values\n","    isim_values.append(calculate_isim_real(total_fingerprints, n_ary=n_ary))\n","    pair_values.append(pairwise_average_real(total_fingerprints, n_ary = n_ary))\n","\n","# Print results of the comparison of the iSIM and pairwise values\n","print('R2:', np.corrcoef(isim_values, pair_values)[0,1]**2)\n","print('MAE:', np.mean(np.abs(np.array(isim_values) - np.array(pair_values))))\n","print('RMSE:', np.sqrt(np.mean((np.array(isim_values) - np.array(pair_values))**2)))"]}],"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":"3.10.12"}},"nbformat":4,"nbformat_minor":2}