Prediction_with_trained_model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "# Predict and evaluate DTIs and BR with trained Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-10-29 16:23:10.976659: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n"
     ]
    }
   ],
   "source": [
    "from HoTS.model.hots import *\n",
    "from HoTS.utils.build_features import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Firstly, you need to define encoders for protein sequences and compounds with given encoding types."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "os.environ[\"CUDA_DEVICE_ORDER\"] = \"PCI_BUS_ID\"   # see issue #152\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# define input feature\n",
    "prot_vec = \"Sequence\"\n",
    "drug_vec = \"Morgan\"\n",
    "drug_len = 2048\n",
    "radius = 2\n",
    "protein_encoder = ProteinEncoder(prot_vec)\n",
    "compound_encoder = CompoundEncoder(drug_vec, radius=radius, n_bits=drug_len)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Load trained model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You need to intialize model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hots model initialization done!\n"
     ]
    }
   ],
   "source": [
    "# initialize model\n",
    "dti_model = HoTS()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Then you can load model. You can use `dti_model.load_model(\"output.config.json\")`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Given hyperparamters in ./Model/Model_config.json are loaded\n",
      "\n",
      "protein_grid_size   :  10\n",
      "compound_grid_size  :  None\n",
      "anchors             :  [9]\n",
      "hots_dimension      :  128\n",
      "hots_n_heads        :  4\n",
      "dropout             :  0.1\n",
      "drug_layers         :  [512, 128]\n",
      "protein_strides     :  [5, 10, 15, 20, 25, 30]\n",
      "filters             :  128\n",
      "fc_layers           :  [256, 64]\n",
      "hots_fc_layers      :  [256, 64]\n",
      "learning_rate       :  0.0001\n",
      "prot_vec            :  Sequence\n",
      "drug_vec            :  Morgan\n",
      "drug_len            :  2048\n",
      "activation          :  gelu\n",
      "protein_layers      :  [128, 128, 128, 128]\n",
      "reg_loss_weight     :  0.1\n",
      "conf_loss_weight    :  1\n",
      "negative_loss_weight:  0.1\n",
      "retina_loss_weight  :  2\n",
      "decay               :  0.0001\n",
      "hots_file           :  ./Model/HoTS.h5\n",
      "dti_file            :  ./Model/DTI.h5\n",
      "hots_validation_results:  {}\n",
      "dti_validation_results:  {}\n",
      "n_stack_hots_prediction:  2\n",
      "protein_encoder_config:  {'feature': 'Sequence'}\n",
      "compound_encoder_config:  {'radius': 2, 'feature': 'Morgan', 'n_bits': 2048, 'n_compound_word': 65}\n",
      " \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-10-29 16:23:13.535617: I tensorflow/compiler/jit/xla_cpu_device.cc:41] Not creating XLA devices, tf_xla_enable_xla_devices not set\n",
      "2022-10-29 16:23:13.537012: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcuda.so.1\n",
      "2022-10-29 16:23:13.615713: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties: \n",
      "pciBusID: 0000:24:00.0 name: NVIDIA TITAN RTX computeCapability: 7.5\n",
      "coreClock: 1.77GHz coreCount: 72 deviceMemorySize: 23.65GiB deviceMemoryBandwidth: 625.94GiB/s\n",
      "2022-10-29 16:23:13.615751: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
      "2022-10-29 16:23:13.618547: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
      "2022-10-29 16:23:13.618663: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.10\n",
      "2022-10-29 16:23:13.621427: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10\n",
      "2022-10-29 16:23:13.622333: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10\n",
      "2022-10-29 16:23:13.624580: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10\n",
      "2022-10-29 16:23:13.626016: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.10\n",
      "2022-10-29 16:23:13.630084: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n",
      "2022-10-29 16:23:13.632304: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0\n",
      "2022-10-29 16:23:13.632863: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  SSE4.1 SSE4.2 AVX AVX2 AVX512F FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2022-10-29 16:23:13.633479: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set\n",
      "2022-10-29 16:23:13.634650: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties: \n",
      "pciBusID: 0000:24:00.0 name: NVIDIA TITAN RTX computeCapability: 7.5\n",
      "coreClock: 1.77GHz coreCount: 72 deviceMemorySize: 23.65GiB deviceMemoryBandwidth: 625.94GiB/s\n",
      "2022-10-29 16:23:13.634682: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
      "2022-10-29 16:23:13.634733: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
      "2022-10-29 16:23:13.634744: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublasLt.so.10\n",
      "2022-10-29 16:23:13.634754: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcufft.so.10\n",
      "2022-10-29 16:23:13.634764: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcurand.so.10\n",
      "2022-10-29 16:23:13.634774: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusolver.so.10\n",
      "2022-10-29 16:23:13.634784: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcusparse.so.10\n",
      "2022-10-29 16:23:13.634795: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n",
      "2022-10-29 16:23:13.636843: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1862] Adding visible gpu devices: 0\n",
      "2022-10-29 16:23:13.636884: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudart.so.10.1\n",
      "2022-10-29 16:23:14.275094: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1261] Device interconnect StreamExecutor with strength 1 edge matrix:\n",
      "2022-10-29 16:23:14.275134: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1267]      0 \n",
      "2022-10-29 16:23:14.275141: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1280] 0:   N \n",
      "2022-10-29 16:23:14.278737: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1406] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 13349 MB memory) -> physical GPU (device: 0, name: NVIDIA TITAN RTX, pci bus id: 0000:24:00.0, compute capability: 7.5)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\tHoTS Model is loaded from ./Model/HoTS.h5\n",
      "\tDTI Model is loaded from ./Model/DTI.h5\n"
     ]
    }
   ],
   "source": [
    "# load model\n",
    "dti_model.load_model(\"./Model/Model_config.json\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Load BR dataset, predict and evaluate Performance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You can use the `parse_HoTS_data` function to load HoTS data as input of the HoTS model.\n",
    "\n",
    "`parse_HoTS_data` is located in `HoTS.utils.build_features`.\n",
    "\n",
    "If it contains `binding_region` columns you parsed, then set `binding_region=True`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parsing HoTS data: ./SampleData/HoTS/Validation_HoTS.tsv\n",
      "Number of 3D-complexes : 232\n",
      "Number of proteins : 207\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 232/232 [00:00<00:00, 31168.43it/s]\n"
     ]
    }
   ],
   "source": [
    "hots_data = parse_HoTS_data(\"./SampleData/HoTS/Validation_HoTS.tsv\",\n",
    "                           compound_encoder=compound_encoder, protein_encoder=protein_encoder, \n",
    "                          binding_region=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You can predict BRs with `HoTS_prediction`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-10-29 16:23:18.142627: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:116] None of the MLIR optimization passes are enabled (registered 2)\n",
      "2022-10-29 16:23:18.187538: I tensorflow/core/platform/profile_utils/cpu_utils.cc:112] CPU Frequency: 3800000000 Hz\n",
      "2022-10-29 16:23:18.774938: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10\n",
      "2022-10-29 16:23:19.135566: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcudnn.so.7\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:5 out of the last 5 calls to <function Model.make_predict_function.<locals>.predict_function at 0x7fc90802f670> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has experimental_relax_shapes=True option that relaxes argument shapes that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n",
      "WARNING:tensorflow:6 out of the last 6 calls to <function Model.make_predict_function.<locals>.predict_function at 0x7fc8a46d8f70> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has experimental_relax_shapes=True option that relaxes argument shapes that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n"
     ]
    }
   ],
   "source": [
    "hots_prediction_result = dti_model.HoTS_prediction(**hots_data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "It returns a tuple of DTI predictions and BR predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(316, 331, 0.7528640627861023),\n",
       " (307, 320, 0.718887209892273),\n",
       " (378, 393, 0.7044767737388611),\n",
       " (358, 372, 0.6984079480171204),\n",
       " (297, 311, 0.6941364407539368),\n",
       " (267, 281, 0.6887148022651672),\n",
       " (334, 349, 0.6546329259872437),\n",
       " (207, 221, 0.6542425751686096),\n",
       " (189, 201, 0.6519545316696167),\n",
       " (366, 378, 0.6451152563095093)]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hots_prediction_result[1][0][0:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Please see under sections for interpretation of this result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You can evaluate BR prediction results with the `HoTS_validation` function,\n",
    "\n",
    "then AP will be calculated. __(It takes a lot of time for AP calculation, so that you can skip it)__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\tAP :  0.6485652597577035\n",
      "=================================================\n"
     ]
    }
   ],
   "source": [
    "hots_evalution_result = dti_model.HoTS_validation(**hots_data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "evaluation results are summarized as a dictionary object"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'AP': 0.6485652597577035}"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hots_evalution_result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Load DTI dataset, predict and evaluate Performance"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You can use the `parse_DTI_data` function to load HoTS data as the input of HoTS model.\n",
    "\n",
    "`parse_DTI_data` is located in `HoTS.utils.build_features`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parsing ./SampleData/DTI/Validation/Validation_DTI.csv , ./SampleData/DTI/Validation/Validation_Compound.csv, ./SampleData/DTI/Validation/Validation_Protein.csv with length 2500, type None\n",
      "Encoding compound with Morgan type\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 499/499 [00:00<00:00, 648.56it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Encoding compound ends!\n",
      "\tPositive data : 370\n",
      "\tNegative data : 507\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "dti_data = parse_DTI_data(\"./SampleData/DTI/Validation/Validation_DTI.csv\", \n",
    "                   \"./SampleData/DTI/Validation/Validation_Compound.csv\", \n",
    "                   \"./SampleData/DTI/Validation/Validation_Protein.csv\", \n",
    "                   compound_encoder=compound_encoder, protein_encoder=protein_encoder)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You can predict DTI with the `DTI_prediction` function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "dti_prediction = dti_model.DTI_prediction(**dti_data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Also, you can easily evaluate perforamnce with the `DTI_validation` function.\n",
    "\n",
    "Basically, HoTS will calculate AUC and AUPR.\n",
    "\n",
    "You can put `threshold` argument, then sensitivity, specificity, precision, accuracy, F1 scores are also calculated."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\tSen :  0.6351351351351351\n",
      "\tSpe :  0.8224852071005917\n",
      "\tPrecision :  0.7230769230769231\n",
      "\tAcc :  0.7434435575826682\n",
      "\tF1 :  0.6762589928057553\n",
      "\tArea Under ROC Curve(AUC): 0.838\n",
      "\tArea Under PR Curve(AUPR): 0.811\n",
      "\tOptimal threshold(AUC)   : 0.000 \n",
      "\tOptimal threshold(AUPR)  : 0.218\n",
      "=================================================\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/dlsrnsi/DTI/HoTS_github/HoTS/HoTS/model/hots.py:295: RuntimeWarning: divide by zero encountered in divide\n",
      "  EERs = (1-recall)/(1-precision)\n"
     ]
    }
   ],
   "source": [
    "dti_evalution_result = dti_model.DTI_validation(threshold=0.2, **dti_data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Evaluation results are summarized as a dictionary object."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'Sen': 0.6351351351351351,\n",
       " 'Spe': 0.8224852071005917,\n",
       " 'Acc': 0.7434435575826682,\n",
       " 'Pre': 0.7230769230769231,\n",
       " 'F1': 0.6762589928057553,\n",
       " 'AUC': 0.8381043765659151,\n",
       " 'AUPR': 0.8106569321431619}"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dti_evalution_result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Predict BRs and DTIs without loading data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "import urllib\n",
    "uniprot_url = \"https://www.uniprot.org/uniprot/{0}.fasta\"\n",
    "from Bio import Entrez\n",
    "\n",
    "def get_smiles_from_cid(cid):\n",
    "    return urllib.request.urlopen(\"https://pubchem.ncbi.nlm.nih.gov/rest/pug/compound/cid/%d/property/CanonicalSMILES/txt\"%cid).read().decode(\"utf-8\").strip()\n",
    " \n",
    "def get_seq_from_uniprot_acc(uniprot_acc):\n",
    "    opened = urllib.request.urlopen(uniprot_url.format(uniprot_acc))\n",
    "    lines = opened.readlines()\n",
    "    return \"\".join([line.decode(\"utf-8\").rstrip() for line in lines[1:]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You can download SMILES with PubChem CID"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# SMILES of drug (should be listed)\n",
    "drugs = [get_smiles_from_cid(5291), get_smiles_from_cid(219024)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['CC1=C(C=C(C=C1)NC(=O)C2=CC=C(C=C2)CN3CCN(CC3)C)NC4=NC=CC(=N4)C5=CN=CC=C5', 'CNC(=O)C1=CN(N=C1)C2=NC(=C3C(=N2)N(C=N3)C4C(C(C(O4)CO)O)O)N']\n"
     ]
    }
   ],
   "source": [
    "print(drugs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Also, you can download protein sequences with UniProt accession."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# Sequences (should be listed)\n",
    "\n",
    "targets = [get_seq_from_uniprot_acc(\"P00519\"), get_seq_from_uniprot_acc(\"P29274\")]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['MLEICLKLVGCKSKKGLSSSSSCYLEEALQRPVASDFEPQGLSEAARWNSKENLLAGPSENDPNLFVALYDFVASGDNTLSITKGEKLRVLGYNHNGEWCEAQTKNGQGWVPSNYITPVNSLEKHSWYHGPVSRNAAEYLLSSGINGSFLVRESESSPGQRSISLRYEGRVYHYRINTASDGKLYVSSESRFNTLAELVHHHSTVADGLITTLHYPAPKRNKPTVYGVSPNYDKWEMERTDITMKHKLGGGQYGEVYEGVWKKYSLTVAVKTLKEDTMEVEEFLKEAAVMKEIKHPNLVQLLGVCTREPPFYIITEFMTYGNLLDYLRECNRQEVNAVVLLYMATQISSAMEYLEKKNFIHRDLAARNCLVGENHLVKVADFGLSRLMTGDTYTAHAGAKFPIKWTAPESLAYNKFSIKSDVWAFGVLLWEIATYGMSPYPGIDLSQVYELLEKDYRMERPEGCPEKVYELMRACWQWNPSDRPSFAEIHQAFETMFQESSISDEVEKELGKQGVRGAVSTLLQAPELPTKTRTSRRAAEHRDTTDVPEMPHSKGQGESDPLDHEPAVSPLLPRKERGPPEGGLNEDERLLPKDKKTNLFSALIKKKKKTAPTPPKRSSSFREMDGQPERRGAGEEEGRDISNGALAFTPLDTADPAKSPKPSNGAGVPNGALRESGGSGFRSPHLWKKSSTLTSSRLATGEEEGGGSSSKRFLRSCSASCVPHGAKDTEWRSVTLPRDLQSTGRQFDSSTFGGHKSEKPALPRKRAGENRSDQVTRGTVTPPPRLVKKNEEAADEVFKDIMESSPGSSPPNLTPKPLRRQVTVAPASGLPHKEEAGKGSALGTPAAAEPVTPTSKAGSGAPGGTSKGPAEESRVRRHKHSSESPGRDKGKLSRLKPAPPPPPAASAGKAGGKPSQSPSQEAAGEAVLGAKTKATSLVDAVNSDAAKPSQPGEGLKKPVLPATPKPQSAKPSGTPISPAPVPSTLPSASSALAGDQPSSTAFIPLISTRVSLRKTRQPPERIASGAITKGVVLDSTEALCLAISRNSEQMASHSAVLEAGKNLYTFCVSYVDSIQQMRNKFAFREAINKLENNLRELQICPATAGSGPAATQDFSKLLSSVKEISDIVQR', 'MPIMGSSVYITVELAIAVLAILGNVLVCWAVWLNSNLQNVTNYFVVSLAAADIAVGVLAIPFAITISTGFCAACHGCLFIACFVLVLTQSSIFSLLAIAIDRYIAIRIPLRYNGLVTGTRAKGIIAICWVLSFAIGLTPMLGWNNCGQPKEGKNHSQGCGEGQVACLFEDVVPMNYMVYFNFFACVLVPLLLMLGVYLRIFLAARRQLKQMESQPLPGERARSTLQKEVHAAKSLAIIVGLFALCWLPLHIINCFTFFCPDCSHAPLWLMYLAIVLSHTNSVVNPFIYAYRIREFRQTFRKIIRSHVLRQQEPFKAAGTSARVLAAHGSDGEQVSLRLNGHPPGVWANGSAPHPERRPNGYALGLVSGGSAQESQGNTGLPDVELLSHELKGVCPEPPGLDDPLAQDGAGVS']\n"
     ]
    }
   ],
   "source": [
    "print(targets)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You need to encode SMILES and sequences with a defined encoder."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "drugs_fp = [compound_encoder.encode(drug) for drug in drugs]\n",
    "targets_encoded = [protein_encoder.encode(target) for target in targets]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "For protein sequences with different lengths, you need to pad sequence."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "target_encoded = protein_encoder.pad(targets_encoded)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "### Prediction of DTIs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "you can simply predict DTIs with the `DTI_prediction` function of the `dti_model` object"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "dti_prediction_results = dti_model.DTI_prediction(drugs_fp, targets_encoded)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "The model will result in [N, 1] shape prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.9627259 ]\n",
      " [0.97712976]]\n"
     ]
    }
   ],
   "source": [
    "print(dti_prediction_results)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "### Prediction of BRs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You can predict BRs with the `HoTS_prediction` function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    },
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "dti_predictions,  br_predictions = dti_model.HoTS_prediction(drugs_fp, targets_encoded)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "BR prediction results contain a list of (BR_start, BR_end, confidence score) for each protein grid.\n",
    "\n",
    "Also, they are sorted by confidence score."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(307, 330, 0.7390334010124207), (367, 382, 0.721824049949646), (260, 274, 0.7199005484580994), (358, 373, 0.7056204080581665), (288, 303, 0.7008621692657471), (246, 262, 0.6984542012214661), (298, 312, 0.697698175907135), (376, 389, 0.6810927391052246), (240, 254, 0.6167941093444824), (277, 291, 0.6081487536430359)]\n"
     ]
    }
   ],
   "source": [
    "print(br_predictions[0][0:10])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "### Visualization of BR predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You can visualize results with the `HoTS_visualization` function, but you need to give a list of names for each visualiztion.\n",
    "\n",
    "You need to put names DTI pairs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "names = [\"ABL1_imatinib\", \"AA2AR_Regadenoson\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction with 0.600000\n",
      "ABL1_imatinib\n",
      "DTI score :  [0.9627259]\n",
      "  Sequence : MLEICLKLVGCKSKKGLSSSSSCYLEEALQRPVASDFEPQGLSEAARWNSKENLLAGPSENDPNLFVALYDFVASGDNTLSITKGEKLRVLGYNHNGEWC\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : EAQTKNGQGWVPSNYITPVNSLEKHSWYHGPVSRNAAEYLLSSGINGSFLVRESESSPGQRSISLRYEGRVYHYRINTASDGKLYVSSESRFNTLAELVH\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : HHSTVADGLITTLHYPAPKRNKPTVYGVSPNYDKWEMERTDITMKHKLGGGQYGEVYEGVWKKYSLTVAVKTLKEDTMEVEEFLKEAAVMKEIKHPNLVQ\n",
      "Prediction :                                         DITMKHKLGGGQYGEVYEGVWKKYSLTVAVKTLK   MEVEEFLKEAAVMKEIKHPNLVQ\n",
      "     Score :                                         61%   69%           71%              60%        70%       69\n",
      "  Sequence : LLGVCTREPPFYIITEFMTYGNLLDYLRECNRQEVNAVVLLYMATQISSAMEYLEKKNFIHRDLAARNCLVGENHLVKVADFGLSRLMTGDTYTAHAGAK\n",
      "Prediction : LLGVCTREPPFYIITEFMTYGNLLD                                 FIHRDLAARNCLVGENHLVKVADFGLSRLMT           \n",
      "     Score : %      73%                                                70%      72%      68%                     \n",
      "  Sequence : FPIKWTAPESLAYNKFSIKSDVWAFGVLLWEIATYGMSPYPGIDLSQVYELLEKDYRMERPEGCPEKVYELMRACWQWNPSDRPSFAEIHQAFETMFQES\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : SISDEVEKELGKQGVRGAVSTLLQAPELPTKTRTSRRAAEHRDTTDVPEMPHSKGQGESDPLDHEPAVSPLLPRKERGPPEGGLNEDERLLPKDKKTNLF\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : SALIKKKKKTAPTPPKRSSSFREMDGQPERRGAGEEEGRDISNGALAFTPLDTADPAKSPKPSNGAGVPNGALRESGGSGFRSPHLWKKSSTLTSSRLAT\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : GEEEGGGSSSKRFLRSCSASCVPHGAKDTEWRSVTLPRDLQSTGRQFDSSTFGGHKSEKPALPRKRAGENRSDQVTRGTVTPPPRLVKKNEEAADEVFKD\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : IMESSPGSSPPNLTPKPLRRQVTVAPASGLPHKEEAGKGSALGTPAAAEPVTPTSKAGSGAPGGTSKGPAEESRVRRHKHSSESPGRDKGKLSRLKPAPP\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : PPPAASAGKAGGKPSQSPSQEAAGEAVLGAKTKATSLVDAVNSDAAKPSQPGEGLKKPVLPATPKPQSAKPSGTPISPAPVPSTLPSASSALAGDQPSST\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : AFIPLISTRVSLRKTRQPPERIASGAITKGVVLDSTEALCLAISRNSEQMASHSAVLEAGKNLYTFCVSYVDSIQQMRNKFAFREAINKLENNLRELQIC\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : PATAGSGPAATQDFSKLLSSVKEISDIVQR\n",
      "Prediction :                               \n",
      "     Score :                               \n",
      "AA2AR_Regadenoson\n",
      "DTI score :  [0.97712976]\n",
      "  Sequence : MPIMGSSVYITVELAIAVLAILGNVLVCWAVWLNSNLQNVTNYFVVSLAAADIAVGVLAIPFAITISTGFCAACHGCLFIACFVLVLTQSSIFSLLAIAI\n",
      "Prediction :                                                            IPFAITISTGFCAACHG    ACFVLVLTQSSIFSLL    \n",
      "     Score :                                                            65%                  65%                 \n",
      "  Sequence : DRYIAIRIPLRYNGLVTGTRAKGIIAICWVLSFAIGLTPMLGWNNCGQPKEGKNHSQGCGEGQVACLFEDVVPMNYMVYFNFFACVLVPLLLMLGVYLRI\n",
      "Prediction :                                                                      DVVPMNYMVYFNFFA                \n",
      "     Score :                                                                      67%                            \n",
      "  Sequence : FLAARRQLKQMESQPLPGERARSTLQKEVHAAKSLAIIVGLFALCWLPLHIINCFTFFCPDCSHAPLWLMYLAIVLSHTNSVVNPFIYAYRIREFRQTFR\n",
      "Prediction :                                                        TFFCPDCSHAPLWLMYL     HTNSVVNPFIYAYRIR       \n",
      "     Score :                                                        61%                   61%                    \n",
      "  Sequence : KIIRSHVLRQQEPFKAAGTSARVLAAHGSDGEQVSLRLNGHPPGVWANGSAPHPERRPNGYALGLVSGGSAQESQGNTGLPDVELLSHELKGVCPEPPGL\n",
      "Prediction :                                                                                                     \n",
      "     Score :                                                                                                     \n",
      "  Sequence : DDPLAQDGAGVS\n",
      "Prediction :             \n",
      "     Score :             \n"
     ]
    }
   ],
   "source": [
    "dti_model.HoTS_visualization(drugs_fp, targets_encoded, targets, protein_names=names, th=0.6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "HoTS-tf2_version",
   "language": "python",
   "name": "hots-tf2_version"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}