RUN_ME.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "bb76e1d1",
   "metadata": {},
   "source": [
    "**In this notebook, we will use SQUID to generate shape-similar analogues for a handful of 3D compounds from our test set.**\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "386261eb",
   "metadata": {},
   "source": [
    "Before you run these cells, you will need to download some data from Figshare: https://figshare.com/s/3d2f8fd57d9a65fe237e\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "119f5b68",
   "metadata": {},
   "source": [
    " "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "139ed1f5",
   "metadata": {},
   "source": [
    "First, we import some modules."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c9c2c253",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch_geometric\n",
    "import torch\n",
    "import torch_scatter\n",
    "import math\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from copy import deepcopy\n",
    "import rdkit\n",
    "import rdkit.Chem\n",
    "import rdkit.Chem.AllChem\n",
    "import rdkit.Chem.rdMolTransforms\n",
    "import networkx as nx\n",
    "import random\n",
    "from tqdm import tqdm\n",
    "from rdkit.Chem import rdMolTransforms\n",
    "import itertools\n",
    "import os\n",
    "import pickle\n",
    "\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from models.vnn.models.vn_layers import *\n",
    "from models.vnn.models.utils.vn_dgcnn_util import get_graph_feature\n",
    "\n",
    "from utils.general_utils import *\n",
    "\n",
    "from models.EGNN import *\n",
    "from models.models import *"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ec74200b",
   "metadata": {},
   "source": [
    "We will now specify the following generation hyperparameters:\n",
    "\n",
    "- The interpolation factor $\\lambda$ controlling the degree of lienar interpolation between the posterior and prior\n",
    "- The threshold for local stopping\n",
    "- Scale factor to scale the standard deviation after interpolation (default is 1.0)\n",
    "- Whether to ablate equivariance\n",
    "    \n",
    "We also specify some sampling limits. Here, we will generate 20 molecules for 10 different targets.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "36997247",
   "metadata": {},
   "outputs": [],
   "source": [
    "interpolate_to_GNN_prior = 1.0 # 'prior'\n",
    "stop_threshold = 0.01\n",
    "variational_GNN_factor = 1.0\n",
    "ablateEqui = False\n",
    "\n",
    "repetitions = 20\n",
    "total_evaluations = 10 "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5fdbf1aa",
   "metadata": {},
   "source": [
    "We now load in some necessary data containing:\n",
    "- The atom/fragment library\n",
    "- bond distance lookup table\n",
    "- array of unique atom types\n",
    "- 3D molecules in the test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "1cd875be",
   "metadata": {},
   "outputs": [],
   "source": [
    "use_artificial_mols = True\n",
    "\n",
    "AtomFragment_database = pd.read_pickle('data/MOSES2/MOSES2_training_val_AtomFragment_database.pkl')\n",
    "AtomFragment_database = AtomFragment_database.iloc[1:].reset_index(drop = True) # removing stop token from AtomFragment_database\n",
    "fragment_library_atom_features = np.concatenate(AtomFragment_database['atom_features'], axis = 0).reshape((len(AtomFragment_database), -1))\n",
    "\n",
    "bond_lookup = pd.read_pickle('data/MOSES2/MOSES2_training_val_bond_lookup.pkl')\n",
    "unique_atoms = np.load('data/MOSES2/MOSES2_training_val_unique_atoms.npy') \n",
    "\n",
    "test_mol_df = pd.read_pickle('data/MOSES2/test_MOSES_filtered_artificial_mols.pkl')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "08a280f8",
   "metadata": {},
   "source": [
    "We will encode 'artificial molecules' with fixed bonding geometries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "24d3ed3b",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_mols = list(test_mol_df.artificial_mol)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3bb14032",
   "metadata": {},
   "source": [
    "and shuffle the test set..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "221eddbb",
   "metadata": {},
   "outputs": [],
   "source": [
    "random.seed(0)\n",
    "indices_to_evaluate = list(range(0, len(test_mols)))\n",
    "random.shuffle(indices_to_evaluate)\n",
    "\n",
    "val_mols = [test_mols[i] for i in indices_to_evaluate]\n",
    "val_mols_index = indices_to_evaluate"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "18659848",
   "metadata": {},
   "source": [
    "We will generate new molecules using our two trained models:\n",
    "1. graph generator\n",
    "2. rotatable bond scorer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "5ee72dba",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not ablateEqui:\n",
    "    model_3D_PATH = 'trained_models/graph_generator.pt'\n",
    "    rocs_model_3D_PATH = 'trained_models/scorer.pt'\n",
    "else:\n",
    "    model_3D_PATH = 'trained_models/graph_generator_ablateEqui.pt'\n",
    "    rocs_model_3D_PATH = 'trained_models/scorer_ablateEqui.pt'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "b94d8fa4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ROCS_Model_Point_Cloud(\n",
       "  (Encoder): Encoder_point_cloud(\n",
       "    (fragment_encoder): FragmentLibraryEncoder(\n",
       "      (EGNN_layers): ModuleList(\n",
       "        (0): EGNN_static(\n",
       "          (edge_mlp): Sequential(\n",
       "            (0): Linear(in_features=96, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "            (3): LeakyReLU(negative_slope=0.2)\n",
       "          )\n",
       "          (node_mlp): Sequential(\n",
       "            (0): Linear(in_features=109, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "          )\n",
       "        )\n",
       "        (1): EGNN_static(\n",
       "          (edge_mlp): Sequential(\n",
       "            (0): Linear(in_features=193, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "            (3): LeakyReLU(negative_slope=0.2)\n",
       "          )\n",
       "          (node_mlp): Sequential(\n",
       "            (0): Linear(in_features=128, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "          )\n",
       "        )\n",
       "        (2): EGNN_static(\n",
       "          (edge_mlp): Sequential(\n",
       "            (0): Linear(in_features=193, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "            (3): LeakyReLU(negative_slope=0.2)\n",
       "          )\n",
       "          (node_mlp): Sequential(\n",
       "            (0): Linear(in_features=128, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "    )\n",
       "    (GraphEncoder): EGNN_VN_Encoder_point_cloud(\n",
       "      (std_feature_mix_shape_to_nodes): VNStdFeature(\n",
       "        (vn1): VNLinearLeakyReLU(\n",
       "          (map_to_feat): Linear(in_features=128, out_features=64, bias=False)\n",
       "          (map_to_dir): Linear(in_features=128, out_features=64, bias=False)\n",
       "        )\n",
       "        (vn2): VNLinearLeakyReLU(\n",
       "          (map_to_feat): Linear(in_features=64, out_features=32, bias=False)\n",
       "          (map_to_dir): Linear(in_features=64, out_features=32, bias=False)\n",
       "        )\n",
       "        (vn_lin): Linear(in_features=32, out_features=3, bias=False)\n",
       "      )\n",
       "      (EGNN_layers): ModuleList(\n",
       "        (0): EGNN_static(\n",
       "          (edge_mlp): Sequential(\n",
       "            (0): Linear(in_features=992, out_features=128, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=128, out_features=64, bias=True)\n",
       "            (3): LeakyReLU(negative_slope=0.2)\n",
       "          )\n",
       "          (node_mlp): Sequential(\n",
       "            (0): Linear(in_features=557, out_features=128, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=128, out_features=64, bias=True)\n",
       "          )\n",
       "        )\n",
       "        (1): EGNN_static(\n",
       "          (edge_mlp): Sequential(\n",
       "            (0): Linear(in_features=193, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "            (3): LeakyReLU(negative_slope=0.2)\n",
       "          )\n",
       "          (node_mlp): Sequential(\n",
       "            (0): Linear(in_features=128, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "          )\n",
       "        )\n",
       "        (2): EGNN_static(\n",
       "          (edge_mlp): Sequential(\n",
       "            (0): Linear(in_features=193, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "            (3): LeakyReLU(negative_slope=0.2)\n",
       "          )\n",
       "          (node_mlp): Sequential(\n",
       "            (0): Linear(in_features=128, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (project_h_embeddings): Sequential(\n",
       "        (0): Linear(in_features=64, out_features=64, bias=True)\n",
       "        (1): LeakyReLU(negative_slope=0.2)\n",
       "        (2): Linear(in_features=64, out_features=2048, bias=True)\n",
       "      )\n",
       "      (Equi_linear_leaky_mixing): VNLinearAndLeakyReLU(\n",
       "        (linear): VNLinear(\n",
       "          (map_to_feat): Linear(in_features=96, out_features=64, bias=False)\n",
       "        )\n",
       "        (leaky_relu): VNLeakyReLU(\n",
       "          (map_to_dir): Linear(in_features=64, out_features=64, bias=False)\n",
       "        )\n",
       "      )\n",
       "      (conv1): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=2, out_features=10, bias=False)\n",
       "        (map_to_dir): Linear(in_features=2, out_features=10, bias=False)\n",
       "      )\n",
       "      (conv2): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=20, out_features=10, bias=False)\n",
       "        (map_to_dir): Linear(in_features=20, out_features=10, bias=False)\n",
       "      )\n",
       "      (conv3): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=20, out_features=21, bias=False)\n",
       "        (map_to_dir): Linear(in_features=20, out_features=21, bias=False)\n",
       "      )\n",
       "      (conv4): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=42, out_features=42, bias=False)\n",
       "        (map_to_dir): Linear(in_features=42, out_features=42, bias=False)\n",
       "      )\n",
       "      (conv5): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=83, out_features=64, bias=False)\n",
       "        (map_to_dir): Linear(in_features=83, out_features=1, bias=False)\n",
       "      )\n",
       "      (std_feature): VNStdFeature(\n",
       "        (vn1): VNLinearLeakyReLU(\n",
       "          (map_to_feat): Linear(in_features=128, out_features=64, bias=False)\n",
       "          (map_to_dir): Linear(in_features=128, out_features=64, bias=False)\n",
       "        )\n",
       "        (vn2): VNLinearLeakyReLU(\n",
       "          (map_to_feat): Linear(in_features=64, out_features=32, bias=False)\n",
       "          (map_to_dir): Linear(in_features=64, out_features=32, bias=False)\n",
       "        )\n",
       "        (vn_lin): Linear(in_features=32, out_features=3, bias=False)\n",
       "      )\n",
       "      (point_invariant_mlp): Sequential(\n",
       "        (0): Linear(in_features=448, out_features=128, bias=True)\n",
       "        (1): LeakyReLU(negative_slope=0.2)\n",
       "        (2): Linear(in_features=128, out_features=64, bias=True)\n",
       "        (3): LeakyReLU(negative_slope=0.2)\n",
       "        (4): Linear(in_features=64, out_features=64, bias=True)\n",
       "      )\n",
       "    )\n",
       "    (SubGraphEncoder): EGNN_VN_Encoder_point_cloud(\n",
       "      (std_feature_mix_shape_to_nodes): VNStdFeature(\n",
       "        (vn1): VNLinearLeakyReLU(\n",
       "          (map_to_feat): Linear(in_features=128, out_features=64, bias=False)\n",
       "          (map_to_dir): Linear(in_features=128, out_features=64, bias=False)\n",
       "        )\n",
       "        (vn2): VNLinearLeakyReLU(\n",
       "          (map_to_feat): Linear(in_features=64, out_features=32, bias=False)\n",
       "          (map_to_dir): Linear(in_features=64, out_features=32, bias=False)\n",
       "        )\n",
       "        (vn_lin): Linear(in_features=32, out_features=3, bias=False)\n",
       "      )\n",
       "      (EGNN_layers): ModuleList(\n",
       "        (0): EGNN_static(\n",
       "          (edge_mlp): Sequential(\n",
       "            (0): Linear(in_features=992, out_features=128, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=128, out_features=64, bias=True)\n",
       "            (3): LeakyReLU(negative_slope=0.2)\n",
       "          )\n",
       "          (node_mlp): Sequential(\n",
       "            (0): Linear(in_features=557, out_features=128, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=128, out_features=64, bias=True)\n",
       "          )\n",
       "        )\n",
       "        (1): EGNN_static(\n",
       "          (edge_mlp): Sequential(\n",
       "            (0): Linear(in_features=193, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "            (3): LeakyReLU(negative_slope=0.2)\n",
       "          )\n",
       "          (node_mlp): Sequential(\n",
       "            (0): Linear(in_features=128, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "          )\n",
       "        )\n",
       "        (2): EGNN_static(\n",
       "          (edge_mlp): Sequential(\n",
       "            (0): Linear(in_features=193, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "            (3): LeakyReLU(negative_slope=0.2)\n",
       "          )\n",
       "          (node_mlp): Sequential(\n",
       "            (0): Linear(in_features=128, out_features=64, bias=True)\n",
       "            (1): LeakyReLU(negative_slope=0.2)\n",
       "            (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (project_h_embeddings): Sequential(\n",
       "        (0): Linear(in_features=64, out_features=64, bias=True)\n",
       "        (1): LeakyReLU(negative_slope=0.2)\n",
       "        (2): Linear(in_features=64, out_features=2048, bias=True)\n",
       "      )\n",
       "      (Equi_linear_leaky_mixing): VNLinearAndLeakyReLU(\n",
       "        (linear): VNLinear(\n",
       "          (map_to_feat): Linear(in_features=96, out_features=64, bias=False)\n",
       "        )\n",
       "        (leaky_relu): VNLeakyReLU(\n",
       "          (map_to_dir): Linear(in_features=64, out_features=64, bias=False)\n",
       "        )\n",
       "      )\n",
       "      (conv1): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=2, out_features=10, bias=False)\n",
       "        (map_to_dir): Linear(in_features=2, out_features=10, bias=False)\n",
       "      )\n",
       "      (conv2): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=20, out_features=10, bias=False)\n",
       "        (map_to_dir): Linear(in_features=20, out_features=10, bias=False)\n",
       "      )\n",
       "      (conv3): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=20, out_features=21, bias=False)\n",
       "        (map_to_dir): Linear(in_features=20, out_features=21, bias=False)\n",
       "      )\n",
       "      (conv4): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=42, out_features=42, bias=False)\n",
       "        (map_to_dir): Linear(in_features=42, out_features=42, bias=False)\n",
       "      )\n",
       "      (conv5): VNLinearLeakyReLU(\n",
       "        (map_to_feat): Linear(in_features=83, out_features=64, bias=False)\n",
       "        (map_to_dir): Linear(in_features=83, out_features=1, bias=False)\n",
       "      )\n",
       "      (std_feature): VNStdFeature(\n",
       "        (vn1): VNLinearLeakyReLU(\n",
       "          (map_to_feat): Linear(in_features=128, out_features=64, bias=False)\n",
       "          (map_to_dir): Linear(in_features=128, out_features=64, bias=False)\n",
       "        )\n",
       "        (vn2): VNLinearLeakyReLU(\n",
       "          (map_to_feat): Linear(in_features=64, out_features=32, bias=False)\n",
       "          (map_to_dir): Linear(in_features=64, out_features=32, bias=False)\n",
       "        )\n",
       "        (vn_lin): Linear(in_features=32, out_features=3, bias=False)\n",
       "      )\n",
       "      (point_invariant_mlp): Sequential(\n",
       "        (0): Linear(in_features=448, out_features=128, bias=True)\n",
       "        (1): LeakyReLU(negative_slope=0.2)\n",
       "        (2): Linear(in_features=128, out_features=64, bias=True)\n",
       "        (3): LeakyReLU(negative_slope=0.2)\n",
       "        (4): Linear(in_features=64, out_features=64, bias=True)\n",
       "      )\n",
       "    )\n",
       "    (Equi_linear_leaky_1): VNLinearAndLeakyReLU(\n",
       "      (linear): VNLinear(\n",
       "        (map_to_feat): Linear(in_features=256, out_features=128, bias=False)\n",
       "      )\n",
       "      (leaky_relu): VNLeakyReLU(\n",
       "        (map_to_dir): Linear(in_features=128, out_features=128, bias=False)\n",
       "      )\n",
       "    )\n",
       "    (Equi_linear_leaky_2): VNLinearAndLeakyReLU(\n",
       "      (linear): VNLinear(\n",
       "        (map_to_feat): Linear(in_features=128, out_features=64, bias=False)\n",
       "      )\n",
       "      (leaky_relu): VNLeakyReLU(\n",
       "        (map_to_dir): Linear(in_features=64, out_features=64, bias=False)\n",
       "      )\n",
       "    )\n",
       "    (Equi_linear_leaky_3): VNLinearAndLeakyReLU(\n",
       "      (linear): VNLinear(\n",
       "        (map_to_feat): Linear(in_features=64, out_features=64, bias=False)\n",
       "      )\n",
       "      (leaky_relu): VNLeakyReLU(\n",
       "        (map_to_dir): Linear(in_features=64, out_features=64, bias=False)\n",
       "      )\n",
       "    )\n",
       "    (T_layer): VNLinearAndLeakyReLU(\n",
       "      (linear): VNLinear(\n",
       "        (map_to_feat): Linear(in_features=64, out_features=3, bias=False)\n",
       "      )\n",
       "      (leaky_relu): VNLeakyReLU(\n",
       "        (map_to_dir): Linear(in_features=3, out_features=3, bias=False)\n",
       "      )\n",
       "    )\n",
       "  )\n",
       "  (ROCS_scorer): EGNN_MLP(\n",
       "    (net): Sequential(\n",
       "      (0): Linear(in_features=448, out_features=64, bias=True)\n",
       "      (1): LeakyReLU(negative_slope=0.2)\n",
       "      (2): Linear(in_features=64, out_features=64, bias=True)\n",
       "      (3): LeakyReLU(negative_slope=0.2)\n",
       "      (4): Linear(in_features=64, out_features=64, bias=True)\n",
       "      (5): LeakyReLU(negative_slope=0.2)\n",
       "      (6): Linear(in_features=64, out_features=1, bias=True)\n",
       "    )\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# HYPERPARAMETERS for 3D graph generator\n",
    "pointCloudVar = 1. / (12. * 1.7) \n",
    "\n",
    "model_3D = Model_Point_Cloud_Switched(\n",
    "    input_nf = 45, \n",
    "    edges_in_d = 5, \n",
    "    n_knn = 5, \n",
    "    conv_dims = [32, 32, 64, 128], \n",
    "    num_components = 64, \n",
    "    fragment_library_dim = 64, \n",
    "    N_fragment_layers = 3, \n",
    "    append_noise = False, \n",
    "    N_members = 125 - 1, \n",
    "    EGNN_layer_dim = 64, \n",
    "    N_EGNN_layers = 3, \n",
    "    output_MLP_hidden_dim = 64, \n",
    "    pooling_MLP = False, \n",
    "    shared_encoders = False, \n",
    "    subtract_latent_space = True,\n",
    "    variational = False,\n",
    "    variational_mode = 'inv', # not used\n",
    "    variational_GNN = True,\n",
    "    \n",
    "    mix_node_inv_to_equi = True,\n",
    "    mix_shape_to_nodes = True,\n",
    "    ablate_HvarCat = False,\n",
    "    \n",
    "    predict_pairwise_properties = False,\n",
    "    predict_mol_property = False,\n",
    "    \n",
    "    ablateEqui = ablateEqui,\n",
    "    \n",
    "    old_EGNN = False,\n",
    "    \n",
    ").float()\n",
    "\n",
    "model_3D.load_state_dict(torch.load(model_3D_PATH, map_location=next(model_3D.parameters()).device), strict = True)\n",
    "model_3D.eval()\n",
    "\n",
    "\n",
    "# HYPERPARAMETERS for ROCS scorer\n",
    "rocs_pointCloudVar = 1. / (12. * 1.7) \n",
    "\n",
    "rocs_model_3D = ROCS_Model_Point_Cloud(\n",
    "    input_nf = 45, \n",
    "    edges_in_d = 5, \n",
    "    n_knn = 10, \n",
    "    conv_dims = [32, 32, 64, 128], \n",
    "    num_components = 64, \n",
    "    fragment_library_dim = 64,\n",
    "    N_fragment_layers = 3, \n",
    "    append_noise = False, \n",
    "    N_members = 125 - 1, \n",
    "    EGNN_layer_dim = 64, \n",
    "    N_EGNN_layers = 3, \n",
    "    output_MLP_hidden_dim = 64, \n",
    "    pooling_MLP = False, \n",
    "    shared_encoders = False, \n",
    "    subtract_latent_space = True,\n",
    "    variational = False,\n",
    "    variational_mode = 'inv', # not used\n",
    "    variational_GNN = False,\n",
    "    \n",
    "    mix_node_inv_to_equi = True,\n",
    "    mix_shape_to_nodes = True,\n",
    "    ablate_HvarCat = False,\n",
    "    \n",
    "    ablateEqui = ablateEqui,\n",
    "    \n",
    "    old_EGNN = False,\n",
    "    \n",
    ").float()\n",
    "\n",
    "rocs_model_3D.load_state_dict(torch.load(rocs_model_3D_PATH, map_location=next(rocs_model_3D.parameters()).device), strict = True)\n",
    "rocs_model_3D.eval()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "493d815e",
   "metadata": {},
   "source": [
    "Finally, we use the models to generate new 3D molecules conditioned on the target shape.\n",
    "Here, we report the chemical similarity and the shape similarity before ROCS-alignment. \n",
    "If you have access to an Openeye license, we will compute the aligned shape similarities in the next cells.\n",
    "\n",
    "We will generate 20*10 = 200 molecules total. Each molecule takes about 2-3 seconds to generate on a cpu, so this next cell\n",
    "will run for a few minutes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "fd91c6c0",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Encoded Molecule 1:\n",
      "(non-aligned) shape scores: [0.399, 0.645, 0.502, 0.713, 0.531, 0.587, 0.582, 0.601, 0.502, 0.622, 0.378, 0.576, 0.443, 0.645, 0.522, 0.792, 0.765, 0.536, 0.801, 0.491]\n",
      "tanimoto chemical similarity: [0.261, 0.253, 0.271, 0.326, 0.285, 0.259, 0.289, 0.291, 0.368, 0.254, 0.191, 0.359, 0.274, 0.157, 0.237, 0.305, 0.291, 0.354, 0.349, 0.249]\n",
      "\n",
      "Encoded Molecule 2:\n",
      "(non-aligned) shape scores: [0.413, 0.533, 0.354, 0.484, 0.626, 0.499, 0.769, 0.567, 0.755, 0.737, 0.767, 0.781, 0.575, 0.635, 0.711, 0.729, 0.775, 0.704, 0.799, 0.472]\n",
      "tanimoto chemical similarity: [0.338, 0.416, 0.155, 0.28, 0.278, 0.259, 0.454, 0.251, 0.203, 0.204, 0.266, 0.238, 0.262, 0.298, 0.307, 0.327, 0.346, 0.274, 0.237, 0.182]\n",
      "\n",
      "Encoded Molecule 3:\n",
      "(non-aligned) shape scores: [0.488, 0.54, 0.377, 0.793, 0.387, 0.545, 0.457, 0.701, 0.556, 0.513, 0.488, 0.484, 0.423, 0.731, 0.406, 0.345, 0.581, 0.627, 0.454, 0.526]\n",
      "tanimoto chemical similarity: [0.271, 0.163, 0.128, 0.267, 0.297, 0.216, 0.18, 0.237, 0.275, 0.288, 0.271, 0.112, 0.187, 0.25, 0.15, 0.165, 0.161, 0.165, 0.298, 0.227]\n",
      "\n",
      "Encoded Molecule 4:\n",
      "(non-aligned) shape scores: [0.228, 0.248, 0.501, 0.334, 0.465, 0.426, 0.466, 0.386, 0.471, 0.425, 0.424, 0.318, 0.546, 0.444, 0.381, 0.324, 0.316, 0.315, 0.457, 0.4]\n",
      "tanimoto chemical similarity: [0.238, 0.236, 0.139, 0.206, 0.204, 0.217, 0.22, 0.202, 0.237, 0.176, 0.213, 0.205, 0.231, 0.211, 0.192, 0.17, 0.223, 0.167, 0.216, 0.183]\n",
      "\n",
      "Encoded Molecule 5:\n",
      "(non-aligned) shape scores: [0.421, 0.531, 0.592, 0.587, 0.555, 0.602, 0.595, 0.618, 0.623, 0.372, 0.652, 0.555, 0.375, 0.428, 0.578, 0.414, 0.562, 0.699, 0.542, 0.384]\n",
      "tanimoto chemical similarity: [0.163, 0.159, 0.153, 0.18, 0.136, 0.213, 0.165, 0.138, 0.171, 0.16, 0.166, 0.135, 0.147, 0.139, 0.113, 0.152, 0.144, 0.154, 0.16, 0.154]\n",
      "\n",
      "Encoded Molecule 6:\n",
      "(non-aligned) shape scores: [0.697, 0.547, 0.688, 0.455, 0.654, 0.574, 0.733, 0.813, 0.574, 0.451, 0.303, 0.476, 0.683, 0.547, 0.786, 0.805, 0.538, 0.597, 0.679, 0.755]\n",
      "tanimoto chemical similarity: [0.214, 0.145, 0.191, 0.188, 0.227, 0.216, 0.214, 0.18, 0.196, 0.164, 0.114, 0.206, 0.196, 0.201, 0.204, 0.215, 0.221, 0.171, 0.189, 0.183]\n",
      "\n",
      "Encoded Molecule 7:\n",
      "(non-aligned) shape scores: [0.466, 0.361, 0.593, 0.402, 0.461, 0.46, 0.349, 0.262, 0.533, 0.471, 0.199, 0.363, 0.532, 0.263, 0.366, 0.199, 0.307, 0.469, 0.275, 0.327]\n",
      "tanimoto chemical similarity: [0.224, 0.216, 0.175, 0.176, 0.285, 0.258, 0.118, 0.189, 0.237, 0.209, 0.128, 0.315, 0.175, 0.178, 0.361, 0.122, 0.193, 0.15, 0.18, 0.201]\n",
      "\n",
      "Encoded Molecule 8:\n",
      "(non-aligned) shape scores: [0.525, 0.801, 0.371, 0.542, 0.647, 0.71, 0.64, 0.703, 0.651, 0.373, 0.622, 0.524, 0.615, 0.661, 0.533, 0.416, 0.605, 0.587, 0.564, 0.356]\n",
      "tanimoto chemical similarity: [0.23, 0.357, 0.205, 0.254, 0.128, 0.284, 0.29, 0.212, 0.277, 0.337, 0.188, 0.291, 0.224, 0.269, 0.212, 0.348, 0.246, 0.255, 0.237, 0.346]\n",
      "\n",
      "Encoded Molecule 9:\n",
      "(non-aligned) shape scores: [0.487, 0.554, 0.626, 0.494, 0.682, 0.605, 0.684, 0.676, 0.604, 0.673, 0.64, 0.583, 0.766, 0.623, 0.565, 0.675, 0.57, 0.885, 0.713, 0.708]\n",
      "tanimoto chemical similarity: [0.256, 0.204, 0.222, 0.281, 0.309, 0.407, 0.339, 0.337, 0.33, 0.299, 0.353, 0.157, 0.459, 0.19, 0.214, 0.187, 0.187, 0.602, 0.441, 0.326]\n",
      "\n",
      "Encoded Molecule 10:\n",
      "(non-aligned) shape scores: [0.506, 0.537, 0.427, 0.646, 0.33, 0.522, 0.229, 0.78, 0.464, 0.64, 0.399, 0.618, 0.441, 0.69, 0.556, 0.537, 0.41, 0.636, 0.361, 0.713]\n",
      "tanimoto chemical similarity: [0.444, 0.347, 0.303, 0.313, 0.203, 0.278, 0.15, 0.336, 0.237, 0.372, 0.218, 0.245, 0.288, 0.175, 0.271, 0.363, 0.203, 0.381, 0.239, 0.388]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# muting noisy warnings\n",
    "from rdkit import RDLogger\n",
    "import warnings\n",
    "lg = RDLogger.logger()\n",
    "lg.setLevel(RDLogger.CRITICAL)\n",
    "warnings.filterwarnings(\"ignore\", category=DeprecationWarning) \n",
    "warnings.filterwarnings(\"ignore\", category=UserWarning) \n",
    "warnings.filterwarnings(\"ignore\", category=Warning) \n",
    "np.warnings.filterwarnings('ignore', category=np.VisibleDeprecationWarning)\n",
    "\n",
    "\n",
    "file_idx = 0\n",
    "\n",
    "reference_mols_list = []\n",
    "unaligned_mols_list = []\n",
    "\n",
    "for m_idx, m_ in enumerate(val_mols):\n",
    "    \n",
    "    seed = 0\n",
    "    random.seed(seed)\n",
    "    np.random.seed(seed = seed)\n",
    "    torch.manual_seed(seed)\n",
    "    \n",
    "    m = deepcopy(m_)\n",
    "    \n",
    "    mol = deepcopy(m)\n",
    "    xyz = np.array(mol.GetConformer().GetPositions())\n",
    "    center_of_mass = np.sum(xyz, axis = 0) / xyz.shape[0]\n",
    "    xyz_centered = xyz - center_of_mass\n",
    "    for i in range(0, mol.GetNumAtoms()):\n",
    "        x,y,z = xyz_centered[i]\n",
    "        mol.GetConformer().SetAtomPosition(i, Point3D(x,y,z))\n",
    "    \n",
    "    m = deepcopy(mol)\n",
    "    \n",
    "    mol_target = deepcopy(m)\n",
    "    ring_fragments = get_ring_fragments(mol_target)\n",
    "    all_possible_seeds = get_all_possible_seeds(mol_target, ring_fragments)\n",
    "    terminal_seeds = filter_terminal_seeds(all_possible_seeds, mol_target)\n",
    "    \n",
    "    select_seeds = get_starting_seeds(mol_target, AtomFragment_database, fragment_library_atom_features, unique_atoms, bond_lookup)\n",
    "    \n",
    "    if len(select_seeds) == 0:\n",
    "        continue\n",
    "    \n",
    "    random_seed_selection = random.randint(0, len(select_seeds) - 1)\n",
    "    select_seeds = [select_seeds[random_seed_selection]] * repetitions\n",
    "    \n",
    "    repeated_rocs = []\n",
    "    repeated_tanimoto = []\n",
    "    \n",
    "    reference_mols = []\n",
    "    unaligned_mols = []\n",
    "    \n",
    "    \n",
    "    for seed in select_seeds:\n",
    "\n",
    "        mol = deepcopy(m)\n",
    "        \n",
    "        # extracting starting seed and preparing to generate\n",
    "        \n",
    "        frame_generation, frame_rocs = get_frame_terminalSeeds(mol, seed, AtomFragment_database, include_rocs = True)\n",
    "        positions = list(frame_rocs.iloc[0].positions_before)\n",
    "        start = 0\n",
    "        for i in range(len(frame_generation)):\n",
    "            if (set(frame_generation.iloc[i].partial_graph_indices) == set(positions)) & (frame_generation.iloc[i].next_atom_index == -1):\n",
    "                start = i + 1\n",
    "                break\n",
    "        \n",
    "        if len(frame_generation.iloc[0].partial_graph_indices) == 1: # seed is a terminal ATOM\n",
    "            terminalSeed_frame = frame_generation.iloc[0:start].reset_index(drop = True)\n",
    "                    \n",
    "            sequence = get_ground_truth_generation_sequence(terminalSeed_frame, AtomFragment_database, fragment_library_atom_features)\n",
    "                \n",
    "            mol = deepcopy(terminalSeed_frame.iloc[0].rdkit_mol_cistrans_stereo)\n",
    "            partial_indices = deepcopy(terminalSeed_frame.iloc[0].partial_graph_indices_sorted)\n",
    "            \n",
    "            final_partial_indices = deepcopy(terminalSeed_frame.iloc[-1].partial_graph_indices_sorted)\n",
    "            ring_fragments = get_ring_fragments(mol)\n",
    "            add_to_partial = [list(f) for p in final_partial_indices for f in ring_fragments if p in f]\n",
    "            add_to_partial = [item for sublist in add_to_partial for item in sublist]\n",
    "            final_partial_indices = list(set(final_partial_indices).union(add_to_partial))\n",
    "                \n",
    "            queue_indices = deepcopy(terminalSeed_frame.iloc[0].focal_indices_sorted)\n",
    "            \n",
    "            _, seed_mol, queue, positioned_atoms_indices, atom_to_library_ID_map, _, _, _ = generate_seed_from_sequence(sequence, mol, partial_indices, queue_indices, AtomFragment_database, unique_atoms, bond_lookup, stop_after_sequence = True)\n",
    "    \n",
    "            seed_node_features = getNodeFeatures(seed_mol.GetAtoms())\n",
    "            \n",
    "            for k in atom_to_library_ID_map:\n",
    "                seed_node_features[k] = AtomFragment_database.iloc[atom_to_library_ID_map[k]].atom_features\n",
    "                \n",
    "            G = get_substructure_graph(mol, final_partial_indices)\n",
    "            G_seed = get_substructure_graph(seed_mol, list(range(0, seed_mol.GetNumAtoms())), node_features = seed_node_features)\n",
    "            nm = nx.algorithms.isomorphism.generic_node_match(['atom_features'], [None], [np.allclose])\n",
    "            em = nx.algorithms.isomorphism.numerical_edge_match(\"bond_type\", 1.0)\n",
    "            GM = nx.algorithms.isomorphism.GraphMatcher(G, G_seed, node_match = nm, edge_match = em)\n",
    "            assert GM.is_isomorphic()\n",
    "            idx_map = GM.mapping\n",
    "                \n",
    "        else: # seed is a terminal FRAGMENT\n",
    "            partial_indices = deepcopy(frame_generation.iloc[0].partial_graph_indices_sorted)\n",
    "            final_partial_indices = partial_indices\n",
    "            seed_mol = generate_conformer(get_fragment_smiles(mol, partial_indices))\n",
    "            idx_map = get_reindexing_map(mol, partial_indices, seed_mol)\n",
    "            positioned_atoms_indices = sorted([idx_map[f] for f in final_partial_indices])\n",
    "            \n",
    "            atom_to_library_ID_map = {} # no individual atoms yet generated\n",
    "            queue = [0] # 0 can be considered the focal root node\n",
    "    \n",
    "        for i in final_partial_indices:\n",
    "            x,y,z = mol.GetConformer().GetPositions()[i]\n",
    "            seed_mol.GetConformer().SetAtomPosition(idx_map[i], Point3D(x,y,z)) \n",
    "        \n",
    "        \n",
    "        \n",
    "        starting_queue = deepcopy(queue)\n",
    "        try:\n",
    "            _, updated_mol, _, _, _, N_rocs_decisions, _, _, _, chirality_scored = generate_3D_mol_from_sequence(\n",
    "                sequence = [], \n",
    "                partial_mol = deepcopy(seed_mol), \n",
    "                mol = deepcopy(mol_target), \n",
    "                positioned_atoms_indices = deepcopy(positioned_atoms_indices), \n",
    "                queue = starting_queue, \n",
    "                atom_to_library_ID_map = deepcopy(atom_to_library_ID_map), \n",
    "                model = model_3D, \n",
    "                rocs_model = rocs_model_3D,\n",
    "                AtomFragment_database = AtomFragment_database,\n",
    "                unique_atoms = unique_atoms, \n",
    "                bond_lookup = bond_lookup,\n",
    "                N_points = 5, \n",
    "                N_points_rocs = 5,\n",
    "                stop_after_sequence = False,\n",
    "                mask_first_stop = False,\n",
    "                stochastic = False, \n",
    "                chirality_scoring = True,\n",
    "                stop_threshold = stop_threshold,\n",
    "                steric_mask = True,\n",
    "                \n",
    "                variational_factor_equi = 0.0,\n",
    "                variational_factor_inv = 0.0, \n",
    "                interpolate_to_prior_equi = 0.0,\n",
    "                interpolate_to_prior_inv = 0.0, \n",
    "                \n",
    "                use_variational_GNN = True, \n",
    "                variational_GNN_factor = variational_GNN_factor, \n",
    "                interpolate_to_GNN_prior = interpolate_to_GNN_prior, \n",
    "                \n",
    "                rocs_use_variational_GNN = False, \n",
    "                rocs_variational_GNN_factor = 0.0, \n",
    "                rocs_interpolate_to_GNN_prior = 0.0,\n",
    "                \n",
    "                pointCloudVar = pointCloudVar, \n",
    "                rocs_pointCloudVar = rocs_pointCloudVar,\n",
    "            )\n",
    "            \n",
    "            pred_rocs = get_ROCS(torch.tensor(np.array(updated_mol.GetConformer().GetPositions())), torch.tensor(np.array(mol.GetConformer().GetPositions())))\n",
    "            \n",
    "            tanimoto = rdkit.DataStructs.FingerprintSimilarity(*[rdkit.Chem.RDKFingerprint(x) for x in [mol, updated_mol]])\n",
    "                \n",
    "            reference_mols.append(mol)\n",
    "            unaligned_mols.append(updated_mol)\n",
    "            \n",
    "            repeated_rocs.append(pred_rocs.item())\n",
    "            repeated_tanimoto.append(tanimoto)\n",
    "            \n",
    "        except Exception as e:\n",
    "            print(f'error during 3D generation -- {e}')\n",
    "            continue\n",
    "    \n",
    "    print(f'Encoded Molecule {file_idx + 1}:')\n",
    "    print('(non-aligned) shape scores:', [np.round(r, 3) for r in repeated_rocs])\n",
    "    print('tanimoto chemical similarity:', [np.round(r, 3) for r in repeated_tanimoto])\n",
    "    print()\n",
    "    file_idx += 1\n",
    "    \n",
    "    reference_mols_list.append(reference_mols[0])\n",
    "    unaligned_mols_list.append(unaligned_mols)\n",
    "    \n",
    "    if file_idx == total_evaluations:\n",
    "        break\n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7dc8ec80",
   "metadata": {},
   "source": [
    "If you have access to an OpenEye license (placed in the parent directory), we can now use ROCS to compute the (aligned) shape similarity between the generated molecules and the target molecular shape."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "751cc638",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Note that we can safely ignore these aromaticity warnings.\n",
    "\n",
    "from utils.openeye_utils import *\n",
    "\n",
    "all_shape_similarity = []\n",
    "for r, ref in enumerate(reference_mols_list):\n",
    "    ROCS_out = ROCS_shape_overlap(unaligned_mols_list[r], ref)\n",
    "    aligned_shape_similarity = [ROCS_out[i][1] for i in range(len(ROCS_out))]\n",
    "    all_shape_similarity.append(aligned_shape_similarity)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "800ac93b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# selecting the best generated molecules (by shape similarity to their respective targets) \n",
    "\n",
    "best_generated_indices = [np.argmax(s) for s in all_shape_similarity]\n",
    "best_generated_mols = [unaligned_mols_list[i][best_generated_indices[i]] for i in range(len(best_generated_indices))]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f5ee8f92",
   "metadata": {},
   "source": [
    "We now can compare the encoded molecule (with the target shape) to the SQUID-generated molecules.\n",
    "\n",
    "Although we will visualize the molecules in 2D, note that we encode/decode molecules in 3D.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "c97e7f87",
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.core.display import Image, display\n",
    "from rdkit.Chem import Draw\n",
    "\n",
    "def get_2D_mol(mol):\n",
    "    mol_2D = deepcopy(mol)\n",
    "    rdkit.Chem.rdDepictor.Compute2DCoords(mol_2D)\n",
    "    return mol_2D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "0f217c19",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.8706264928476455\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.8426096427387321\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.8269169602590691\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.784029297579313\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.7881463164790653\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.8460011299850846\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.7455217055009342\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.8530059142417641\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.9316405575384616\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "encoded, decoded\n",
      "aligned shape similarity: 0.8097806553751086\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Displaying most shape-similar molecules\n",
    "\n",
    "for i in range(len(best_generated_mols)):\n",
    "    print('encoded, decoded')\n",
    "    print('aligned shape similarity:', all_shape_similarity[i][best_generated_indices[i]])\n",
    "    \n",
    "    display(Draw.MolsToGridImage([get_2D_mol(reference_mols_list[i]), get_2D_mol(best_generated_mols[i])]))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "37c62587",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.9.7"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}