examples/dataset.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Working with the Kinodata-3D dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If you plan to use Kinodata-3D within any projects that involve geometric deep learning based on [PyTorch Geometric](https://pytorch-geometric.readthedocs.io/en/latest/) you can re-use [our PyG dataset](#pyg-dataset).\n",
    "\n",
    "For all other cases, we provide a [framework-agnostic dataset](#framework-agnostic-dataset)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Framework-agnostic dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/joschka/mambaforge/envs/kinodata/lib/python3.10/site-packages/tqdm/auto.py:22: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading raw data from /Users/joschka/projects/kinodata-docked-rescore/data/raw...\n",
      "Reading data frame from /Users/joschka/projects/kinodata-docked-rescore/data/raw/kinodata_docked_with_rmsd.sdf.gz...\n",
      "Deduping data frame (current size: 124761)...\n",
      "122747 complexes remain after deduplication.\n",
      "Checking for missing pocket mol2 files...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2885/2885 [00:00<00:00, 59948.91it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Adding pocket sequences from cached file /Users/joschka/projects/kinodata-docked-rescore/data/raw/pocket_sequences.csv.\n",
      "Converting to data list...\n",
      "Done!\n"
     ]
    }
   ],
   "source": [
    "from kinodata.data import KinodataDockedAgnostic\n",
    "from kinodata.data.data_module import make_data_module\n",
    "\n",
    "dataset = KinodataDockedAgnostic()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "ComplexInformation(kinodata_ident='32336', compound_smiles='Nc1ncnc2c1c(-c1cccc(Oc3ccccc3)c1)cn2C1CCCC1', molecule=<rdkit.Chem.rdchem.Mol object at 0x14fc9a730>, activity_value=5.468521, activity_type='pIC50', pocket_mol2_file=PosixPath('/Users/joschka/projects/kinodata-docked-rescore/data/raw/mol2/pocket/5553_pocket.mol2'), docking_score=-10.307055, posit_probability=0.24, klifs_structure_id=5553, pocket_sequence='DVIGEG__GQVLKAAIKRM____ELEVLCKLGPNIINLLGAYLAIEYAPHGNLLDFLRKYLSQKQFIHRNLAARNILVIADFGLS', predicted_rmsd=4.790116, remove_hydrogen=True)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "next(iter(dataset))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## PyG Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from kinodata.data.dataset import KinodataDocked"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create/load the dataset\n",
    "Instantiating the dataset for the first time will trigger the data preprocessing, which creates the PyG data objects.\n",
    "This process takes a while (~ 1 hour on a 2021 MacBook Pro), but the result is cached on disk to accelerate future instantiations.\n",
    "\n",
    "**Note**: you may also obtain our [preprocessed version of the dataset](https://zenodo.org/records/10886085)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "KinodataDocked(104836)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset = KinodataDocked()\n",
    "dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data representation and data loading\n",
    "Docked complexes are stored/represented as [heterogenous graph data](https://pytorch-geometric.readthedocs.io/en/latest/generated/torch_geometric.data.HeteroData.html?highlight=heterodata) objects."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "HeteroData(\n",
      "  kissim_fp=[1, 85, 12],\n",
      "  y=[1],\n",
      "  docking_score=[1],\n",
      "  posit_prob=[1],\n",
      "  predicted_rmsd=[1],\n",
      "  pocket_sequence='KSLGRGAFGKVVQVAVKMLALMTELKILTHIGLNVVNLLGAMVIVEYCKYGNLSNYLKSFLSSRKCIHRDLAARNILLICDFGLA',\n",
      "  scaffold='C1CCC(CC2CCCC2)CC1',\n",
      "  activity_type='pIC50',\n",
      "  ident='53974',\n",
      "  smiles='Cc1cnc(Nc2ccccn2)s1',\n",
      "  \u001b[1mligand\u001b[0m={\n",
      "    z=[13],\n",
      "    x=[13, 12],\n",
      "    pos=[13, 3]\n",
      "  },\n",
      "  \u001b[1mpocket\u001b[0m={\n",
      "    z=[652],\n",
      "    x=[652, 12],\n",
      "    pos=[652, 3]\n",
      "  },\n",
      "  \u001b[1mpocket_residue\u001b[0m={ x=[85, 23] },\n",
      "  \u001b[1m(ligand, bond, ligand)\u001b[0m={\n",
      "    edge_index=[2, 28],\n",
      "    edge_attr=[28, 4]\n",
      "  },\n",
      "  \u001b[1m(pocket, bond, pocket)\u001b[0m={\n",
      "    edge_index=[2, 1306],\n",
      "    edge_attr=[1306, 4]\n",
      "  }\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "data = dataset[42]\n",
    "print(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Node types: ligand, pocket, pocket_residue\n",
      "Edge types: ('ligand', 'bond', 'ligand'), ('pocket', 'bond', 'pocket')\n",
      "Number of ligand heavy atoms: 13\n",
      "Number of pocket heavy atoms: 652\n",
      "Position of ligand atom indexed 0: tensor([ 1.3422, 21.6702, 38.2683])\n",
      "Position of pocket atom indexed 0: tensor([ 9.7198, 18.2420, 49.8679])\n"
     ]
    }
   ],
   "source": [
    "node_types, edge_types = data.metadata()\n",
    "print(f\"Node types: {', '.join([nt for nt in node_types])}\")\n",
    "print(f\"Edge types: {', '.join([str(et) for et in edge_types])}\")\n",
    "print(f\"Number of ligand heavy atoms: {data['ligand'].x.size(0)}\")\n",
    "print(f\"Number of pocket heavy atoms: {data['pocket'].x.size(0)}\")\n",
    "print(f\"Position of ligand atom indexed 0: {data['ligand'].pos[0]}\")\n",
    "print(f\"Position of pocket atom indexed 0: {data['pocket'].pos[0]}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Typically you'll want split data into train, test (and validation) sets.\n",
    "You can choose to reuse our code for creating a data module that hosts datasets and loaders for train, test and validation splits.\n",
    "This is detailed in [examples/data_splits.ipynb](data_splits.ipynb)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Complex graph representation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "import kinodata.transform as T"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If you want to use the complex graph representation from the Kinodata-3D publication use"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = KinodataDocked(\n",
    "    transform=T.TransformToComplexGraph(remove_heterogeneous_representation=True)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Node types: pocket_residue, complex\n",
      "Edge types: ('complex', 'bond', 'complex')\n",
      "Number of complex heavy atoms: 665\n"
     ]
    }
   ],
   "source": [
    "data = dataset[42]\n",
    "\n",
    "node_types, edge_types = data.metadata()\n",
    "print(f\"Node types: {', '.join([nt for nt in node_types])}\")\n",
    "print(f\"Edge types: {', '.join([str(et) for et in edge_types])}\")\n",
    "print(f\"Number of complex heavy atoms: {data['complex'].x.size(0)}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When creating a data module, holding different (i.e. train, validation, and test) datasets, the transform is passed down to the datasets via the `train_kwargs`, `val_kwargs` and `test_kwargs` keyword arguments;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# this could take a while, expect ~60 seconds\n",
    "from kinodata.data.grouped_split import KinodataKFoldSplit\n",
    "\n",
    "demo_split = KinodataKFoldSplit(\"random-k-fold\", k=5).split(dataset)[0]\n",
    "data_module = make_data_module(\n",
    "    split=demo_split,\n",
    "   batch_size=32,\n",
    "   dataset_cls=KinodataDocked,\n",
    "   num_workers=0,\n",
    "   train_kwargs=dict(\n",
    "       transform=T.TransformToComplexGraph(),\n",
    "   ),\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "HeteroDataBatch(\n",
       "  kissim_fp=[32, 85, 12],\n",
       "  y=[32],\n",
       "  docking_score=[32],\n",
       "  posit_prob=[32],\n",
       "  predicted_rmsd=[32],\n",
       "  pocket_sequence=[32],\n",
       "  scaffold=[32],\n",
       "  activity_type=[32],\n",
       "  ident=[32],\n",
       "  smiles=[32],\n",
       "  \u001b[1mligand\u001b[0m={\n",
       "    z=[931],\n",
       "    x=[931, 12],\n",
       "    pos=[931, 3],\n",
       "    batch=[931],\n",
       "    ptr=[33]\n",
       "  },\n",
       "  \u001b[1mpocket\u001b[0m={\n",
       "    z=[21094],\n",
       "    x=[21094, 12],\n",
       "    pos=[21094, 3],\n",
       "    batch=[21094],\n",
       "    ptr=[33]\n",
       "  },\n",
       "  \u001b[1mpocket_residue\u001b[0m={\n",
       "    x=[2720, 23],\n",
       "    batch=[2720],\n",
       "    ptr=[33]\n",
       "  },\n",
       "  \u001b[1mcomplex\u001b[0m={\n",
       "    x=[22025, 12],\n",
       "    z=[22025],\n",
       "    pos=[22025, 3],\n",
       "    batch=[22025],\n",
       "    ptr=[33]\n",
       "  },\n",
       "  \u001b[1m(ligand, bond, ligand)\u001b[0m={\n",
       "    edge_index=[2, 2066],\n",
       "    edge_attr=[2066, 4]\n",
       "  },\n",
       "  \u001b[1m(pocket, bond, pocket)\u001b[0m={\n",
       "    edge_index=[2, 42460],\n",
       "    edge_attr=[42460, 4]\n",
       "  },\n",
       "  \u001b[1m(complex, bond, complex)\u001b[0m={\n",
       "    edge_index=[2, 44526],\n",
       "    edge_attr=[44526, 4]\n",
       "  }\n",
       ")"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "demo_batch = next(iter(data_module.train_dataloader()))\n",
    "demo_batch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Relative structural information\n",
    "\n",
    "Note that the structural information is encoded as absolute 3D coordinates in the above data sets.\n",
    "\n",
    "Our transformer adds distance-based edges on the fly, which is implemented in a separated PyTorch `nn.Module`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "from kinodata.model.complex_transformer import StructuralInteractions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "interaction_module = StructuralInteractions(\n",
    "    32,\n",
    "    interaction_radius=5.0,\n",
    "    max_num_neighbors=16,\n",
    "    rbf_size=32\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAeoAAAIRCAYAAACF0P59AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAA9hAAAPYQGoP6dpAABnAElEQVR4nO3deXiTVd4+8PvJ2j3dN7qylQJFkLW4gbIquOsoIwO/UXRGERlBR8dXBV8FlwGdAfdXRUXFcQQXRARFNtmXQguFFmihLd2XdE+znN8faWJDS6Ft2jxp78919ZImJ3m+CZI75zznOUcSQggQERGRLClcXQARERFdHIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaSCZWrVoFSZIcfkJCQjBu3DisX7++3c/71ltvYdWqVe16bHZ2NiRJwj//+c92H/9SFi1aBEmSOu35W5KSkoKbbroJMTEx8PT0RGBgIJKTk7F69WqHdmazGcuXL8eUKVMQFRUFLy8vJCYm4qmnnkJFRUWX1kw9F4OaSGY++ugj7N69G7t27cJ7770HpVKJ6dOn4/vvv2/X83UkqLvCAw88gN27d3fpMSsqKhAdHY0lS5Zgw4YN+OSTTxAXF4eZM2fixRdftLerq6vDokWLEBsbizfeeAMbNmzAnDlz8N577+Gqq65CXV1dl9ZNPZPK1QUQkaPBgwdjxIgR9t+nTJmCgIAAfPHFF5g+fboLK+scUVFRiIqK6tJjjhs3DuPGjXO4bdq0acjKysJ7772H//mf/wEAeHp6IisrC0FBQQ6PjYmJwV133YWvv/4a9913X1eWTj0Qe9REMufh4QGNRgO1Wu1w++LFizF69GgEBgbCz88PV155JT744AM03WcnLi4Ox44dw7Zt2+zD6XFxcfb7KyoqsGDBAvTu3RtarRahoaG48cYbceLEiWZ1LF++HPHx8fDx8UFycjL27Nlzydpra2uxcOFCxMfHw8PDA4GBgRgxYgS++OILe5sLh75bOgVg+2karkIIvPXWWxg6dCg8PT0REBCAO++8E2fOnLmct7VFwcHBUKl+778olUqHkLYZNWoUACAnJ6fdxyK6XOxRE8mM2WyGyWSCEAKFhYV47bXXUFNTgxkzZji0y87OxkMPPYSYmBgAwJ49e/Doo48iLy8Pzz33HABg3bp1uPPOO6HT6fDWW28BALRaLQCgqqoKV199NbKzs/H3v/8do0ePRnV1NbZv3478/HwMGDDAfqw333wTAwYMwBtvvAEAePbZZ3HjjTciKysLOp3uoq/l8ccfx6effooXX3wRw4YNQ01NDdLS0lBaWnrRx9x0003NhsJ3796Nxx9/HIMGDbLf9tBDD2HVqlWYN28eXnnlFZSVleGFF17A2LFjceTIEYSFhV3qrYbFYoHFYkF5eTm++uor/PTTT1i5cuUlH7dlyxYAcKiHqNMIIpKFjz76SABo9qPVasVbb73V6mPNZrMwGo3ihRdeEEFBQcJisdjvGzRokLjuuuuaPeaFF14QAMTmzZsv+rxZWVkCgEhKShImk8l++759+wQA8cUXX7Ra1+DBg8Wtt97aapvnn39etPZRdOLECREUFCTGjx8vDAaDEEKI3bt3CwBi2bJlDm1zcnKEp6enePLJJ1s9ps1DDz1kf581Gs0l32chhMjNzRVhYWFixIgRwmw2X9ZxiDqCPWoimfnkk0+QmJgIACgpKcG6devwyCOPwGw2Y+7cufZ2W7ZswZIlS7B//35UVlY6PEdRUdEle5Q//vgj+vfvjwkTJlyypptuuglKpdL++5AhQwAAZ8+ebfVxo0aNwmeffYannnoKU6ZMwejRo+Hp6XnJ49kUFBRgypQpiIiIwLp166DRaAAA69evhyRJuO+++2Aymeztw8PDccUVV2Dr1q2X9fz/+Mc/8MADD6CoqAjff/895s6di5qaGixcuLDF9mVlZbjxxhshhMCXX34JhYJnD6nzMaiJZCYxMbHZZLKzZ8/iySefxH333Qd/f3/s27cPkyZNwrhx4/D+++8jKioKGo0G33zzDV566aXLmo1cXFxsHza/lAvP09qGzy91nH//+9+IiorCl19+iVdeeQUeHh6YPHkyXnvtNfTr16/Vx1ZVVeHGG2+E0WjEjz/+6DDEXlhYCCHERb+M9O7d+3JeFmJiYuzvwY033ggAePrppzFr1iyEhIQ4tC0vL8fEiRORl5eHLVu2XPYxiDqKQU3kBoYMGYKffvoJGRkZGDVqFNasWQO1Wo3169fDw8PD3u6bb7657OcMCQlBbm5uJ1T7O29vbyxevBiLFy9GYWEhfvzxRzz11FOYPn16ixPWbIxGI+644w6cPn0aO3bsaDYrPDg4GJIkYceOHfYvDU21dNvlGDVqFN555x2cOXPGIajLy8sxYcIEZGVl4ZdffrGPKBB1BY7bELmBlJQUALCHhyRJUKlUDsPRdXV1+PTTT5s9VqvVttjznTp1KjIyMuwTozpbWFgYZs+ejXvvvRcnT55EbW3tRdvef//92Lp1K9auXdtiKE6bNg1CCOTl5WHEiBHNfpKSktpV46+//gqFQuHQW7aF9JkzZ7Bp0yYMGzasXc9N1F7sURPJTFpamv28a2lpKdauXYvNmzfjtttuQ3x8PADrOePly5djxowZePDBB1FaWop//vOfLfYkk5KSsGbNGnz55Zfo3bs3PDw8kJSUhPnz5+PLL7/ELbfcgqeeegqjRo1CXV0dtm3bhmnTpmH8+PEdfi2jR4/GtGnTMGTIEAQEBCA9PR2ffvopkpOT4eXl1eJjXnvtNXz66ad49NFH4e3t7XAZmJ+fHwYOHIirrroKDz74IP7f//t/OHDgAK699lp4e3sjPz8fO3fuRFJSEv76179etK4HH3wQfn5+GDVqFMLCwlBSUoKvvvoKX375JZ544gn7F6K6ujpMnjwZhw8fxhtvvAGTyeRQT0hICPr06dPh94moVS6ezEZEjVqa9a3T6cTQoUPF8uXLRX19vUP7Dz/8UCQkJAitVit69+4tli5dKj744AMBQGRlZdnbZWdni0mTJglfX18BQMTGxtrvKy8vF4899piIiYkRarVahIaGiptuukmcOHFCCPH7rO/XXnutWb0AxPPPP9/qa3rqqafEiBEjREBAgL3Ov/3tb6KkpMTe5sJZ37NmzWpx9juAZrPXP/zwQzF69Gjh7e0tPD09RZ8+fcSf/vQnceDAgVbr+vDDD8U111wjgoODhUqlEv7+/uK6664Tn376qUM72+u/2M+sWbNaPQ6RM0hCNFkdgYiIiGSF56iJiIhkjEFNREQkYwxqIiIiGWNQExERyRiDmoiISMYY1E4khEBlZSU4kZ6IiJyFQe1EVVVV0Ol0qKqqcnUpRETUTTCoiYiIZIxBTUREJGMMaiIiIhljUBMREckYg5qIiEjGGNREREQyxqAmIiKSMQY1ERGRjDGoiYiIZIxBTUREJGMMaiIiIhljUBMREckYg5qIiEjGGNREREQyxqAmIiKSMQY1ERGRjDGoiYiIZIxBTUREJGMMaiIiIhljUBMREckYg5qIiEjGGNREREQyxqAmIiKSMQY1ERGRjDGoiYiIZIxBTUREJGMMaiIiIhljUBMREckYg5qIiEjGGNREREQyxqAmIiKSMQY1ERGRjDGoiYiIZIxBTUREJGMMaiIiIhljUBMREckYg5qIiEjGGNREREQyxqAmIiKSMZWrC6DmTCYT0tPT7b8nJiZCpeJfFRFRT8RPfxlKT0/Hsq+2IiQqDsW52VhwF5CUlOTqsoiIyAUY1DIVEhWHyPgEV5dBREQuxnPUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxlwa1EuXLsXIkSPh6+uL0NBQ3HrrrTh58qRDGyEEFi1ahMjISHh6emLcuHE4duyYQxuDwYBHH30UwcHB8Pb2xs0334zc3FyHNuXl5Zg5cyZ0Oh10Oh1mzpyJiooKhzbnzp3D9OnT4e3tjeDgYMybNw8NDQ2d8tqJiIguh0uDetu2bXjkkUewZ88ebN68GSaTCZMmTUJNTY29zauvvorly5dj5cqV2L9/P8LDwzFx4kRUVVXZ28yfPx/r1q3DmjVrsHPnTlRXV2PatGkwm832NjNmzEBKSgo2btyIjRs3IiUlBTNnzrTfbzabcdNNN6GmpgY7d+7EmjVr8PXXX2PBggVd82YQERG1RMhIUVGRACC2bdsmhBDCYrGI8PBw8fLLL9vb1NfXC51OJ9555x0hhBAVFRVCrVaLNWvW2Nvk5eUJhUIhNm7cKIQQ4vjx4wKA2LNnj73N7t27BQBx4sQJIYQQGzZsEAqFQuTl5dnbfPHFF0Kr1Qq9Xt9ivfX19UKv19t/cnJyBICLtr9cR48eFQvf/U4s33RSLHz3O3H06NEOPR8REbkvWZ2j1uv1AIDAwEAAQFZWFgoKCjBp0iR7G61Wi+uuuw67du0CABw8eBBGo9GhTWRkJAYPHmxvs3v3buh0OowePdreZsyYMdDpdA5tBg8ejMjISHubyZMnw2Aw4ODBgy3Wu3TpUvtQuk6nQ3R0tDPeBiIiIjvZBLUQAo8//jiuvvpqDB48GABQUFAAAAgLC3NoGxYWZr+voKAAGo0GAQEBrbYJDQ1tdszQ0FCHNhceJyAgABqNxt7mQk8//TT0er39Jycnp60vm4iIqFUqVxdgM3fuXBw9ehQ7d+5sdp8kSQ6/CyGa3XahC9u01L49bZrSarXQarWt1kFERNQRsuhRP/roo/juu+/w66+/Iioqyn57eHg4ADTr0RYVFdl7v+Hh4WhoaEB5eXmrbQoLC5sdt7i42KHNhccpLy+H0Whs1tMmIiLqKi4NaiEE5s6di7Vr12LLli2Ij493uD8+Ph7h4eHYvHmz/baGhgZs27YNY8eOBQAMHz4carXaoU1+fj7S0tLsbZKTk6HX67Fv3z57m71790Kv1zu0SUtLQ35+vr3Npk2boNVqMXz4cOe/eCIiosvg0qHvRx55BJ9//jm+/fZb+Pr62nu0Op0Onp6ekCQJ8+fPx5IlS9CvXz/069cPS5YsgZeXF2bMmGFve//992PBggUICgpCYGAgFi5ciKSkJEyYMAEAkJiYiClTpmDOnDl49913AQAPPvggpk2bhoSEBADApEmTMHDgQMycOROvvfYaysrKsHDhQsyZMwd+fn4ueHeIiIhcHNRvv/02AGDcuHEOt3/00UeYPXs2AODJJ59EXV0dHn74YZSXl2P06NHYtGkTfH197e1ff/11qFQq3H333airq8MNN9yAVatWQalU2tt89tlnmDdvnn12+M0334yVK1fa71cqlfjhhx/w8MMP46qrroKnpydmzJiBf/7zn5306omIiC5NEkIIVxfRXVRWVkKn00Gv13eoF56amopPdmcjMj4B57NO4k/JcUhKSnJipURE5C5kMZmMiIiIWsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxlSuLoDI3ZlMJqSnp9t/T0xMhErFf1pE5Bz8NCHqoPT0dCz7aitCouJQnJuNBXcBSUlJri6LiLoJBjWRE4RExSEyPsHVZRBRN8Rz1ERERDLGoCYiIpIxBjUREZGMMaiJiIhkjEFNREQkYwxqIiIiGWNQExERyRiDmoiISMYY1ERERDLGoCYiIpIxBjUREZGMMaiJiIhkjEFNREQkYwxqIiIiGWNQExERyRiDmoiISMYY1ERERDLGoCYiIpIxBjUREZGMMaiJiIhkjEFNREQkYwxqIiIiGXNpUG/fvh3Tp09HZGQkJEnCN99843D/7NmzIUmSw8+YMWMc2hgMBjz66KMIDg6Gt7c3br75ZuTm5jq0KS8vx8yZM6HT6aDT6TBz5kxUVFQ4tDl37hymT58Ob29vBAcHY968eWhoaOiMl01ERHTZXBrUNTU1uOKKK7By5cqLtpkyZQry8/PtPxs2bHC4f/78+Vi3bh3WrFmDnTt3orq6GtOmTYPZbLa3mTFjBlJSUrBx40Zs3LgRKSkpmDlzpv1+s9mMm266CTU1Ndi5cyfWrFmDr7/+GgsWLHD+iyYiImoDlSsPPnXqVEydOrXVNlqtFuHh4S3ep9fr8cEHH+DTTz/FhAkTAACrV69GdHQ0fv75Z0yePBnp6enYuHEj9uzZg9GjRwMA3n//fSQnJ+PkyZNISEjApk2bcPz4ceTk5CAyMhIAsGzZMsyePRsvvfQS/Pz8Wjy+wWCAwWCw/15ZWdnm94CIiKg1sj9HvXXrVoSGhqJ///6YM2cOioqK7PcdPHgQRqMRkyZNst8WGRmJwYMHY9euXQCA3bt3Q6fT2UMaAMaMGQOdTufQZvDgwfaQBoDJkyfDYDDg4MGDF61t6dKl9uF0nU6H6Ohop71uIiIiQOZBPXXqVHz22WfYsmULli1bhv379+P666+392ILCgqg0WgQEBDg8LiwsDAUFBTY24SGhjZ77tDQUIc2YWFhDvcHBARAo9HY27Tk6aefhl6vt//k5OR06PUSERFdyKVD35fyhz/8wf7nwYMHY8SIEYiNjcUPP/yA22+//aKPE0JAkiT7703/3JE2F9JqtdBqtZd8HURERO0l6x71hSIiIhAbG4vMzEwAQHh4OBoaGlBeXu7QrqioyN5DDg8PR2FhYbPnKi4udmhzYc+5vLwcRqOxWU+biIioK7lVUJeWliInJwcREREAgOHDh0OtVmPz5s32Nvn5+UhLS8PYsWMBAMnJydDr9di3b5+9zd69e6HX6x3apKWlIT8/395m06ZN0Gq1GD58eFe8NCIioha5dOi7uroap06dsv+elZWFlJQUBAYGIjAwEIsWLcIdd9yBiIgIZGdn4x//+AeCg4Nx2223AQB0Oh3uv/9+LFiwAEFBQQgMDMTChQuRlJRknwWemJiIKVOmYM6cOXj33XcBAA8++CCmTZuGhIQEAMCkSZMwcOBAzJw5E6+99hrKysqwcOFCzJkz56IzvomIiLqCS4P6wIEDGD9+vP33xx9/HAAwa9YsvP3220hNTcUnn3yCiooKREREYPz48fjyyy/h6+trf8zrr78OlUqFu+++G3V1dbjhhhuwatUqKJVKe5vPPvsM8+bNs88Ov/nmmx2u3VYqlfjhhx/w8MMP46qrroKnpydmzJiBf/7zn539FhAREbXKpUE9btw4CCEuev9PP/10yefw8PDAihUrsGLFiou2CQwMxOrVq1t9npiYGKxfv/6SxyMiIupKbnWOmoiIqKdhUBMREckYg5qIiEjGGNREREQyxqAmIiKSMQY1ERGRjDGoiYiIZIxBTUREJGMMaiIiIhljUBMREckYg5qIiEjGXLrWNxF1DyaTCenp6Q63JSYmQqXiRwxRR/FfERF1WHp6OpZ9tRUhUXEAgOLcbCy4C0hKSnJtYUTdAIOaiJwiJCoOkfEJri6DqNvhOWoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYy1K6h79+6N0tLSZrdXVFSgd+/eHS6KiIiIrNoV1NnZ2TCbzc1uNxgMyMvL63BRREREZKVqS+PvvvvO/ueffvoJOp3O/rvZbMYvv/yCuLg4pxVHRETU07UpqG+99VYAgCRJmDVrlsN9arUacXFxWLZsmdOKIyIi6unaFNQWiwUAEB8fj/379yM4OLhTiiIiIiKrNgW1TVZWlrPrICIioha0K6gB4JdffsEvv/yCoqIie0/b5sMPP+xwYURERNTOoF68eDFeeOEFjBgxAhEREZAkydl1EREREdoZ1O+88w5WrVqFmTNnOrseIiIiaqJd11E3NDRg7Nixzq6FiIiILtCuoH7ggQfw+eefO7sWIiIiukC7hr7r6+vx3nvv4eeff8aQIUOgVqsd7l++fLlTiiMiIurp2hXUR48exdChQwEAaWlpDvdxYhkREZHztCuof/31V2fXQURERC3gNpdEREQy1q4e9fjx41sd4t6yZUu7CyIiIqLftSuobeenbYxGI1JSUpCWltZssw4iIiJqv3YF9euvv97i7YsWLUJ1dXWHCiIiIqLfOfUc9X333cd1vomIiJzIqUG9e/dueHh4OPMpiYiIerR2DX3ffvvtDr8LIZCfn48DBw7g2WefdUphRERE1M6g1ul0Dr8rFAokJCTghRdewKRJk5xSGBEREbUzqD/66CNn10FEREQtaFdQ2xw8eBDp6emQJAkDBw7EsGHDnFUXERERoZ1BXVRUhHvuuQdbt26Fv78/hBDQ6/UYP3481qxZg5CQEGfXSURE1CO1a9b3o48+isrKShw7dgxlZWUoLy9HWloaKisrMW/ePGfXSERE1GO1q0e9ceNG/Pzzz0hMTLTfNnDgQLz55pucTEZERORE7epRWyyWZntQA4BarYbFYulwUURERGTVrqC+/vrr8dhjj+H8+fP22/Ly8vC3v/0NN9xwg9OKIyIi6unaFdQrV65EVVUV4uLi0KdPH/Tt2xfx8fGoqqrCihUrnF0jERFRj9Wuc9TR0dE4dOgQNm/ejBMnTkAIgYEDB2LChAnOro+I3JDFbEZGRob998TERKhUHboalKjHatO/nC1btmDu3LnYs2cP/Pz8MHHiREycOBEAoNfrMWjQILzzzju45pprOqVYInIPpQU5WH2iAn1LNSjOzcaCu4CkpCRXl0Xklto09P3GG29gzpw58PPza3afTqfDQw89hOXLlzutOCJyX4ER0YiMT0BIVJyrSyFya20K6iNHjmDKlCkXvX/SpEk4ePBgh4siIiIiqzYFdWFhYYuXZdmoVCoUFxd3uCgiIiKyalNQ9+rVC6mpqRe9/+jRo4iIiOhwUURERGTVpqC+8cYb8dxzz6G+vr7ZfXV1dXj++ecxbdo0pxVHRETU07Vp1vf//M//YO3atejfvz/mzp2LhIQESJKE9PR0vPnmmzCbzXjmmWc6q1YiIqIep01BHRYWhl27duGvf/0rnn76aQghAACSJGHy5Ml46623EBYW1imFEhER9URtXoEgNjYWGzZsQHl5OU6dOgUhBPr164eAgIDOqI+IiKhHa/dSQQEBARg5cqQzayEiIqILtGutbyIiIuoaDGoiIiIZY1ATERHJGIOaiIhIxhjUREREMsagJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGIOaiIhIxlwa1Nu3b8f06dMRGRkJSZLwzTffONwvhMCiRYsQGRkJT09PjBs3DseOHXNoYzAY8OijjyI4OBje3t64+eabkZub69CmvLwcM2fOhE6ng06nw8yZM1FRUeHQ5ty5c5g+fTq8vb0RHByMefPmoaGhoTNeNhER0WVzaVDX1NTgiiuuwMqVK1u8/9VXX8Xy5cuxcuVK7N+/H+Hh4Zg4cSKqqqrsbebPn49169ZhzZo12LlzJ6qrqzFt2jSYzWZ7mxkzZiAlJQUbN27Exo0bkZKSgpkzZ9rvN5vNuOmmm1BTU4OdO3dizZo1+Prrr7FgwYLOe/FERESXod3bXDrD1KlTMXXq1BbvE0LgjTfewDPPPIPbb78dAPDxxx8jLCwMn3/+OR566CHo9Xp88MEH+PTTTzFhwgQAwOrVqxEdHY2ff/4ZkydPRnp6OjZu3Ig9e/Zg9OjRAID3338fycnJOHnyJBISErBp0yYcP34cOTk5iIyMBAAsW7YMs2fPxksvvQQ/P78ueDeIuieL2YyMjAz774mJiVCpXPrRQ+RWZHuOOisrCwUFBZg0aZL9Nq1Wi+uuuw67du0CABw8eBBGo9GhTWRkJAYPHmxvs3v3buh0OntIA8CYMWOg0+kc2gwePNge0gAwefJkGAwGHDx48KI1GgwGVFZWOvwQkaPSghys3n4cn+zOxrKvtiI9Pd3VJRG5FdkGdUFBAQAgLCzM4fawsDD7fQUFBdBoNAgICGi1TWhoaLPnDw0NdWhz4XECAgKg0WjsbVqydOlS+3lvnU6H6OjoNr5Kop4hMCIakfEJCImKc3UpRG5HtkFtI0mSw+9CiGa3XejCNi21b0+bCz399NPQ6/X2n5ycnFbrIiIiaivZBnV4eDgANOvRFhUV2Xu/4eHhaGhoQHl5eattCgsLmz1/cXGxQ5sLj1NeXg6j0disp92UVquFn5+fww8REZEzyTao4+PjER4ejs2bN9tva2howLZt2zB27FgAwPDhw6FWqx3a5OfnIy0tzd4mOTkZer0e+/bts7fZu3cv9Hq9Q5u0tDTk5+fb22zatAlarRbDhw/v1NdJRETUGpdOvayursapU6fsv2dlZSElJQWBgYGIiYnB/PnzsWTJEvTr1w/9+vXDkiVL4OXlhRkzZgAAdDod7r//fixYsABBQUEIDAzEwoULkZSUZJ8FnpiYiClTpmDOnDl49913AQAPPvggpk2bhoSEBADApEmTMHDgQMycOROvvfYaysrKsHDhQsyZM4e9ZCIicimXBvWBAwcwfvx4+++PP/44AGDWrFlYtWoVnnzySdTV1eHhhx9GeXk5Ro8ejU2bNsHX19f+mNdffx0qlQp333036urqcMMNN2DVqlVQKpX2Np999hnmzZtnnx1+8803O1y7rVQq8cMPP+Dhhx/GVVddBU9PT8yYMQP//Oc/O/stICIiapVLg3rcuHEQQlz0fkmSsGjRIixatOiibTw8PLBixQqsWLHiom0CAwOxevXqVmuJiYnB+vXrL1kzERFRV5LtOWoiIiJiUBMREckag5qIiEjGuOAuEbWLyWSyLweakZEBYVG7uCKi7olBTUTtkp6ejmVfbUVIVBwyDh5AcHwCerm6KKJuiEPfRNRuIVFxiIxPQEB45KUbE1G7MKiJiIhkjEFNREQkYwxqIiIiGeNkMiKibqTpbHzAut+BSsWPenfGvz0iom6k6Wz84txsLLgLSEpKcnVZ1AEMaiKibsY2G5+6B56jJiIikjEGNRERkYwxqImIiGSMQU1ERCRjDGoiIiIZY1ATERHJGC/PIiLqpixmMzIyMuy/c/ET98S/MSKibqq0IAerT1Sgb6mGi5+4MQY1EVE3FhgRzcVP3BzPURMREckYg5qIiEjGGNREREQyxqAmIiKSMU4mIyKnq7RoYTZqXF0GUbfAoCYip0o7r8chYyRglJBYWe/qcojcHoe+ichpjuZW4Jf0IgASACAlp8Kl9RB1BwxqInKaQ+cqAAAhimoAQEZhFQxC6cKKiNwfg5qInKLeooC+zghJAhJUxdApjbAI4LzZz9WlEbk1BjUROUW5SQ0ACPXVQiUJRGvrAAD5Zl8I4crKiNwbg5qInKK8cZZ3VIAXACBEbYBCAhqgQr3gRw1Re/FfDxF1mBC/96ijAzwBAAoJCPS2hneViReYELUXg5qIOqweKtQLJRQSEOnvab891NcDAFBlZlATtReDmog6rNxiDecwPw+olb9/rIT4agEwqIk6gkFNRB1WZbEGcq8mvWnAOrEMAKoZ1ETtxqAmog6rFdbz00E+jsuGBvtoAQgYhBI1BpMLKiNyfwxqIuqwWmEN6EAvx6DWqBTwlIwAgOJqQ5fXRdQdMKiJqEPqjGYYYV19zN+r+UYcvlIDAKCoikFN1B4MaiLqkPIaaxBrJTM0quYfKT6SNaCLGdRE7cKgJqIOKa+1BrW30tzi/T4K6/1ljYFORG3DoCaiDimvsZ6D9lK0PFnMo/Ectb7OyKVEidqB10wQUYeUXaJH7QETJAiYLUB9y02IqBXsURNRh9jOUXspWk5hhQRoFRYAQLWxy8oi6jYY1ETUbmZhHdIGLt6jBn4P8WpeSk3UZhz6JqJ2qzYCAoASFmgky0XbeTYGdY1R6qLKejZ9A9BQUgOjUDReOEfujEFNRO1W1TiR20tqgNRKBtuCmkPfnauoqh7LfivB1lwFkHseQCzChQEDLZzF584Y1ETUbjWNQ9meUutj2p4c+u50+joj7nl3D86U1AIA/DxUqKw3ocDoga0ZxeinNCMjI8PePjExESoVI8Ad8G+JiNqt1mTtRmsvFdRK29B3p5fUI5ktAo+tOYwzJTUI9lLiCn8jBifG4+dtO3DMFIbUPD3qlHpknTyHvqUaFOdmY8FdQFJSkqtLp8vAyWRE1G61jfnsccketfX8dYNFQnXDxc9lU/u8s+00tp4shlalwDPXhSDQug04QpS16OdRAwDIMgfCNywGkfEJCImKc12x1GYMaiJqN1tQa9F6UKskAU+1dVpTQRXHv50pt7wWK7ZkAgBevHUw+gY6rrcera1DkLcGZiiQY/Bs6SlI5hjURE2YTCakpqbaf0wmhkprfu9RX3pM29/LuhVmPmeUOdXSDSdQb7RgdHwg7hwe1ex+SQJGxwcCAHIMnjAYueqMu2FQEzWRnp6OZV9txSe7s7Hsq61IT093dUmyVWe0oMFiPUd9qaFvANB5WoOaPWrn2XOmFD+k5kMhAYtuHgTpIlPv+4b6wFtqgAkKHMnVd3GV1FGcTEZ0gZCoOETGJ7i6DNkrqbX2zDRKBVTSpS//sQc1p363i8lkcvjiOGDAACzbdBIAcO+oGCRG+F30sZIkIVpZgROmUBzPr0Sv8E4vl5yIQU1E7VLUeG2Wr6cKuIzRbF8P68eNLeCpbWyjPSFRcSjOzcb1V9dhf3Y5NCoF5t3Q75KPD1HUIBMW6OuMKOWOo26FQU1E7VJcYw1cX+3lBrW68XHsUbeXbbRHCOCzo9Yh7PtGxyLMz+OSj1VKAqEaA/IbPJFdxRXi3AnPURNRu9gC1xbAl2LrURfVmCG432WHFNYBJ0sa4KFW4C/jel/24yI01q50TjVQb+Jlcu6CQU1E7WIb+vbzuLyBOV+ttZ3BLFBRy5nfHXGiwtojvndUDEJ9L92btvFXGqHzVMMkJOzJqeus8sjJGNRE1C7FjeeaL7dHrVIqoFVae9J5FQyJ9srX16G4XoJKATx47eX3pgHrpVr9w3wAALtyajujPOoEDGoiapffh74vf6qLV2PT8wzqdtufXQ4AGB/vjQhd2xcw6RtqDeqD5+tR28D5Au6AQU1EbWa2CJTae9RtD2r2qNtH3wBkldQAELhj4MUvx2pNiI8W3iqBBrPA1pPFzi2QOgWDmojarKiqHmYBSBDw1rJH3VVs56ajvIFefpd3yuFCkiQhytv65x/TCpxVGnUiXp5FRG1mC1pPFaBobSPqC3irBAAJ5yvqO6my7qugyoRz1dY/J/iZOrRlZS9vgZN6CVvSC1FvNMOjcR12kif2qImozXLLrUHt1cav+hz6br+16ZUAJMQGecFcloPV24+3e6nbQC0Q7KVETYMZOzJLOqdgchr2qImozWw94vYGNYe+26awsh6bT1u70yNjA1FyDAiMiEZkfAIsZrND7zojIwPC0vqwuCQBY6O98N3JKvyYmo+JA8M6tX7qGAY1EbWZLWjbG9RFVQYYTGZoVRxyvRzvbz8DkwUI9hDoFeCJpn3g0oIcrD5Rgb6l1u0tMw4eQHB8Anpd4jnHxnjiu5NV2JxeiAaTBRoVB1jlin8zRNRmtqFr6znny6dRABql9Zx2gZ7nqS9HWU0DPtt7DgCQ6N/y+23rXUfGJyAgPPKynndAsBYhvlpU1Zvw22kOf8sZg5qI2qy9PWpJAkK8rb1onqdunW1v9FfW7UWd0YwoLwtCtc5belWpkDB5kHXIe2MqZ3/LGYOaiNosr51BDQAhjQ/izO/Wpaen45X/bMPa49bNN5RFJ1GhL3PqMW4cHAEA2HS8ACYz1/6WKwY1EbVJZb0RVfXWFa3aFdS2HnU5e9SXUuETC6NFQqC3BnGBWqc//6j4QAR4qVFea8TeLOd+CSDnYVATtZFtSNL2Y51l23N2g7INe/tqFGjP/KMQb1uPmkHdmnqTBZl66/n8kbEBaMPl6pdNpVRg8qBwAMCPafnOPwA5BYOaqI3S09Ox7Kut+GR3tvVn0wGUV/Sc3ogtYG0947YK8bI+7ryeQd2aTaeqYbBI0Hmq0T/Mt9OOM2WwNag3phXC3IO+cLoTBjVRO4RExbV5lm13YRuytvWM2yq08XGcTHZxBpMZa49XAQCGxwZAoeiE7nSjsX2C4euhQkm1AQfPlnfacaj9GNRE1CZ5jZPA2t2jbnzc+Yo6CMEeXEvWHspDaZ0ZnkqBxIjO600DgEalsC940hnD3xeeKjKZuGNXWzGoiahN7EPf7ZlJBiC48XH1RgvKahqcVld3YTJb8PbW0wCA/v4CKkXnf0zbZn9/fyQfRifP/m56qqg9y50SVyYjojayDVmHeqtQrW/749VKCSG+WhRXGXC+oh5BPs6fzezOvk05j3NltdBpFejt2zWXTF2XEIJgHy1Kqg34Jb0QUxqDuz1MJpNDGGdkZCA4MhaR8QnOKLVHYo+aehwOxXVMRyeTAUCkvycAnqe+kMlswYotmQCAWxP92jWrvj3USgXuGhEFAPh8X06HnqunT7bsDOxRU49j+yAJiYpDcW42FtwFJCUltfl5agwm5JbXwSwk9JQVq41mCworbeeo2//x0cvfA0dyeInWhb4+mIPs0lr4aRVIUBTg0CU212ivCzfyAIA7h8Xi7a2nsSOzGDlltYgO9Lrs52vai76wB12cl+20unsqBjX1SLZZ2+1hsggcLZVwOjsbJotAkCIUV4haJ1coTwX6elgEoFEq4O/R/u5er8YeNYP6d0azBa9vOgEAiPM24ctfLm9zjfa4cCMP6xfWcbiqbxB+O1WKz/edw9+nDLjs52v65bfppiCVdUYUmH1grPdE9ZlSmCot2HI4E0IISJLU5n20eyq+Q0RttO54JU7qJQDWGculFm/kGATa3id3P7ZgjfD3gKIDK3Bw6Lu5dYfyUFBtglYpcPWQfjhek92px7Nt5NHUn5Lj8NupUnyyKxsPXB3fpvkDti+/RbnZKGjQInXfORRVGQCEAibgdFYZABUOlqrwfuY5hFrK8MqdFowcdoVzX1g3xHPURG2grzPi6/RKAMB1/UMwrn8IAOBUvTdKqw2uLK1L2BYpidR5duh5ItmjdmA0W7DiV+u56QSdgFrpmo/mSQPDMLiXH2oazHh3+5k2P76wsh4HjL1wrNYPRVUGSAD8pHpEqOsxONIP/lItFBCoNko4Yw7CX77Lx/qj553/QroZWQf1okWLIEmSw094eLj9fiEEFi1ahMjISHh6emLcuHE4duyYw3MYDAY8+uijCA4Ohre3N26++Wbk5uY6tCkvL8fMmTOh0+mg0+kwc+ZMVFRUdMVLJDfzwc4s1DQI+KkFhkTpMCRKhwCpFgISThZWubq8Tmdb7KRXQMeCupe9R82NOQDg64O5yCmrg7+HAn38XFeHJElYMMnay/54V7Z9PsKlmC0Cx8uB/xzIQY3QQiVZkNw7CA9cE48rNecx0LsKNySGYaimANfoSjAhMRReSoHSOjPmfn4Ys9/Zgl37DnKS50XIOqgBYNCgQcjPz7f/pKam2u979dVXsXz5cqxcuRL79+9HeHg4Jk6ciKqq3z8w58+fj3Xr1mHNmjXYuXMnqqurMW3aNJjNZnubGTNmICUlBRs3bsTGjRuRkpKCmTNndunrJPnT1xnx4c4sAMCgAAFF45fHMGU1ACCrpMaV5XUJW7DaesTtZXt8SbUB9UbzJVp3b/VGM1ZsOQUAuGNg1830vphx/UMwPDYABpMFC786csllRc8UV+PJTYU4Vq6ARQAhimok+5ZhVHwgvDTNz66qJGBQpA7DldkIN+YDENiaXYf7/3sG7+7I5vXWLZB9UKtUKoSHh9t/QkKsQ41CCLzxxht45plncPvtt2Pw4MH4+OOPUVtbi88//xwAoNfr8cEHH2DZsmWYMGEChg0bhtWrVyM1NRU///wzAOskiI0bN+L//u//kJycjOTkZLz//vtYv349Tp486bLXTfKzMS0f1QYTov1U6OX9++2BiloAAiXVDajt5p0A21B1L3+PDj1PgJcanmrrXPl8fc/uVa/ecxZ5FXUI9/PA1H4+ri4HkiRh6e1J8FQrsSOzBP/+JbPFdkIIfLI7Gzf+ewcyShugVghMHhSGgaoiaBSXXnFOKQkMClHhzuHRUMOEOoUXthZq4BEaj5CoOCe/Kvcm+6DOzMxEZGQk4uPjcc899+DMGet5k6ysLBQUFGDSpEn2tlqtFtdddx127doFADh48CCMRqNDm8jISAwePNjeZvfu3dDpdBg9erS9zZgxY6DT6extLsZgMKCystLhh7qv745Yz6WNi/d22MlII1mgU1oTOr+bT/7Oswf15V+60xJJkuzD57nl3fxNa0VlvRErf7X2pv82sR+0ru5ON+of5osltw8GAPzrl0ws/v6YfeRDCIFdp0tw65u/4blvj6HeaEE/PzMmRlowINyvzbt89fL3xJWa8/BSmFBtMOHrQ7mo4oJ1DmQ963v06NH45JNP0L9/fxQWFuLFF1/E2LFjcezYMRQUFAAAwsLCHB4TFhaGs2fPAgAKCgqg0WgQEBDQrI3t8QUFBQgNDW127NDQUHubi1m6dCkWL17c7tdH7qOoqh67T5cCAK6N9camNMfNC4LVBujNauTXdt7mCa4mhLD3qCP9PVDT+j+PS4oO8MSpomrklF3+hLILV70C4JaX+NhexycpFaioNSLKT4UETQUyMjIhOuna6ba6bVgUMgur8dbW0/jot2z892Au+oT4IKesFqWNS78qYMGQIMByajcMHVh5zFMyYbhPBY5bIlFa3YCt+RKGp/5+nbc7/h07k6xf+dSpU+1/TkpKQnJyMvr06YOPP/4YY8aMAWD9Zt6U7fq81lzYpqX2l/M8Tz/9NB5//HH775WVlYiOjm71MeSefjiaD4sAhkb7I9y3+T+bYHUDTtcDhXXWfYTd2YVhaPuQ1NcZUdtg7VVF+nsis4NBHdO4oMa5ssvvUTe9XhdAhxascaX09HQs+c8O7DPFAJCgLj6Jz/Y2OFyDLAdPThmAkfGBeOrroyisNCAlpwIA4KFWYHycF1SGSvTul4Aj1dkdPpZGIXD7Fb2sG5LUNGBZisBkfTb0+e75d+xMsg7qC3l7eyMpKQmZmZm49dZbAVh7xBERv69LW1RUZO9lh4eHo6GhAeXl5Q696qKiIowdO9beprCwsNmxiouLm/XWL6TVaqHVcp3insA27H3zFZEAqpvd760ww1urRI3BjNNlDRjZxfU508VWbsttnPEd7KOBh7rja7HZVr7KaePQd0cWq5GTQo8YmKsk+En1iA/1Q2R8gixX8RqfEIodT16P08XVOFNcgxBfLa6I1iEj/Tg+2e3c031eGhVuGRqJ1b+dQr3CEwcrPTG6V5xTj+GO5HFC5DIZDAakp6cjIiIC8fHxCA8Px+bNm+33NzQ0YNu2bfYQHj58ONRqtUOb/Px8pKWl2dskJydDr9dj37599jZ79+6FXq+3t6GeraiyHofPVQAApg1pebMCSQJCfa0TrE6XGbuqtE5jC8Omk3p+H/bu2IxvG1tQ57ahR91d5OiNyGq8OKW3qrTN53W7mkalQGKEH24aEoFR8YHQqjpv0VxfDzWGqAughAW55XVILZP5m9MFZB3UCxcuxLZt25CVlYW9e/fizjvvRGVlJWbNmgVJkjB//nwsWbIE69atQ1paGmbPng0vLy/MmDEDAKDT6XD//fdjwYIF+OWXX3D48GHcd999SEpKwoQJEwBYh/WmTJmCOXPmYM+ePdizZw/mzJmDadOmISHB/b+1U8dtzSgGAFwRpUOo38VnO4f4WkdXzpR3z5kw9qDu4GInNtEBbR/67g6EEPjgUDkEJMQHe8Nf0f0XymkrH0UDBnpZv8lk6CVsz+7+lz62RtZD37m5ubj33ntRUlKCkJAQjBkzBnv27EFsbCwA4Mknn0RdXR0efvhhlJeXY/To0di0aRN8fX/faP3111+HSqXC3Xffjbq6Otxwww1YtWoVlMrfvxF+9tlnmDdvnn12+M0334yVK1d27Ysl2dp6sggAMC6h+aTDpkK7eVDbZ3x3cLETm+hA6/OU1xpRVW+Er4c8JlF1tk3HC3HwfD0kCFzTLxjnDrm6ouabdLQ0eaul7Ss7c+JbqKYBI0ICcOBsOf69uwT9gw8j3EfVIyeWyfrVrlmzptX7JUnCokWLsGjRoou28fDwwIoVK7BixYqLtgkMDMTq1avbWyZ1YyaLwI6MEgDA+AGtB3VI47rI5yqMMJjMnTo86ArnnbTYiY2vhxoBXmqU1xqRU1aHgZHdP6jrGsx44fvjAIAEfyDAS4Nzri0JgOMmHReboHfhRL6umPiW3CcImefOQ2/xxFMb8zDIkoUn7u55E8tkPfRN5GrpxQZUGUwI8tZgSC9dq219PVTQKATMAsgsbD7hzN3lOmmxk6Zi2jmhzF29+esp5FXUIcRLiUT/Sy8K0pVsm3S0ttiIbe5CZHwCAsIjO70mhSRhgLoYSlhQapBQ6nPx2rozBjVRKw6et4bTdf1DoFC0PqlFkiT4N14EcOy8vrNL63K2db6jAjq22ElT9pnfPeA89ZniarzXuNHFnBEBLl8q9HKYTCb72tvWoe6u/3LhKZmQ4GX94nu8XEJGSc87py/roW8iVzuQZx3uvS4h5LLa+2uAojrg2PnutUpdXYMZJY27g0U56Rw10HOC2mg0YuHne9FgtmB4pAcCa3Nls7DJhZqer87IyMB3KXkIjY536TXe4WoDTKE+yCiqxrJdpZgy1gRvbc+JLzf4TkfkGrUm4KzeCIUEXNvv8oI6QGvtcXS3oM6rsAapr1YFnafzAqY9i564o7c2HMCh/HooIBCqrMWnmw+gvKLM1WW1qLQgB6u3H8cnu7PxyaYDUHn7ddlQ98VIknWOiKdS4HyVCa/91LP2YWBQE12Ebd3uYTEBCPDWXNZj/BubpedXwuKCYcLOktNke8tLrdjXFrZLtGzP3x3l6+vw3gHrkrOj+wSjf3/Xht7lsJ2vllOdHmolRoRY/019vDsbh86VX+IR3QeDmugiChrX7R5/mcPeAOCjBpQSUNtgRsFl7uXrDmyLkjjz/DTQZDJZWW23+mJjY7EIPPnfo6gxCgRqBUbEBFz6QXRRoVozRgSZIQQw/7P9OJRytEfsW82gJmqByWJBYWMnL0pZedkb2Ssk2NcCP1PcfRZpsC0farv22Vki/T2gVkowmCw4r+9+veq3t53GjswSaJQSRoWIS05IpNaVFuTAkJsGrULgnN6Ix7860iP2rWZQE7Ugr7wOZiFBJRqw88T5Nm1kH+VnPYd7pqT7XKKV2wkzvgFApVSgd7B1D+bMou7zfgHAnjOlWLbJei71LyMD4Ht5Z0/oEkIjIjE+0bqU71lzAM7p3X/J3kthUMtQrdEC0f1GAd1Kdql1qDdEa0av3q1fW3qhXt2wR227ztmZM75t+oZZg/pUN7r2PKesFo98dggWAdx+ZS9M6O3t6pK6lf5hPogL8oKAhJV7SrvlaZOmGNQyc/hcOeauz8fp7jVp2O2cLbWGbLCq7cuB9mrsUZ8u7j7BYx/6dnKPGgD6hdp61FVOf25XqKw34s+r9qO0pgGDIv3w4q2DnToBj6xrFowfEAqVJJBe0oDP9p51dUmdqudciOYmDp4tR3GtGSW1Evrru89kJHdSbbSuPy1BIFDd9mG1Xn7dq0ddZ7SgrMb6hcVZ63w31S/Uujb/xYa+m64x3dnrS3dUbYMJD6w6gMyiaoT5avHEaB+cPpku+7rdkZ+HGoP8LThSrsSSH46jl1SOa0ckdct1wLvfK3Jz918dj1+OZmN3Th02pOVjfOtbYlMnKGi8LEsn1UMltX1IzXaO+ry+DvVGs1P2bnalohrrJDo/D+deQ23Tr8nQtxCiWe+z6RrTrlx042JsXyQMJgte3FaClIJ6+GpVePrqAKz6Yads6+4O/GvOwtscihr44IlvM/GRlxJDhgxxdVlOx6FvmZEkCY+NCYK3SqCq3oQMPYfMulp+42VZgYr2LcLhp1VA56mGEEBWifv3qgurzQB+X0WsI2yrXtmWpTSZTIgL8oZSIaHKYEJhZcvLQ9rWmJbTdb026enpePk/2/HQNzlIKaiHh0rCqj+PQp9Ajazr7g4kCRjs3wCFBJQKb+w81z0XzmFQy5C3RoFBAdaeXEH3/P9OtupNFhQ1nnEIamdQS5KE3iHWyUPdYfjb1qN2xkSypqte2WbSa1QKxAVZvwS443nq85VGpCIWpQYJaoXAovEhGB7L66W7io/SjBFxgQCA9w6UQ1/b/WaBM6hlKqzxM7GiAdDXm11bTA+SVmiARUjw9VDBS2r/P3jbJUdnusGEssJqW1A7ZyJZS7s02c5TZ7jZzO+Nafn428YCVBoleGuVGB8pMCjUebuL0eUZGRcAX7VARb0FSzZ0v+uqGdQy5aECgn00ACQcKeCksq5yoHG3rLggb3Rkoq69R90Nhr7zG4M6uhMmktnYz1O7SY+6orYBf/syBX9ZfQi1RoEgrcA9I2Pgq/x9aN9Vu031RCqFAlcGWjs0Xx7IwWeb93erFcs4mUzGYgK9UFLdgBQGdZcQQmB/ni2ovVBV0v7n6mMf+navHmJL8qusH3hxwZ13LXDfxku05N6jNpkt+GJ/DpZvOonyWuuGLbcl+gJ1evhoVThdkIPVJyrQt1TDCWRdTKo4h2CTL0pUIXjxlzwkBGswYtgVri7LKdijljHbOsiH8+shuAJKpztdXIOiGjMUEB2eOBUbZA0128Ip7koIoKCxRx3fiUE9KNIPAJCWp0eDydJpx2kvo9mC/+zPwQ3Lt+HZb9JQXmtEjE6NVyaFYax3MaQm/z7luKFFTzEoyHoKog5qfJnWfRajYI9axnr5e0IhCZTUmnG6uMbe66DOsfVkEQAgxBNQKzv2HTa2cXKUvs6IitoG+Hu55/qRdWagwSygUkjo5d95Q999QnwQ4KVGea0RqXn6LpuM1fQabdtQadPrcHv3S8A3R/Lx1q+nkVdhHW3x0yoQZirCkMBA7MvMZ89ZRlSSwPiEUKw/mo+1xyvx/85XYmDjl0B3xqCWMZVSgSAtUFwPHMguY1B3sl8bgzrCq+OjF14aFUJ9tSiqMuBsaa3bBnV143y6mEAvqDr45aU1kiRhRFwgNh8vxP7ssi4LasdrtHdB4eGFvoOGwmwBjuaUoUKbh5Ja67nPYB8tHrq2N4b6VuE/+02IjE8AABTnZXdJrXR5+oT4oJeXQF6thKfXHsXah6+C0s03Q+HQt8wFaK3/Tc/vPsM4clRtMGFfVhkAINxJq2TG2Ye/3XdCWVVjUNtGCDrTqMZLbA5kl3X6sZpqeq1zQHg0StRh+Om8GqfMwSipNSPQU4k/D/PDysmBGBNQg3NnTnGSmMwNCxbwVks4kqvHql3Zri6nw9ijljl/jQAgIT3fPWbDuqtdp0pgNAtE+Kjgq277+t4tiQ3ywr7sMpx14/PU1UZrT6QzJ5LZjIy3BvX+7HJYLJe3JaRtARWbxMTEdi8hWWXRIKPaH1UniwEAWpgQ0nAeo+KicHDLBhzeZe1tc6hb/rSSGVMjzfjvWTVe25iOGxKCERfi6+qy2o1BLXP+TXrULS2vSM6xI9M6xfvKSA/A5JygtoWbO/eobUPfnTmRzGZQpB881Uro64zILKpGQvilP1hLm8yyLs7NxoK7gKSkpFYfc+Ha4Q0mNbaeLMIRYy8AEjRKBZL7BEFk74Xa3xfRfRJQlp8NpacvIuMTONTtBkoLclBRUYHggGEoqQce/3wfvp53vdt+fnLoW+Z81YBKAVQZTPYdjMj5tmdae1HDIpy3WIVtuNi9e9TW/9pmsXcmtVKBYTH+AIB9bRj+bmkBldbYzkt/vCsb//o1Gz/mAEdy9QAkhKnr8afkWAyN9oebn9bs8YIionHj0DgoIHAovx7fHTnv6pLajUEtcwoJiNFZN0I4zvPUnSK/yoizpbVQKSQkhTkvqG3nqM+6aY9aCIHGK7MQ3wVBDQCjGoe/t54ouqz29UKJYqMGR3MrUFALlNZe3iIXHmFx2Kf3wTmfAWgQSvh7qjFEnY/B3lXw1nKgsbsI8NYgsXE55sXfH7fvAuduGNRuID7AOmP4+HkGdWc4nG9dUGZ4bAC81M77JxHT2KMuqW5AVb37rT9cZTDBIiSoFECkf9csizltiPXa4y0ni5BTdvGRiPLaBnx35Dz2NMTiaI0Ov54sxo4CBWavO4/b3voN728/g9xyx8cLIXCqqAor95bipxwJ58pqIcGCeI8a/HF0DAIVHLHqjgb4A3H+apTVNOB/1x93dTntwq+ObqB3gBq/gDO/O8uhxqC+tn8IAOd9WPt5qBHkrUFpTQPOltZicC+d0567K1Q0bm4Q5q3q1Euzmuob6oNr+gVjR2YJVu85i6dvTHS4Xwggx+CJX/echXXitYCPwozgAD+UVdagyijh8LkKHD5XgZc2pCMqwBP9Qn1gFtaRjd9PQ0iIDfJCWNUJ+Hp4dtnrIxewmHFLRDX+XaHFusN5uMK/AffdcKVb7VvN/zvdgL1HzaB2OosAjjYu0XptvxCnP787n6euqLUOE0b4du0H2p+S4wAAa/bnoK7h9w1pTGYLTphCkFHnA4uwvrej1LkY7VeOW4b2wuRogVW398ILtwzCmN6BUEhAbnkdfj1ZjO0ZxThbWguNUoGRvTwwPtKCW4f2gpfUfdaDppaVFuRg+6Fj6OtnHQJf+ut5HDp6zMVVtY37fKXoweL9rUGdW16Hynoj/DzULq6o+yitB+pMAoHeGgyK9MOxcuc+f1yQNw6dq3DLmd+283nRuq79/+36AaGICvBEbnkdnv02DTP6Sag2AlsP5qLY4gsJAtf0C8HQaH8c/S0NwO/D8oGeSlyXFIc/JcdBX2vE8fxKZJXUQKtSINBbg5HxgcjKSMcnu7O79DWRawVGRCN5aF8U7jmLynpg9RE9Rl3p6qouH3vUbsBHq0CEzvphlFnI66mdqbDOOrX36r7Bl3XdblvF2HvU7hfUpY1BHdPFQa1USHhicgIkCfjvwVzMXpeHH3MUKK4yQA0zhnnrMSwm4JKX2ui81EjuE4QZo2Nwx/AojB8QCh9OFOux1EoFrh8QCgD4/mQVDp9z8rfyTsSgdhO25UNPFcl7dyF3U9B4Stp6ftr5fp/57X5D32UuCmoAuGVoL3w0eyR8PVSoNFg36ejl74nhmlwEqN1vYh7JQ2yQN2K8LRAA5n+2H4dSjrrFdpgMajfBoHa+ugYzyg3WP1/TL7hTjuGu56gNZqC28fxwVw9924xLCMWWBeOw+PoQ3BxrwZ3Do+AhmS/9QKJWRBnPQSWMOKs3YsFXR+yL38gZx4HcBIPa+c6V1QKQEOevRphf51x+ZOtRF1TWo67BDE+NslOO42yVjZebeqkEPJ14yVpbhfhqcWWEJ9KyO/Y8F65GJiyc59FTaSQLErzrcKxWjXOWABRVs0dNTtI3xBrUmQxqpznXeJ2uM1cju5C/lxp+HiqH47mDysbRZb9ukme21cg+2Z2NTzYdQHlF1278QfISpjagl78nzELC/x2S/7lqBrWbsPWo8yrqHC5ZofYRomuCWpIkt1zzu7LBOlFL5567c7ao6S5Z1LNJEjAuIQQSBHbn1GF7RrGrS2oVg9pNBPloEeClhhDA6WL2qjuq2mTd2lIBgYEh2k49VqwbLiWqbxz69tNwO0fqnoJ9tOjja52o+OzXh5FyRL4TyxjUbqRfqHU3IQZ1xxU1zvYO8gC0qs79ZxDXOKEs240mlLnj0Ldty8vU1FSkpqbK9kOX5CPMkAOlMOGs3ogn/3NIthPLOJnMjfQJ9cG+7DJkFjKoO6qo8frpUM/O7zG6W49aX2+GwWx9f/zcaOi76ZaXhedO45ZhGejfvz8ATiCjlqklC3p71iOz3gfnEIw6o8XVJbWIQe1GOPPbOSxC2HvUoZ6dfzx7j7rEPXrU5/TW7rSfhwoqhXvtNmTb8rI4Lxurtx9H31LrN42MgwcQHJ+AXi6uj+QnSluHIikA+joj1mdUyXLFMg59uxF7UHPou0POVhjRYJGgVkoI7NzT0wB+71Gf19fBYJL/RMDTZdZwDvHtgjenE9lCmxPIqDUKCRjduL3quuNVstzpjkHtRmxBnV1SA6NZnkM07uBooXUTjkh/T3TCqqHNBPto4KVRWnd+KpP/Voq2oA717ZqtLYlcLSHcF74qgaoGC15eu1d2cxwY1G4kUucBb40SJotwm/OdcnS0wLocWXSAV5ccT5IktzpP3V161ESXSyFJiEYRAOA/aRV45T/bZDWxjEHtRiRJQh+ep+4Qk9mCtCJrjzo6oAtOUDeKD7Z+KcgqkXdQ1zaYkFtp7UmEMqipBwlR1MBLYYLRIqHSN9bV5TjgZDI30zfEB0dz9QzqdkrN06PWKKBWCAT7alFQ2jXH7dO4spzcL607fr4SAoCHUsBbq4Le1QURdRFJAmK1dUiv80WmXsLxExkO9ycmJkKlck1kMqjdDHvUHbPrtDWZQz2sw11dxR7URfLuUaflWaM5gJ1p6oHCNfU4Z/FHjcGMN3cXYHS59aqB4txsLLgLSEpKckldDGo30y+Ua353xG5bULfx+mnbYhpA+67JdZcZ+2nnKwEAAW50/TSRsygkYFh0AHaeKkGpRxQi4vpfct/zLqnL1QVQ29g+8E8XV8Ni4fKObWEwmbE/27oZQ1uvny4tyMHq7cfbvalD7xDrZLKymgb7Ps9yZOtR+2v5/xb1TIN7+UEBC2osKuSWy+MqDQa1m4kJ9IJGqUC90YK8Cnn8T+QuDp2tgMFkgb+HAr7tWKTKdl1ue67J9dKo0Mvf+u1Aruep641m+0gNh76pp9KqlAhXWP8dHMmtAOD65WkZ1G5GpVQgrnEGsdyHUeVm9+kSAMCQMA+4YjTLNr/gtExPWxzJqYDZIhDoqYSne2ybTdQpeimtI0tnimtQWW90GFFb9tXWLr90i0HthuznO7nmd5vsOGUN6qGduK1la/o0Dn/LdSLggbPWfXkTQzQu+SJDJBfeCiMCVA0QAFJzraFtG1ELiYrr8noY1G6ob+MuWnL9wJejynojjuRUAACuCHdNUDedXyBHB+1BzXFvoiiN9dRien4lXD0diEHthmwf+BlFVS6uxH3sOV0KiwB6B3sj1Ns1Fzv8fi21/C7RslgEg5qoiWB1AzzVStQ0mFFm6ZpVDC+GQe2GBoRbe9QZBVWc+X2Zfmsc9r6qb7DLarB9wcopr0W9UV6bc5wuroa+zghPtRK9eW0WERQSkBhh/azNt/i6thaXHp3apXewNzQqBWoazMgpd4+tE11tZxcFdWuzQ4O8NdB5qiGE/Ia/beenr4jWQdUVO5UQuYFBkToAQKnFCwaL6+KSQe2GVEoF+odZe2fp+ZUurkb+8vV1OF1cA4UEJPcO6tRjtTY7VJIk+2hIer68TlvYhr1HxAa6uBIi+Qj01iBC5wFAQn6D604JMajd1IBwPwDy+8CXo52Z1t50UpQ/dF7tuIC6jVqbHWr7hm5bWEQubAvBDI8LcHElRPIyKNL6WXu+wQNCuOZUI5cQdVOJEbagZo/6Umznp69x4flpm8G9rH9vx8/L5+8tp6wWZ0troVJIGBkXiKyMQvt9TZdOtXHl5gREXa1fqC9+TS9AnUWFvIo6l/Ru+a/NTSXahlAL5POBL0dCCOw8ZV3f25UTyWxsPerj+ZWwWAQUMjgfvKNxxGFYjD98tI4fCaUFOVh9ogJ9S+WxOQFRV9OoFAhVVCPf4odj5yuR5IIJ4AxqN2XrUeeU1aGq3ghfj84f0nVHGYXVKKk2wEOtwJWx/q4uB31CvKFVKVBtMOFsWS3ig71dXRJ2nioGAFzdN6TF+21D+c7G3jq5iwhlFfItfsgsqkZCdNcfn/8i3FSAtwbhfh4oqKzHyYIqjIjjJKCW7Mi0htDIuABkpB8H0L7dr5xFpVRgQIQfjuRU4Nh5vcuD2mwR+K1xxCFSoUdqamqXvT+t9dZNJpN9Ip4r/76IAMBXMsBbYUKNRYUcFyyDwKB2YwMifFFQWY/0/EoG9UXYzk/39TFj2VdbERIVh4yDBxAcn4BeLqppUKQ1qNPyKjFtSNs3+HCmtDw99HVGeKslfP/rbuyL7tr352K99fT0dNn8fRFJEhCpqUdmvQ+yqrr+dBVnfbuxgY3D32l5PE/dkgaTBXuzrLOZh0V4ICQqrt27XznT4Mbz1MfOu37mt+368qQwD4RFy+P9sZHL3xcRAIRr6qGQgHKDhOzyrt2qlkHtxoZG+wMADp0rd20hMrU3qxS1DWYE+2gR6y+foVPb5R7Hzle67HIPmy0nigBYv8gQ0cVpFAK9g63rV2w+07Xj3wxqN3ZlrPWa18yiauhrjS6uRn42H7deZjQhMRQKGW0HlRDuC41SgbKaBmSXum5luaLKevtCJ6OjPF1WB5G7GNj4JfvXrBo0mCxddlwGtRsL9tEiLsh6rcChHPaqmxJC4OfGoJ44MMzF1TjyUCsxLMYfALCrcY9sV/ip8f0ZFuOPIK/Lm67SdIlU6yQvrjVPPUdsoBc8lAKBHkqcr6jrsuMyqN2crVd96CyDuqlj5ytxXl8PT7VSFtdPXyi5j3Up092nS11Ww09pBQCAKYPCL/sxTZdI/WTTAZRXlHVWeUSyo1BImNhLYMVN4Yjrwis2GNRuzrY280EGtQPbsPe1/YPhoVa6uJrmbGuO7zlT6pLz1BW1Ddh9xvolYXIbghr4faY2J3lRT+Shsq7b35UY1G5ueGOPOiWnAiZz150zkbtN9mHvtoVQVxka4w8PtQIl1Q3ILOr6nbQ2Hy+E2SIwINy3S3sGRNR2DGo31y/UB75aFWobzDhRwA06AOBkQRXS8yuhVkq4YUCoq8tpkValtI+GuGL4+78HcwEANyVFdPmxiahtGNRuTqGQ7OepXXm+U07WHraG0PiEUAR4a1xczcXZzlN39YSyrJIa7M0qg0IC7hwR1aXHJqK2Y1B3A+MSrGs0b04vvETL7s9sEfj28HkAwO1Xynstq6sbJ7ntyCxBjcHUZcf9z4EcAMB1/UMQoeNlWURyx6DuBmyXHx3ILkNZTdeumCM3u0+XoqCyHjpPNcbLdNjbZkiUDnFBXqhtMOOnYwVdckyj2WIf9v7DyJguOebl4qVfRC1jUHcDUQFeSIzwg0X8vtJUT/XVQWtvcdqQCGhV8pvt3ZQkSbh1mLXXv+5wXpccc2NaAYqrDAj20eCGRHl9keGlX0QtY1B3E7Ze9ebjXdMzk6O8ijr8cDQfAPCHkS7Yi64dbh9mPUf826kSFFbWd+qxLBaBlVtOAQBmjomDWim/f/689IuoOUm4erHhbqSyshI6nQ56vR5+fn7tfp64p34AIABIuKu3BX9KjkNSUlKrj0nL02Paip3wVCtx+LmJLrl2+Ph5PRZ/fxzPTx+IgY0bT3T28zVt89+Defjwtywk9w7Cs9MSsfj740gM98HHu8/BXwuUG6zvqfW9RZM/X3hN5MXatfRnXKKdZP9tUKQfbhoSjhW/nIJGpUBFXdvPS/fy94CHyhqwnhoVXr1zSLP35vsjefjH2lTccWUvfH0oDw+P74Mfjhagss6Ic+WtraZ0qdd9qdfa1nZd9Rh3fW7WI9/Xar09++Wb0BW4zaVste2C+kGRfujl74m8ijp8f+Q87hrR9T3KjMJq7M0qQ0ZhtVOC+nKez9YmJacCa/afAwD8ZVwf++2lNQ0QAMoNwO/vadP3tqX3+VLtLnzMpdsJAGnnK+HnqUat0YJaY/uuec+rcOx1t/Te7D1ThiqDGTtPW/+7I7MUaecvZ4e1y3l/nPeedN1j3PW5WY/rnvtyH9M15Df2Re0iSRJmJscCAN7fccbluzJ1tV9PFKG2wYzECD9c209+S4YSEbUXg7obmTE6Bj5aFTIKq7E1o9jV5XQp2yS6+RP6dfnyfnLXYDK7ugQi6gAGdTfi56HGPY2TqN7ZerpH9Kptr9EsgAmJYZgks52y5OBkYdcvUUpEzsOg7mb+fHU81EoJe7PKsGZ/jqvL6XQ/pllneWuUEhbfMoi9aQfWLzFV9V23mAoROR+DupuJ9PfEwkkJAIDF3x9DZmH3Xf/70z1nsem4dcj79it7oZc/V9lqqkDfuZd7EVHXYFBf4K233kJ8fDw8PDwwfPhw7Nixw9Ultdmca3rjmn7BqDdaMOvDfUjN1bu6JKeqN5qx+PtjePabNPttY3pzAhlgXUL1XFktAKDaYD033TeUu2MRuTNentXEl19+ifnz5+Ott97CVVddhXfffRdTp07F8ePHERMjr+UWW6NQSFh+91D84d3dOFNSgzve2YW/XNsb942JRaifh6vLa7eymgZ8l5KHj3Zl42ypNYzGDwjBryd61sS5prZnFKGy3oiqehMyCquwL6sM+Y09aSUAM4BQXw+cKqpxaZ1E1H4M6iaWL1+O+++/Hw888AAA4I033sBPP/2Et99+G0uXLnVxdW0T4qvFN3Ovwt/WpOCXE0X495ZTWPnrKSSE+yExwhdB3hr4e2ng76WGh0oJSQIUkmT/r+3Ptsv7hQAERON/rWwTuWz3Hci2Lvm4L6sUJouAEMLatvH+39s6Pl9jE6CxvRDW597f+HyrfsvGW1tPIaPJpKggbw1eu2sIKutMbhXUQjh3z/C1h89jbeMmJDZqpQSjWSA22AtnSmqdejwi6noM6kYNDQ04ePAgnnrqKYfbJ02ahF27drX4GIPBAIPBYP9dr7cOMVdWXs7iEhdnMdg+XC3IO30WKZ4VqK5u38zdP/cXGOTliR8zq3CypAHHsmtxLLtzlxldveMkVjvxjMGh07+HTYxOhQl9fHBtrBc8y8/gwNlaWAy1yDh5Arvrz9nbZTTeXl1ugMUgn8uT8s/pYTGonfZ8fQPVCPJUwlOtQKi3Cn2D1NiTU4stWXWo1Vtfe3mxCRZD42YtRiOgdt7xiXqyjn7WA4Cvr+8lJ8EyqBuVlJTAbDYjLMzx8p6wsDAUFLQcbEuXLsXixYub3R4d7bxVwb5o/CGrHAC/tXD731tpLyfOrqe158u5jDZE1H66Nzr+HJez5DSD+gIXfrMRQlz0287TTz+Nxx9/3P67xWJBWVkZgoKCOnSZUGVlJaKjo5GTk9OhNcO7K74/reP70zq+P63j+9M6Z78/vr6+l2zDoG4UHBwMpVLZrPdcVFTUrJdto9VqodVqHW7z9/d3Wk1+fn78h9IKvj+t4/vTOr4/reP707qufH94eVYjjUaD4cOHY/PmzQ63b968GWPHjnVRVURE1NOxR93E448/jpkzZ2LEiBFITk7Ge++9h3PnzuEvf/mLq0sjIqIeikHdxB/+8AeUlpbihRdeQH5+PgYPHowNGzYgNja2S+vQarV4/vnnmw2rkxXfn9bx/Wkd35/W8f1pnSveH0n0hJ0biIiI3BTPURMREckYg5qIiEjGGNREREQyxqAmIiKSMQa1zHSHbTY7y/bt2zF9+nRERkZCkiR88803ri5JNpYuXYqRI0fC19cXoaGhuPXWW3Hy5ElXlyUbb7/9NoYMGWJfpCI5ORk//vijq8uSraVLl0KSJMyfP9/VpcjGokWLIEmSw094eHiXHJtBLSO2bTafeeYZHD58GNdccw2mTp2Kc+fOXfrBPUBNTQ2uuOIKrFy50tWlyM62bdvwyCOPYM+ePdi8eTNMJhMmTZqEmhpubwkAUVFRePnll3HgwAEcOHAA119/PW655RYcO3bM1aXJzv79+/Hee+9hyJAhri5FdgYNGoT8/Hz7T2pqatccWJBsjBo1SvzlL39xuG3AgAHiqaeeclFF8gVArFu3ztVlyFZRUZEAILZt2+bqUmQrICBA/N///Z+ry5CVqqoq0a9fP7F582Zx3XXXiccee8zVJcnG888/L6644gqXHJs9apmwbbM5adIkh9tb22aT6GJsW64GBga6uBL5MZvNWLNmDWpqapCcnOzqcmTlkUcewU033YQJEya4uhRZyszMRGRkJOLj43HPPffgzJkzXXJcrkwmE+3ZZpOoJUIIPP7447j66qsxePBgV5cjG6mpqUhOTkZ9fT18fHywbt06DBw40NVlycaaNWtw6NAh7N+/39WlyNLo0aPxySefoH///igsLMSLL76IsWPH4tixYwgKCurUYzOoZaYt22wStWTu3Lk4evQodu7c6epSZCUhIQEpKSmoqKjA119/jVmzZmHbtm0MawA5OTl47LHHsGnTJnh4eLi6HFmaOnWq/c9JSUlITk5Gnz598PHHHztsd9wZGNQy0Z5tNoku9Oijj+K7777D9u3bERUV5epyZEWj0aBv374AgBEjRmD//v3417/+hXfffdfFlbnewYMHUVRUhOHDh9tvM5vN2L59O1auXAmDwQClUunCCuXH29sbSUlJyMzM7PRj8Ry1THCbTeoIIQTmzp2LtWvXYsuWLYiPj3d1SbInhIDBYHB1GbJwww03IDU1FSkpKfafESNG4I9//CNSUlIY0i0wGAxIT09HREREpx+LPWoZ4TabrauursapU6fsv2dlZSElJQWBgYGIiYlxYWWu98gjj+Dzzz/Ht99+C19fX/vIjE6ng6enp4urc71//OMfmDp1KqKjo1FVVYU1a9Zg69at2Lhxo6tLkwVfX99m8xm8vb0RFBTEeQ6NFi5ciOnTpyMmJgZFRUV48cUXUVlZiVmzZnX6sRnUMiKXbTbl6sCBAxg/frz9d9t5oVmzZmHVqlUuqkoe3n77bQDAuHHjHG7/6KOPMHv27K4vSGYKCwsxc+ZM5OfnQ6fTYciQIdi4cSMmTpzo6tLITeTm5uLee+9FSUkJQkJCMGbMGOzZs6dLPp+5zSUREZGM8Rw1ERGRjDGoiYiIZIxBTUREJGMMaiIiIhljUBMREckYg5qIiEjGGNREREQyxqAmIiKSMQY1kYtIkoRvvvnG1WV0ioaGBvTt2xe//fabq0uRnTvvvBPLly93dRnkRhjURE40e/ZsSJIESZKgVqsRFhaGiRMn4sMPP4TFYnFom5+f77B1XmvcLdTfe+89xMbG4qqrrnJ1KQ7GjRuH+fPnu7SG5557Di+99BIqKytdWge5DwY1kZNNmTIF+fn5yM7Oxo8//ojx48fjsccew7Rp02AymeztwsPDodVqXVhp51mxYgUeeOABV5fRLkIIh78nZxsyZAji4uLw2WefddoxqHthUBM5mVarRXh4OHr16oUrr7wS//jHP/Dtt9/ixx9/dNg8pGkvuaGhAXPnzkVERAQ8PDwQFxeHpUuXAgDi4uIAALfddhskSbL/fvr0adxyyy0ICwuDj48PRo4ciZ9//tmhlri4OCxZsgR//vOf4evri5iYGLz33nsObXJzc3HPPfcgMDAQ3t7eGDFiBPbu3Wu///vvv8fw4cPh4eGB3r17Y/Hixa0G2aFDh3Dq1CncdNNNDrf//e9/R//+/eHl5YXevXvj2WefhdFotN+/aNEiDB06FJ9++ini4uKg0+lwzz33oKqqyt6mqqoKf/zjH+Ht7Y2IiAi8/vrrzXrJb731Fvr16wcPDw+EhYXhzjvvBGAd7di2bRv+9a9/2Uc9srOzsXXrVkiShJ9++gkjRoyAVqvFjh07YDAYMG/ePISGhsLDwwNXX3019u/fbz9O08cNGzYMnp6euP7661FUVIQff/wRiYmJ8PPzw7333ova2lqH9+Lmm2/GF198cdH3kMiBICKnmTVrlrjllltavO+KK64QU6dOtf8OQKxbt04IIcRrr70moqOjxfbt20V2drbYsWOH+Pzzz4UQQhQVFQkA4qOPPhL5+fmiqKhICCFESkqKeOedd8TRo0dFRkaGeOaZZ4SHh4c4e/as/RixsbEiMDBQvPnmmyIzM1MsXbpUKBQKkZ6eLoQQoqqqSvTu3Vtcc801YseOHSIzM1N8+eWXYteuXUIIITZu3Cj8/PzEqlWrxOnTp8WmTZtEXFycWLRo0UXfg9dff10MGDCg2e3/+7//K3777TeRlZUlvvvuOxEWFiZeeeUV+/3PP/+88PHxEbfffrtITU0V27dvF+Hh4eIf//iHvc0DDzwgYmNjxc8//yxSU1PFbbfdJnx9fcVjjz0mhBBi//79QqlUis8//1xkZ2eLQ4cOiX/9619CCCEqKipEcnKymDNnjsjPzxf5+fnCZDKJX3/9VQAQQ4YMEZs2bRKnTp0SJSUlYt68eSIyMlJs2LBBHDt2TMyaNUsEBASI0tJSIYSwP27MmDFi586d4tChQ6Jv377iuuuuE5MmTRKHDh0S27dvF0FBQeLll192eC82bNggtFqtqK+vv+j7SGTDoCZyotaC+g9/+INITEy0/940qB999FFx/fXXC4vF0uJjm7ZtzcCBA8WKFSvsv8fGxor77rvP/rvFYhGhoaHi7bffFkII8e677wpfX197+FzommuuEUuWLHG47dNPPxUREREXreGxxx4T119//SVrffXVV8Xw4cPtvz///PPCy8tLVFZW2m974oknxOjRo4UQQlRWVgq1Wi2++uor+/0VFRXCy8vLHtRff/218PPzc3iOpq677jp7Wxtb4H7zzTf226qrq4VarRafffaZ/baGhgYRGRkpXn31VYfH/fzzz/Y2S5cuFQDE6dOn7bc99NBDYvLkyQ7HPHLkiAAgsrOzW32PiIQQgvtRE3URIQQkSWrxvtmzZ2PixIlISEjAlClTMG3aNEyaNKnV56upqcHixYuxfv16nD9/HiaTCXV1dTh37pxDuyFDhtj/LEkSwsPDUVRUBABISUnBsGHDEBgY2OIxDh48iP379+Oll16y32Y2m1FfX4/a2lp4eXk1e0xdXR08PDya3f7f//4Xb7zxBk6dOoXq6mqYTCb4+fk5tImLi4Ovr6/994iICHutZ86cgdFoxKhRo+z363Q6JCQk2H+fOHEiYmNj0bt3b0yZMgVTpkzBbbfd1mKdFxoxYoT9z6dPn4bRaHSYDKdWqzFq1Cikp6c7PK7p+xsWFmYf2m962759+xwe4+npCQDNhsSJWsJz1ERdJD09HfHx8S3ed+WVVyIrKwv/+7//i7q6Otx99932c6sX88QTT+Drr7/GSy+9hB07diAlJQVJSUloaGhwaKdWqx1+lyTJPgPdFhgXY7FYsHjxYqSkpNh/UlNTkZmZ2WIYA0BwcDDKy8sdbtuzZw/uueceTJ06FevXr8fhw4fxzDPPtKlWIYT9tqZstwOAr68vDh06hC+++AIRERF47rnncMUVV6CioqLV1wkA3t7ezZ6zpWNdeFvTmm2z/S/2GmzKysoAACEhIZesi4hBTdQFtmzZgtTUVNxxxx0XbePn54c//OEPeP/99/Hll1/i66+/tn+gq9VqmM1mh/Y7duzA7NmzcdtttyEpKQnh4eHIzs5uU11DhgxBSkqK/TgXuvLKK3Hy5En07du32Y9C0fLHx7Bhw3DixAmHAP3tt98QGxuLZ555BiNGjEC/fv1w9uzZNtXap08fqNVqh95pZWUlMjMzHdqpVCpMmDABr776Ko4ePYrs7Gxs2bIFAKDRaJq9jy3p27cvNBoNdu7cab/NaDTiwIEDSExMbFPdLUlLS0NUVBSCg4M7/FzU/XHom8jJDAYDCgoKYDabUVhYiI0bN2Lp0qWYNm0a/vSnP7X4mNdffx0REREYOnQoFAoFvvrqK4SHh8Pf3x+AdUj4l19+wVVXXQWtVouAgAD07dsXa9euxfTp0yFJEp599tlmPbdLuffee7FkyRLceuutWLp0KSIiInD48GFERkYiOTkZzz33HKZNm4bo6GjcddddUCgUOHr0KFJTU/Hiiy+2+Jzjx49HTU0Njh07hsGDBwOwBt+5c+ewZs0ajBw5Ej/88APWrVvXplp9fX0xa9YsPPHEEwgMDERoaCief/55KBQKey93/fr1OHPmDK699loEBARgw4YNsFgs9uHxuLg47N27F9nZ2fDx8bnokL+3tzf++te/2o8VExODV199FbW1tbj//vvbVHdLduzYcclTG0Q27FETOdnGjRsRERGBuLg4TJkyBb/++iv+/e9/49tvv4VSqWzxMT4+PnjllVcwYsQIjBw5EtnZ2diwYYO917ps2TJs3rwZ0dHRGDZsGABruAcEBGDs2LGYPn06Jk+ejCuvvLJNtWo0GmzatAmhoaG48cYbkZSUhJdfftle5+TJk7F+/Xps3rwZI0eOxJgxY7B8+XLExsZe9DmDgoJw++23O1wnfMstt+Bvf/sb5s6di6FDh2LXrl149tln21QrACxfvhzJycmYNm0aJkyYgKuuugqJiYn2YXh/f3+sXbsW119/PRITE/HOO+/giy++wKBBgwAACxcuhFKpxMCBAxESEtLsfH5TL7/8Mu644w7MnDkTV155JU6dOoWffvoJAQEBba67qfr6eqxbtw5z5szp0PNQzyGJpuNTREROkJqaigkTJuDUqVMOk8OcraamBr169cKyZcuc0tPtCm+++Sa+/fZbbNq0ydWlkJtgj5qInC4pKQmvvvpqm8+ZX8rhw4fxxRdf4PTp0zh06BD++Mc/ArD22N2FWq3GihUrXF0GuRH2qInIbRw+fBgPPPAATp48CY1Gg+HDh2P58uVISkpydWlEnYZBTUREJGMc+iYiIpIxBjUREZGMMaiJiIhkjEFNREQkYwxqIiIiGWNQExERyRiDmoiISMYY1ERERDL2/wG50faxdDqYUQAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 500x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sns\n",
    "edge_index, _, distances = interaction_module.interactions(demo_batch)\n",
    "\n",
    "fig = sns.displot(x=distances.numpy(), kde=True, rug=True)\n",
    "fig.axes[0,0].set_title(f\"Batch size {demo_batch['ligand'].batch.max().item() + 1}\")\n",
    "_ = fig.axes[0,0].set_xlabel(\"Distance (angstrom)\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.8 ('kinodata')",
   "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": "3.10.8"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "528a87b2665604848835e963d9b59f6a2ef408edd78da891397a1684bf62ca9e"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}