PBMC-prepro.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "import scipy.sparse\n",
    "import scipy.io\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "plt.rcParams['pdf.fonttype'] = 42\n",
    "plt.rcParams['ps.fonttype'] = 42\n",
    "\n",
    "from IPython.display import clear_output\n",
    "\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.decomposition import FactorAnalysis\n",
    "from R_interop import ZIFA\n",
    "from R_interop import ZINB\n",
    "from R_interop import SIMLR\n",
    "\n",
    "from benchmarking import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_path = \"/home/ubuntu/single-cell-scVI/data/10xPBMCs/\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# STEP 1: gene filtering once and for all from original matrix, this was done to produce the file that will be loaded in step 2. Do not execute, we provide the files needed to run the rest of the code"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "gene_names = np.array(pd.read_csv(data_path + \"michael_gene_names.csv\", index_col=0)[\"x\"], dtype=str)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(12039, 22156)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "expression = pd.DataFrame(scipy.io.mmread(data_path + \"count.mtx\").T.A, columns=gene_names)\n",
    "expression.shape "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### select genes that are in the DE data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def dumb_prefix(x, n):\n",
    "    if x != \"GS\":\n",
    "        return str(n) + \"_\" + x\n",
    "    else:\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(6394, 19)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# load micro_array data and align existing genes\n",
    "exp_data_1 = pd.read_csv(data_path+\"gene_data/GSE8835_CD4_CD8.tab\", delimiter=\"\\t\")\n",
    "exp_data_1[\"GS\"] = exp_data_1.index\n",
    "exp_data_1.columns = [dumb_prefix(x, \"CD\") for x in exp_data_1.columns] \n",
    "exp_data_2 = pd.read_csv(data_path+\"gene_data/GSE29618_Bcells_mDC.tab\", delimiter=\"\\t\")\n",
    "exp_data_2[\"GS\"] = exp_data_2.index\n",
    "exp_data_2.columns = [dumb_prefix(x, \"BDC\") for x in exp_data_2.columns]\n",
    "exp_data_3 = pd.read_csv(data_path+\"gene_data/GSE22886_Bcells_mDC.tab\", delimiter=\"\\t\")\n",
    "exp_data_3[\"GS\"] = exp_data_3.index\n",
    "exp_data_3.columns = [dumb_prefix(x, \"BDC2\") for x in exp_data_3.columns]\n",
    "exp_data = pd.merge(exp_data_1, \\\n",
    "                          exp_data_2, how=\"inner\", on=\"GS\")\n",
    "exp_data = pd.merge(exp_data, \\\n",
    "                          exp_data_3, how=\"inner\", on=\"GS\")\n",
    "exp_data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# order correspondance from datasets\n",
    "gene_dataset_name = pd.DataFrame(gene_names, columns=[\"ENSG\"])\n",
    "gene_correspondance = pd.read_csv(data_path+\"de/new_rownames.tab\", delimiter=\"\\t\")\n",
    "gene_correspondance.columns = [\"ENSG\", \"GS\"]\n",
    "ordered_corres = pd.merge(gene_dataset_name, \\\n",
    "                          gene_correspondance, how=\"inner\", on=\"ENSG\", left_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3346, 20)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "micro_array_result = pd.merge(ordered_corres, exp_data, how=\"inner\", on=\"GS\", right_index=True)\n",
    "micro_array_result.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "micro_array_result.to_csv(data_path+\"de/gene_info.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "expression = expression[micro_array_result[\"ENSG\"]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "de_expression = np.copy(expression.as_matrix())\n",
    "de_gene_names = expression.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {
    "collapsed": true,
    "deletable": false,
    "editable": false,
    "run_control": {
     "frozen": true
    }
   },
   "outputs": [],
   "source": [
    "np.savetxt(data_path+\"gene_data/gene_names.tab\", gene_names, fmt='%s')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## STEP2: now clusters and qc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "design = pd.read_csv(data_path + \"design.csv\", index_col=0)\n",
    "dim_names = [\"WPC\"+str(i) for i in range(1, 8)]\n",
    "qc_pc = design[dim_names].as_matrix()\n",
    "normalized_qc = pd.read_csv(data_path + \"full_qc.csv\", index_col=0)\n",
    "barcodes = pd.read_csv(data_path + \"barcodes.csv\", index_col=1)\n",
    "batch = pd.read_csv(data_path + \"batch.csv\", index_col=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "raw_qc_8 = pd.read_csv(data_path + \"molecule_qc_8k.txt\", index_col=0, sep=\"\\t\")\n",
    "raw_qc_8.index = raw_qc_8.index.map(lambda s: s+\"-1\")\n",
    "raw_qc_4 = pd.read_csv(data_path + \"molecule_qc_4k.txt\", index_col=0, sep=\"\\t\")\n",
    "raw_qc_8 = raw_qc_8.loc[barcodes.iloc[np.where(batch[\"x\"] == \"pbmc8k\")[0], :].index.values]\n",
    "number_suffix = barcodes.iloc[np.where(batch[\"x\"] == \"pbmc4k\")[0], :].index.map(lambda s: str(s.split(\"-\")[1])).values\n",
    "barcode_prefix = barcodes.iloc[np.where(batch[\"x\"] == \"pbmc4k\")[0], :].index.map(lambda s: str(s.split(\"-\")[0])).values\n",
    "raw_qc_4 = raw_qc_4.loc[barcode_prefix]\n",
    "raw_qc_4.index = raw_qc_4.index +\"-\" +number_suffix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "raw_qc = pd.concat((raw_qc_8, raw_qc_4)).loc[barcodes.index]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((12039, 7), (12039, 9), (12039, 9), (12039, 1))"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "qc_pc.shape, normalized_qc.shape, raw_qc.shape, batch.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dataset_population\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[(0, 'B cells', 1625),\n",
       " (1, 'CD14+ Monocytes', 2237),\n",
       " (2, 'CD4 T cells', 5024),\n",
       " (3, 'CD8 T cells', 1452),\n",
       " (4, 'Dendritic Cells', 339),\n",
       " (5, 'FCGR3A+ Monocytes', 351),\n",
       " (6, 'Megakaryocytes', 88),\n",
       " (7, 'NK cells', 459),\n",
       " (8, 'Other', 464)]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bio = pd.read_csv(data_path + \"bio.csv\", index_col=0)\n",
    "list_clusters = np.unique(bio[\"x\"])\n",
    "def string_to_cluster(string):\n",
    "    return np.where(list_clusters == string)[0][0]\n",
    "data_bio = bio[\"x\"].apply(lambda i: string_to_cluster(i))\n",
    "clusters = data_bio.values\n",
    "counts = data_bio.value_counts()[range(0, 9)].values\n",
    "print \"dataset_population\"\n",
    "zip(list(np.unique(clusters)), list(list_clusters), list(counts))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "_, batch = np.unique(batch, return_inverse=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, \\\n",
    "de_X_train, de_X_test, \\\n",
    "c_train, c_test, \\\n",
    "r_train, r_test, \\\n",
    "b_train, b_test, \\\n",
    "qc_train, qc_test = \\\n",
    "                train_test_split(expression, de_expression, clusters, qc_pc, \\\n",
    "                                 batch, raw_qc.as_matrix(), random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": true,
    "deletable": false,
    "editable": false,
    "run_control": {
     "frozen": true
    }
   },
   "outputs": [],
   "source": [
    "scipy.sparse.save_npz(data_path + \"data_train.npz\", scipy.sparse.coo_matrix(X_train))\n",
    "scipy.sparse.save_npz(data_path + \"data_test.npz\", scipy.sparse.coo_matrix(X_test))\n",
    "\n",
    "np.save(data_path + \"de/data_train\", de_X_train)\n",
    "np.save(data_path + \"de/data_test\", de_X_test)\n",
    "\n",
    "np.savetxt(data_path + \"label_train\", c_train)\n",
    "np.savetxt(data_path + \"label_test\", c_test)\n",
    "\n",
    "np.savetxt(data_path + \"b_train\", b_train)\n",
    "np.savetxt(data_path + \"b_test\", b_test)\n",
    "\n",
    "np.save(data_path + \"design_train\", r_train)\n",
    "np.save(data_path + \"design_test\", r_test)\n",
    "\n",
    "np.save(data_path + \"qc_train\", qc_train)\n",
    "np.save(data_path + \"qc_test\", qc_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Now benchmarking"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_latent = 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#X_train = scipy.sparse.load_npz(data_path + \"data_train.npz\").A\n",
    "#X_test = scipy.sparse.load_npz(data_path + \"data_test.npz\").A\n",
    "X_train = np.load(data_path + \"de/data_train.npy\")\n",
    "X_test = np.load(data_path + \"de/data_test.npy\")\n",
    "r_train = np.load(data_path + \"design_train.npy\")\n",
    "r_test = np.load(data_path + \"design_test.npy\")\n",
    "c_train = np.loadtxt(data_path + \"label_train\")\n",
    "c_test = np.loadtxt(data_path + \"label_test\")\n",
    "b_train = np.loadtxt(data_path + \"b_train\")\n",
    "b_test = np.loadtxt(data_path + \"b_test\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "algorithms = [PCA(n_components=n_latent, svd_solver=\"full\"), \\\n",
    "              ZIFA.ZIFA(n_components=n_latent)]\n",
    "alg_names = [\"PPCA\", \"ZIFA\"]\n",
    "alg_logs = [True, True, False]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 PPCA\n",
      "CPU times: user 1min 52s, sys: 8.6 s, total: 2min 1s\n",
      "Wall time: 31.8 s\n",
      "[0.32141702134965328, 0.68377188523709931, 0.54595098016588517]\n",
      "0.616187184275\n",
      "-411.634143765\n",
      "1 ZIFA\n"
     ]
    }
   ],
   "source": [
    "for i, alg in enumerate(algorithms):\n",
    "    print i, alg_names[i]\n",
    "    \n",
    "    if i == 1:\n",
    "        #ZIFA do not converge in reasonable time\n",
    "        continue\n",
    "    \n",
    "    data_train = X_train\n",
    "    data_test = X_test\n",
    "    \n",
    "    if alg_logs:\n",
    "        data_train = np.log(1 + X_train)\n",
    "        data_test = np.log(1 + X_test)\n",
    "        \n",
    "    %time alg.fit(data_train)\n",
    "    latent = alg.transform(data_train)\n",
    "    \n",
    "    print cluster_scores(latent, 9, c_train)\n",
    "    print entropy_batch_mixing(latent, b_train)\n",
    "    \n",
    "    score = alg.score(data_test)\n",
    "    if alg_logs:\n",
    "        score -= np.mean(np.sum(data_test, axis=-1))\n",
    "        \n",
    "    print score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# ZINB crashes\n",
    "# zinb = ZINB.ZINB(n_components=n_latent, learn_V=True)\n",
    "# % time zinb.fit(X_train)\n",
    "# % time latent = zinb.transform(X_train)\n",
    "# print zinb.score(X_test)\n",
    "# print cluster_scores(latent, 9, c_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.78212435541262204, 0.83751868977433352, 0.88378315126477902]\n",
      "0.628411486367\n"
     ]
    }
   ],
   "source": [
    "simlr = SIMLR(n_clusters=9)\n",
    "%time simlr.fit_transform(X_train)\n",
    "clear_output()\n",
    "print cluster_scores(simlr.F, 9, c_train)\n",
    "print entropy_batch_mixing(simlr.F, b_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## imputation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_zero, i, j, ix = dropout(X_train)\n",
    "np.save(data_path + \"imputation/X_zero.npy\", X_zero)\n",
    "np.save(data_path + \"imputation/i.npy\", i)\n",
    "np.save(data_path + \"imputation/j.npy\", j)\n",
    "np.save(data_path + \"imputation/ix.npy\", ix)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_zero, i, j, ix = \\\n",
    "        np.load(data_path + \"imputation/X_zero.npy\"),\\\n",
    "        np.load(data_path + \"imputation/i.npy\"), \\\n",
    "        np.load(data_path + \"imputation/j.npy\"), \\\n",
    "        np.load(data_path + \"imputation/ix.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# ZIFA Imputation\n",
    "zifa = ZIFA(n_components=10)\n",
    "%time zifa.fit(np.log(1 + X_zero))\n",
    "%time X_ZIFA_log = zifa.output_estimation(np.log(1 + X_zero))[\"EX\"]\n",
    "p_ZIFA = np.exp(- zifa.params[\"decay_coef\"] * X_ZIFA_log**2)\n",
    "print(\"ZIFA\", imputation_error(np.exp(X_ZIFA_log) - 1, X_train, i, j, ix), \\\n",
    "              imputation_dropout(np.log(p_ZIFA), X_zero, i, j, ix))\n",
    "del X_ZIFA_log\n",
    "del p_ZIFA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# ZINB Imputation\n",
    "zinb = ZINB(n_components=10)\n",
    "%time zinb.fit(X_zero)\n",
    "%time zinb.score(X_zero)\n",
    "fit_ZINB = zinb.output_estimation()\n",
    "def softplus(x):\n",
    "    return np.log(1 + np.exp(x))\n",
    "X_ZINB = fit_ZINB[0]\n",
    "logit_ZINB = fit_ZINB[1]\n",
    "theta_ZINB = fit_ZINB[2]\n",
    "log_pi_ZINB = -softplus(- logit_ZINB)\n",
    "log_p_ZINB = softplus(- logit_ZINB + theta_ZINB * np.log(theta_ZINB + 1e-8) \\\n",
    "                      - theta_ZINB * np.log(theta_ZINB + X_ZINB + 1e-8)) -  softplus( - logit_ZINB)\n",
    "print(\"ZINB\", imputation_error(X_ZINB, X_train, i, j, ix),\\\n",
    "      imputation_dropout(log_pi_ZINB, X_zero, i, j, ix), imputation_dropout(log_p_ZINB, X_zero, i, j, ix))\n",
    "del fit_ZINB"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('MAGIC', 0.53948369692054254)\n"
     ]
    }
   ],
   "source": [
    "#MAGIC Imputation\n",
    "mean_MAGIC = np.load(data_path + \"imputation/X_zero_MAGIC.npy\")\n",
    "print(\"MAGIC\", imputation_error(mean_MAGIC, X_train, X_zero, i, j, ix))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# de benchmark"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from R_interop import IDR, deseq, mast, edgeR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "micro_array_result = pd.read_csv(data_path+\"de/gene_info.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>ENSG</th>\n",
       "      <th>GS</th>\n",
       "      <th>CD_logFC</th>\n",
       "      <th>CD_AveExpr</th>\n",
       "      <th>CD_t</th>\n",
       "      <th>CD_P.Value</th>\n",
       "      <th>CD_adj.P.Val</th>\n",
       "      <th>CD_B</th>\n",
       "      <th>BDC_logFC</th>\n",
       "      <th>...</th>\n",
       "      <th>BDC_t</th>\n",
       "      <th>BDC_P.Value</th>\n",
       "      <th>BDC_adj.P.Val</th>\n",
       "      <th>BDC_B</th>\n",
       "      <th>BDC2_logFC</th>\n",
       "      <th>BDC2_AveExpr</th>\n",
       "      <th>BDC2_t</th>\n",
       "      <th>BDC2_P.Value</th>\n",
       "      <th>BDC2_adj.P.Val</th>\n",
       "      <th>BDC2_B</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5</td>\n",
       "      <td>ENSG00000188976</td>\n",
       "      <td>NOC2L</td>\n",
       "      <td>0.248976</td>\n",
       "      <td>4.193230</td>\n",
       "      <td>1.308282</td>\n",
       "      <td>0.201059</td>\n",
       "      <td>0.478547</td>\n",
       "      <td>-5.465606</td>\n",
       "      <td>0.023141</td>\n",
       "      <td>...</td>\n",
       "      <td>0.199196</td>\n",
       "      <td>0.844018</td>\n",
       "      <td>0.936848</td>\n",
       "      <td>-6.964569</td>\n",
       "      <td>0.532702</td>\n",
       "      <td>5.985696</td>\n",
       "      <td>1.671825</td>\n",
       "      <td>0.116222</td>\n",
       "      <td>0.148453</td>\n",
       "      <td>-6.166109</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>7</td>\n",
       "      <td>ENSG00000187608</td>\n",
       "      <td>ISG15</td>\n",
       "      <td>-0.036463</td>\n",
       "      <td>5.769094</td>\n",
       "      <td>-0.216990</td>\n",
       "      <td>0.829737</td>\n",
       "      <td>0.919775</td>\n",
       "      <td>-6.272810</td>\n",
       "      <td>0.987879</td>\n",
       "      <td>...</td>\n",
       "      <td>3.449654</td>\n",
       "      <td>0.002391</td>\n",
       "      <td>0.019573</td>\n",
       "      <td>-2.095965</td>\n",
       "      <td>0.933572</td>\n",
       "      <td>6.303954</td>\n",
       "      <td>2.202136</td>\n",
       "      <td>0.044485</td>\n",
       "      <td>0.062851</td>\n",
       "      <td>-5.300666</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>36</td>\n",
       "      <td>ENSG00000149527</td>\n",
       "      <td>PLCH2</td>\n",
       "      <td>0.505600</td>\n",
       "      <td>3.225463</td>\n",
       "      <td>1.834575</td>\n",
       "      <td>0.076854</td>\n",
       "      <td>0.327638</td>\n",
       "      <td>-4.707168</td>\n",
       "      <td>-0.119499</td>\n",
       "      <td>...</td>\n",
       "      <td>-1.014781</td>\n",
       "      <td>0.321715</td>\n",
       "      <td>0.598441</td>\n",
       "      <td>-6.463128</td>\n",
       "      <td>-0.977612</td>\n",
       "      <td>4.232389</td>\n",
       "      <td>-3.119639</td>\n",
       "      <td>0.007349</td>\n",
       "      <td>0.012455</td>\n",
       "      <td>-3.569136</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>37</td>\n",
       "      <td>ENSG00000157881</td>\n",
       "      <td>PANK4</td>\n",
       "      <td>-0.093142</td>\n",
       "      <td>4.319701</td>\n",
       "      <td>-0.772409</td>\n",
       "      <td>0.446126</td>\n",
       "      <td>0.686981</td>\n",
       "      <td>-6.001137</td>\n",
       "      <td>0.024259</td>\n",
       "      <td>...</td>\n",
       "      <td>0.111785</td>\n",
       "      <td>0.912052</td>\n",
       "      <td>0.963969</td>\n",
       "      <td>-6.978668</td>\n",
       "      <td>0.183735</td>\n",
       "      <td>6.453772</td>\n",
       "      <td>1.306024</td>\n",
       "      <td>0.212100</td>\n",
       "      <td>0.252955</td>\n",
       "      <td>-6.668252</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>40</td>\n",
       "      <td>ENSG00000157873</td>\n",
       "      <td>TNFRSF14</td>\n",
       "      <td>0.073292</td>\n",
       "      <td>4.949553</td>\n",
       "      <td>0.353608</td>\n",
       "      <td>0.726192</td>\n",
       "      <td>0.865967</td>\n",
       "      <td>-6.233949</td>\n",
       "      <td>-0.377352</td>\n",
       "      <td>...</td>\n",
       "      <td>-2.465162</td>\n",
       "      <td>0.022361</td>\n",
       "      <td>0.106855</td>\n",
       "      <td>-4.215261</td>\n",
       "      <td>-0.075492</td>\n",
       "      <td>6.920026</td>\n",
       "      <td>-0.540324</td>\n",
       "      <td>0.597263</td>\n",
       "      <td>0.639609</td>\n",
       "      <td>-7.373536</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0             ENSG        GS  CD_logFC  CD_AveExpr      CD_t  \\\n",
       "0           5  ENSG00000188976     NOC2L  0.248976    4.193230  1.308282   \n",
       "1           7  ENSG00000187608     ISG15 -0.036463    5.769094 -0.216990   \n",
       "2          36  ENSG00000149527     PLCH2  0.505600    3.225463  1.834575   \n",
       "3          37  ENSG00000157881     PANK4 -0.093142    4.319701 -0.772409   \n",
       "4          40  ENSG00000157873  TNFRSF14  0.073292    4.949553  0.353608   \n",
       "\n",
       "   CD_P.Value  CD_adj.P.Val      CD_B  BDC_logFC    ...        BDC_t  \\\n",
       "0    0.201059      0.478547 -5.465606   0.023141    ...     0.199196   \n",
       "1    0.829737      0.919775 -6.272810   0.987879    ...     3.449654   \n",
       "2    0.076854      0.327638 -4.707168  -0.119499    ...    -1.014781   \n",
       "3    0.446126      0.686981 -6.001137   0.024259    ...     0.111785   \n",
       "4    0.726192      0.865967 -6.233949  -0.377352    ...    -2.465162   \n",
       "\n",
       "   BDC_P.Value  BDC_adj.P.Val     BDC_B  BDC2_logFC  BDC2_AveExpr    BDC2_t  \\\n",
       "0     0.844018       0.936848 -6.964569    0.532702      5.985696  1.671825   \n",
       "1     0.002391       0.019573 -2.095965    0.933572      6.303954  2.202136   \n",
       "2     0.321715       0.598441 -6.463128   -0.977612      4.232389 -3.119639   \n",
       "3     0.912052       0.963969 -6.978668    0.183735      6.453772  1.306024   \n",
       "4     0.022361       0.106855 -4.215261   -0.075492      6.920026 -0.540324   \n",
       "\n",
       "   BDC2_P.Value  BDC2_adj.P.Val    BDC2_B  \n",
       "0      0.116222        0.148453 -6.166109  \n",
       "1      0.044485        0.062851 -5.300666  \n",
       "2      0.007349        0.012455 -3.569136  \n",
       "3      0.212100        0.252955 -6.668252  \n",
       "4      0.597263        0.639609 -7.373536  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "micro_array_result.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def boostrap_scores_de(p_value, model, it, p_prior = 0.1, rank_auc=111):\n",
    "    l = []\n",
    "    for i in range(it):\n",
    "        print i\n",
    "        model_p_value = model.fit().astype(np.float64)\n",
    "        res = idr.fit(-np.log(model_p_value), -np.log(p_value), p_prior=p_prior)\n",
    "        auc = auc_score_threshold(p_value, model_p_value, rank_auc, p_value=True)\n",
    "        res = list(res)\n",
    "        res.append([auc])\n",
    "        l.append([x[0] for x in res])    \n",
    "    return l"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "gene_names = micro_array_result[\"ENSG\"]\n",
    "gene_symbols = micro_array_result[\"GS\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# interest = \"CD\"\n",
    "# couple_celltypes = (3, 2)\n",
    "# rank_auc = 111\n",
    "# p_prior = 0.03\n",
    "\n",
    "interest = \"BDC\"\n",
    "couple_celltypes = (0, 4)\n",
    "rank_auc = 800\n",
    "p_prior = 0.25\n",
    "\n",
    "# getting p_values\n",
    "p_value = micro_array_result[interest + \"_adj.P.Val\"].values\n",
    "signed_p_value = - np.log10(p_value) * np.sign(micro_array_result[interest + \"_logFC\"].values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "idr = IDR.IDR()\n",
    "\n",
    "deseq_2 = deseq.DESeq2(100, 100, \\\n",
    "                 data_path + \"de/data_train.npy\",\\\n",
    "                 data_path + \"label_train\", \\\n",
    "                    couple_celltypes)\n",
    "mast_ = mast.MAST(100, 100, \\\n",
    "                 data_path + \"de/data_train.npy\",\\\n",
    "                 data_path + \"label_train\", \\\n",
    "                    couple_celltypes)\n",
    "\n",
    "edger_ = edgeR.edgeR(100, 100, data_path + \"de/data_train.npy\",\\\n",
    "                 data_path + \"label_train\", \\\n",
    "                    couple_celltypes) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "from rpy2.robjects.packages import importr\n",
    "from rpy2.robjects.vectors import FloatVector\n",
    "\n",
    "stats = importr('stats')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "p, fc = mast_.fit(return_fc=True)\n",
    "p = np.array(stats.p_adjust(FloatVector(p), method = 'BH'))\n",
    "signed_p = - np.log10(p) * np.sign(fc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "65\n",
      "13.1367285978\n",
      "26.5712319467\n",
      "33.7548348595\n",
      "10.4782400702\n",
      "8.23924219623\n",
      "20.7548789226\n",
      "10.2734758221\n",
      "27.6477604493\n",
      "25.937779413\n",
      "nan\n",
      "nan\n",
      "nan\n",
      "61.4431415407\n",
      "29.9670701291\n",
      "18.2757823484\n",
      "81.7094506843\n",
      "36.0672537044\n",
      "30.6529078913\n",
      "6.72417677862\n",
      "nan\n",
      "8.42303552961\n",
      "21.6653397101\n",
      "20.6380455189\n",
      "13.5980398166\n",
      "28.5424216977\n",
      "nan\n",
      "19.0292293652\n",
      "27.0393242151\n",
      "26.2492051421\n",
      "15.2507982297\n",
      "50.1387835279\n",
      "nan\n",
      "12.1890198226\n",
      "18.5865161684\n",
      "18.0344313704\n",
      "48.114697771\n",
      "31.5896588295\n",
      "16.4226704709\n",
      "10.3025330168\n",
      "14.8473303664\n",
      "4.28283232277\n",
      "nan\n",
      "27.3485558701\n",
      "53.7988794386\n",
      "22.1124621632\n",
      "nan\n",
      "3.5302301756\n",
      "11.0437654218\n",
      "32.9620562511\n",
      "12.8989672248\n",
      "26.996724755\n",
      "24.6034257986\n",
      "13.6768870668\n",
      "48.9500519776\n",
      "50.2636695469\n",
      "55.3693816358\n",
      "9.86260331224\n",
      "19.9844394201\n",
      "0.97992017688\n",
      "32.7216771748\n",
      "31.694192471\n",
      "14.6373340161\n",
      "1.04352491745\n",
      "19.2413600395\n",
      "59.3833768612\n"
     ]
    }
   ],
   "source": [
    "list_gene = ['AHR', 'ANXA5', 'ARPC5', 'ATP6AP1', 'C1orf54', 'CAPG', 'CD1D', 'CD33', 'CD63', 'CDK2AP1', 'CLEC10A', 'CLIC2', 'COTL1', 'COX8A', 'CSF3R', 'CST3', 'CSTB', 'CTSH', 'CUEDC2', 'CYFIP1', 'DPYD', 'DPYSL2', 'EFHD2', 'FBP1', 'FCER1A', 'FLT3', 'GABARAP', 'GNAI2', 'GNG5', 'GSTO1', 'GSTP1', 'IL18', 'ILK', 'IMPA2', 'ITGAX', 'ITGB2', 'LCP1', 'LMO2', 'MAP2K1', 'NDFIP1', 'NUP93', 'PADI2', 'PAK1', 'PFN1', 'PRCP', 'PSTPIP2', 'PTGS2', 'PTTG1IP', 'PYCARD', 'RAB11FIP1', 'RAB32', 'RHOG', 'RPS6KA4', 'S100A10', 'S100A4', 'S100A6', 'SEC13', 'SEMA4A', 'SLC30A1', 'TGFBI', 'TIMP1', 'TOB1', 'TRIB1', 'UQCRFS1', 'VIM']\n",
    "print len(list_gene)\n",
    "for a in list_gene:\n",
    "    print signed_p[np.where(gene_symbols == a)[0][0]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:11: RuntimeWarning: invalid value encountered in greater_equal\n",
      "  # This is added back by InteractiveShellApp.init_path()\n",
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:12: RuntimeWarning: invalid value encountered in less_equal\n",
      "  if sys.path[0] == '':\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BDC_Detected  1765  genes with MAST\n",
      "BDC_Detected  879  genes with microarray\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:9: RuntimeWarning: invalid value encountered in log10\n",
      "  if __name__ == '__main__':\n",
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:9: RuntimeWarning: invalid value encountered in sign\n",
      "  if __name__ == '__main__':\n",
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:9: RuntimeWarning: invalid value encountered in multiply\n",
      "  if __name__ == '__main__':\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BDC_Detected  340  genes with DESeq\n",
      "BDC_Detected  879 "
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:24: RuntimeWarning: invalid value encountered in less_equal\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " genes with microarray\n",
      "BDC_Detected  1143  genes with edgeR\n",
      "BDC_Detected  879  genes with microarray\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:9: RuntimeWarning: divide by zero encountered in log10\n",
      "  if __name__ == '__main__':\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWAAAAFgCAYAAACFYaNMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJztnXuYHGWV/z9nJpMwSZBJSFAYCARlw8IGEhgBF9YlgKAg\nMAIaWHEVXXFdb4Cbn0FYCYhLNMviKrsiKqgrl4CBGAQJKBeVNWBCEjBKBBISGBASkgGSGTK38/uj\nqiY1PVXV1d116e45n+eZZ7qrqqtO9eXUW+c953tEVTEMwzCypyFvAwzDMEYq5oANwzBywhywYRhG\nTpgDNgzDyAlzwIZhGDlhDtgwDCMnzAEbhmHkRO4OWERuEJFXROQPvmUTReR+EXna/T8h5LUfdbd5\nWkQ+mp3VhmEYlZO7AwZ+CLy3YNlc4FeqegDwK/f5EERkInAZcCRwBHBZmKM2DMOoRkblbYCq/lpE\n9itYfDpwrPv4R8BDwJcKtjkJuF9VtwCIyP04jvyWqONNmjRJ99uv8HBGXbF2rfN/2rR87SikWu0y\nEmfFihWbVXVyse1yd8AhvFVVX3If/wV4a8A2rcDzvucvuMuGISLnA+cDTJkyheXLlydoqlF1HHus\n8/+hh/K0YjjVapeROCKyIc521RCCiEQdsYqKBCtU9XpVbVPVtsmTi16UDMMwMqFaHfDLIrIngPv/\nlYBtOoB9fM/3dpcZhmHUBNXqgJcAXlbDR4GfBWyzFDhRRCa4k28nussMwzBqgtwdsIjcAvwOmCYi\nL4jIJ4D5wHtE5GngBPc5ItImIt8HcCffvgr83v27wpuQMwzDqAVyn4RT1XNCVh0fsO1y4J98z28A\nbkjJNMMwjFTJfQRsGIYxUjEHbBiGkRPmgA3DMHLCHLBhGEZO5D4JZxiGEcTilR0sWLqWFzu72aul\nmTknTaN9ZmCxa81iDtgwjKpj8coOLr7jSbp7+wHo6Ozm4jueBKgrJ2whCMMwqo4FS9cOOl+P7t5+\nFixdm5NF6WAO2DCMquPFzu6Sltcq5oANw6g69mppLml5rWIO2DCMqmPOSdNobmocsqy5qZE5J9WX\nlrJNwhmGUXV4E22WBWEYhpED7TNb687hFmIhCMMwjJwwB2wYhpET5oANwzBywhywYRhGTpgDNgzD\nyAlzwIZhGDlhDtgwDCMnzAEbhmHkhBViGIZhBJCFHrE5YMMwjAKy0iM2B2wYNc5I6ByRNVF6xOaA\nDcMARk7niKzJSo/YJuEMo4YZKZ0jsiYrPWJzwIZRw+TZOWLxyg6Onv8AU+fezdHzH2Dxyo7Uj5kV\nWekRWwjCMGqYvVqa6Qhwtml3jqj30EdWesRV6YBFZBqw0Ldof+ArqvpN3zbHAj8D1ruL7lDVKzIz\n0jCqgDknTRviCCGbzhFZTVLlSRZ6xFXpgFV1LTADQEQagQ7gzoBNf6Oq78/SNsOoJvLqHDFSmmam\nTVU64AKOB55V1Q15G2IY1UgenSPyCn3UG7UwCXc2cEvIuneJyGoR+YWIHBy2AxE5X0SWi8jyTZs2\npWOlYYwgRkrTzLSpagcsIqOB04DbA1Y/DuyrqocC3wYWh+1HVa9X1TZVbZs8eXI6xhrGCKJ9ZitX\nnTGd1pZmBGhtaeaqM6bXTfw3K6o9BPE+4HFVfblwhaq+7nt8j4j8j4hMUtXNmVpoGCOUem+aaVoQ\ncA4h4QcReRvwsqqqiByBM5p/NUvjDGOkk4STqsZS6qzS7Ko2BCEi44D3AHf4lv2ziPyz+/Qs4A8i\nshr4FnC2qmr2lhrGyMRzUh2d3Sg7nVQpBRlJ7CMNsqowrNoRsKpuB3YvWHad7/G1wLVZ22UYhkMS\nucBp5hNXMrI2LQjDMKqaJJxUWo6u0pG1aUEYRh2yeduOutFPSMJJpeXoKg0hZJVmZw7YMGJSqfjM\n5m07WLdpe9XFO8slCSeVlqOrdGTdPrOVMw9vpVEEgEYRzjw8+awPc8CGEYMkJos2bulmoGCeuFal\nI734andv/6CTKicXOK184kpH1otXdrBoRQf97ufVr8qiFR2JXyyrdhLOMKqJJCaLevr6A5fXmn5C\nYYpWv+rgqLUcx5lGPnGlIkVZiQ3ZCNgwYpDEZNHoUY2ByxtEaioMUWl8NQsd4UpH1lllQdgI2DBi\nkIT4zJSJzazbtH3Y8n7VmtLSrcQ5ZakjXO7IevHKDhAgoKqgZWxT5Yb5iOWARWS6qj6Z6JENo4ZI\nQnd30vgxgDOh0x8SC67ECcXNe6208qySi1Gx0XPeFXHeBSKspCvpUq+4IYj/EZHHRORfRGS3ZE0w\njOonqcmiSePHDJuI86jk9jbuJGESk4mVZC6EnaNnR94ZIkEXCD+d3b2JHi/WCFhV/05EDgA+DqwQ\nkceAG1X1/kStMYwqJqnJojS0dONOGiUxuVSJCHzYuTeKVEWHjWIXQS/jIylix4BV9WkRuRRYjqO9\nMFNEBPiyqt4R/WrDMDzSaCMUNy6b1ORSuRejsHMPG3VmnSESdoHwKAwdVUqsEISIHCIi1wB/Ao4D\nTlXVv3YfX5OoRYZR56SR+xo37zWrEtswws69NWG7ys20CAqvpEncEfC3ge/jjHYHLw+q+qI7KjYM\nowSSzn2NO6rOq4mnn7BzL9WusMnEsEyL5Ru28OBTmyLDJt7zCxauSuJUixI3Bvz3Eev+NzlzDMMo\nh7hx2byaeEJ09kWpdkU52VsefT4wy+SmZRsHM8ui0t/aZ7Zy0W2rGAiINjQkGwKOnYZ2AHAVcBCw\ni7dcVfdP1hzDqE2yFBUPO1bcUXWlo+9i5xq0Hiia/xtmV9D+wiYT/U62kMLlUZN8jUKgA27MwwED\nNwKX4cR7ZwHnYVV0hgFkW1yQ5bHKOX7Y+jGjGsrKcgjbX9ikXalTZGGTfL0DwduHLS+XuE60WVV/\nBYiqblDVecApyZpiGLVJVt0Tsj5WOccPWx+WP1ssyyFsf6Wmg4VtXTjJ503eZUVcB7xDRBqAp0Xk\nsyLyAWB8inYZRs2QlW5A1D47Orsz0Rgudq6lnnOxLIew/XkCQH7CnGyjCB8+akrR4hF/kUpWxHXA\nXwDGAp8HDgfOBT6allGGUUtkmdoVtc8sKsiKnWvY+gljm8qqngvbn6fP609nC3OyV3/oUK5sn140\n9a9YFVwaFHXAItIIzFbVbar6gqqep6pnquqyDOwzjKonLHe0q6cvcUcYJ081KCSRlAJZsTLksPWX\nnXpwWbnPYefr6fPOOWka6+efwiNzjyvqZNtntvLI3OMGty88dh6yoEUn4VS1X0SOycIYw6hFvB/y\nvCVrhsQ6t3b1Dpmg2rxtBxu3dHPm3LvLzpQoTNcKm3TyO5MkJ+6KpYvFXV/q8b542+pYAkaVZHgU\nq4JLg7hZECtFZAlwOzCop2clyIbh0D6zlQVL1w6bbPKPRvfatJ0B1SGhAu+1pR7Le83R8x8oqiuR\ntLh4MSeXdJFJ+8xWLgwpjEhy1BpUpJI2cWPAuwCv4pYhu3/vT8sow6hFoiaoFixdm0o7ojjKZGGj\nulrqxJFFnN1fJp0VcWPAT7ixX//fxzOwzzBqhignkVamRDFdiahY727NyYqLp0lWXYq9OHFWxI0B\nn4OJ7hhGZBVYlM5C2Eg3iRFc1C1/1Ag7YWXFSCqtFMyzhDpN4saAHxGRa4GFDI0BP56KVYZRhRSb\nzCrmJBq+I0PCEEF5qEk7mKgRdmdX5eLicWxOahIwjeadYbxlTCOv7xgeC37LmGSV0uI64Bnu/yt8\nyxQnJpwaIvIc8AbQD/SpalvBegH+CzgZ6AI+ZhcFIy3iTGaFOYn2ma1snjyOjVu6ERjmrNIqMY6a\n2a909B3X5qw6DCdJkPONWl4ucdXQZiV61NKYpaqbQ9a9DzjA/TsS+I773zASp9I47qTxY5g0fgzr\n5w+v4k/LSc05aRpzbl9Nb4GyTFOjVBw/jWtzlpWCtUbsjhgicgpwMEPV0K4If0UmnA78WFUVWCYi\nLSKyp6q+lLNdRh2SRishjzQn6WBojvKEsU1cdurBqYU3Ojq7OXr+A4Mj/DTfN8hWiS5p4spRXodT\nijwLR5j9LOCxFO3yUOA+EVHgu6p6fcH6VuB53/MX3GVDHLCInA+cDzBlypT0rDXqmjTFzNN0UmnF\nTqPCG/5wRJrvW97qcJUSNw/4b1X1H4Gtqno58C7gr9Iza5BjVPUwnFDDZ0Tk3eXsRFWvV9U2VW2b\nPHlyshYaI4Y0Wgl5ZJVmlSTFyqL94Yi03re81eEqJW4IwrvMdYnIXjhFGXumY9JOVLXD/f+KiNwJ\nHAH82rdJB7CP7/ne7jLDSIVSR5P+2+NFGzuZMrGZSSH7hdpKs/LbXKzYI61ReK3Hl+M64J+LSAuw\nAHgcJzTwvdSsAkRkHNCgqm+4j09kaBYGwBLgsyJyK87k22sW/zWqhcLb456+ftZt2s5vV3aEZkpk\n5XCTipt6NscpiU6DtOPLaRM3C+Kr7sNFIvJzYBdVfS09swB4K3Cnk2nGKOBmVb1XRP7Ztek64B6c\nFLRncNLQzkvZJsOITdDt8YAqX7xtNRcuXMVuzU2IOPm4pTjBSp1nknHTSxc/GdiDDbIJoVRDk9FK\niJ0F4ePbqnp+4pYUoKrrgEMDll/ne6zAZ9K2xTDKIUpMHBgi3BPXCSbhPJNKebt08ZP8ZNnGwHWt\nMXvFVTrir8XQjZ9yHHBb8U0MwyhV3jCOEwxznt6oOo4DSipuesujzwcubxQZpqeQZrZClqGbpCnH\nAb+SuBWGUQOUOoIrR96w0AkWHjPMoXuj6o7Obub8dDWw07EV7mO35qbAHm2lxk2Dwg5hy2uxGi4L\nSnLAIjIeJwfYMEYU5YzgCm+Pna5l0X17/U4w6JjF9wC9/crld60J7VLc1Cg0NciQ6jhheAFFMRpF\nAp1tUMPMas5WCLqwZkWsPGARmS4iK4E1wBoRWSEif5OuaYZRPZSbb+pvg/OOPcbRECFBVjh5NG/J\nmmHHVMKbT/rZ6grtBNnd26+M32XUoO6t36mX0lPunCP3ib08y755peBvxFkolB9E0gJycQsxvgtc\npKr7quq+wBeBwqo0w6hbkhjBTRo/hv0njxssSGhpbmLC2KZQHd+wVu7qbh/HGYTZ19nVyyNzj6O1\npXnYiDpuIcOV7dM596gpgyPeRhHOPWoKV7ZPH7ZtloUmpfS/C7uwhlHs7qNU4oYgxqnqg4NGqD7k\n5uYaRt3ivzVtCLndLnUEN2n8mFiC31EOsLWleXAfMy6/L9BRt7hi68XyZKP0HOJwZfv0QIdbSFbZ\nCqWGivIOgcR1wOtE5N+A/3WfnwusS8ckw8ifwh9y3DzXJHJ0oyrLgKFhitMOHqZ21tQgzDvtYBav\n7KCrpy/S7rAJOc+WIG3fcs8viWyFYscvdbIvj0acfuI64I8DlwN34IzCf+MuM4y6JOiHDM5t9oBq\n4I9/8coO5vx0Nb394RkJ3nZBTqTQ6QfRIHDhwlUsWLqWWQdO5sGnNtE7oIMTYq2+SaSwfY0ZtTPy\n2Ns/EHqsi+94Yoidsw6czMLHnh909h2d3cy5ffj5pUWx0e3ilR0l978LylQpnKBMk6IO2O0Jd4mq\nfj4DewyjKgj7wQ6oBur5Alx+15pB5+vhz0gA2LxtxzAnMuf21Vx+15rBibMoPL/Q0dk9pAiiX3Vw\nZOuVBoc58s7uXi6+40mWb9jC9p5wZ9/dOzDo0AqPN3h+A8qX73iiIgccd1RdbCI0avIsLFQUFBrp\n6ukL/Sx2aUx2Gi5uT7hjEj2qYVQ5pWgMeA4k7EfrX75xS/fwrIQBjeV8i+G/1S4W2+zu7Q+tYiuV\nrt4BLl38ZGgsOMrBlhKzjYpXB2WMeHgXJr8dLWObUIXXuoeXgU+de3foub7Zn+zIOG4WxEoRWSIi\nHxGRM7y/RC0xjCoi7qy9P40pDj19yba0KcRzUlmnd920bGNgtkFYmpe3bSnpfVHnFBbHBrjqDOfC\n4Ldja1cvnd29gzbNuX01M6+4z3G+GTYrjeuAd8GRoDwOONX9e39aRhlG3sTVsA2LFftpGdL+Pd1f\nt+ek5pw0LUs/ghKcuVHMwZaS3ldMfziI1pZm2me2Fv2cvLsQBUIK/FIhrhqaqYwZI47CWXsvv9R/\nKx0njWneaQf7nqX36/b3eWuf2coFC1clu/8ik1P+96JYNod/pB431ON9FnHPy3/Hkne6WRhxK+F+\n5OoBe88niMgN6ZllGNVF2K10y9imyNeNG904OEP/+MbOko4ZVNIbRVODsGDp2sEChKRGwN4dwIIP\nHsq5R4W39PKcZpywjH+kXkqBRvvM1sEKvkImjG0KvWPJu+IujLhpaIeo6uC3R1W3isjMlGwyjKoj\n7FZ6zKgGmpsaQ29vu3v7uXTxkyxa0cGNJcZ/w8RuwujqHaDLl7WQBC3NTay67MTB5+0zW1m/aRuP\nPLtlyHZ+p1nsdt+/bTkFGmEawFGNRssRRsqCuA64QUQmqOpWABGZWMJrDaPmCbuFfa27l2tmzwi9\n3R5QuPnRjWSUVpo4Q8MnuCP5ob0YBDjz8J3hmqjb/SCd4KgCjagMilKcduFr/FkQuzU3sb2nb1gK\nYRbEdaJXA78Tkdtx3u+zgK+lZpVhVBlRscpiscladb4exUqyFXjwqU2Dz8PeK38JddzjRqWolZp7\nHMfRZ10VFysGrKo/Bs4AXgb+Apyhqv8b/SrDqB+iYpWeo6hGJhSJURdjzk9XM+f21YOx77CwSEdn\n92BqWVLCO2l1PA4S6/FU68Liyx5jm+ImjsUjcgQsIm9R1dfdkMNfgJt96yaq6pbwVxtG/RB22wvw\nxdtWlxyvzQpVxwkHFXqE6fn6KeW2vLCAolLhnajCi8JsFG+is9gxF6/sGKKdUVhOXSxW3Jfw7Uyx\nEMTNOPm+K9gpRer/v3+i1hhGFVPoWOYtWcP2nr6qdb7gFCg0NQhNjTLEmQrOaDaOwHtc/JV4SQjv\nhIUyPPF42BmWWL5hC4tWdBStqJu3ZM2wVLreAWXekjVDbA4LJ/UkHCeOdMCq+n73/9REj2oYNUhh\nTDKq+gqcEeauu4wqul3Ya8Mce6nregeU5qYGBgZ2hhC8LfyjqTgj4mJUmm9bWC4c1LkjSL84qDOz\n1ysPdjrhsM/CvzyNHOowYgc0ROQQETnNSpGNkUqcqjc/5xy5D/NOO7ik6q1vzp7Bc/NP4dmrTg7N\n4+1Xpalh+NqmxnAH2t07ELrOE3i/+kPDmpAH0tzUWFDdt5NK8m0Lc623dvWCOKlwXm5v2OUhqj9d\n3A4feRC3EOMG4AbgTKwU2RghFE7WlDpD7mUG7BJz4qbVl1EB4c7MK4rwO8EJY5tYcNahoY6xGB2d\n3bTPbKU5hq2HTdmN9x+6Z+C6WQdOLuv4EN4+adyYUayffwqPzD2urElF/8Rd2Osrnawsl7hpaEep\n6kGpWmIYVUS5DTH9dHR2x76V9eKab7/4nkFd31kHTh4S1wSGSE4GxVgvv2tNCRYOPf6li5+MNcn0\nf89u4Y8vvRG4zp+OVirFdCEWr+xg25vDBebj4E3cBU1GNjUKl516cMCr0ieuA/6diBykqn9M1RrD\nqBKCRmNpTbX5Hbu/vfyiFR2ceXgrDz61aVjmRVAWADi93spBITCOGrZtmHxmoRMtlpkQp+3Tbu6o\nfsHStRUJpQfdwTSKMPud+2QiKB9EXAf8Yxwn/BdgB+53RlUPSc0yw8iRrMRbWiNa4nT39vPgU5uG\nFC8EjcwvWLiKS+58kqbGhoouEklkc/jDJnE6WBRr+wSwvaePxSs7UvlM+lVZtKKDtn0n5uKE407C\n/QD4CPBedsZ/T03LKBHZR0QeFJE/isgaEflCwDbHishrIrLK/ftKWvYYI4+w+GupAjlRNIoUdSqF\nzjlsInB7T39Z2RaF9lRCU4MMKbYoVkgRd1Kzt19ZsHRtaoI6SRR3lEtcB7xJVZeo6npV3eD9pWhX\nH/BFN+58FPAZEQmKQf9GVWe4f1ekaI8xwgir5jrnyH1K1qQNo9/tLRdFoVNMa2TunVscFxyVnXHh\nwlWD1WXFYrqlnMuLnd2Bn0ljQDZIOeQlVxk3BLFSRG4G7sIJQQCgqnekYZSqvgS85D5+Q0T+BLQC\nFoM2MiGq8u3nq19KRFWrUaTo5F6/KlPn3j14/KS6+I5taqCr12nI2dLcxLzTdiqJRbUqigqZ+PvV\nXXzHk6Edl72LTinn0iDChQtXsVtzE7s0NdDZtbOV0PINWypur5SXXGXcEXAzjuM9kYzT0ERkP2Am\n8GjA6neJyGoR+YWIhE5jisj5IrJcRJZv2lT+LK0xsvD0AbwUKHDKbf1OpalBGDe6vBFxYVFEGH79\n4UrSvDwEBp0vOEUIF7gj17Z9J0a+Lo5eAji39SJEakIEjWi9qr1C+lVR19Y3ewe4ZvYMHpl7HO0z\nW2nbd2JgXnRcytGpSIqq7oghIuOBRcAFqvp6werHgX1VdZuInAwsBg4I2o+qXg9cD9DW1la9daNG\nVROYpzqgDPSGt3ZPku7efm56tPJGmmE/gEGR+SIj17jaup1dvXz4qCmD2RWNIkNkK6PuMqIyI7p7\n+5m3ZM3gNkhwGyHPJRcbaQe1msqKqtX0FZEmHOd7U1Cow++QVfUeEfkfEZmkqpuztNMYOYTFCbPU\ngkj7UN29/ezSNFxkPkpEPSx9rGVsE4tWdAyuC8o4CMtn9rIkwvKoO7t7d14kIt6T9fNPAQgtpCks\nfsmaqnTAIiI4mRd/UtX/DNnmbcDLqqoicgROOOXVDM00Rghermot3DolIa7T2bVTZD4sf9fvOAvT\nycBx2KoEZkEU6jN4+/Afb9aBk1n4++crOo8GkUGpyVkHTg6ME3d29QyJsWftjKvSAQNH46S9PSki\n3iXwy8AUAFW9DkcU/tMi0gd0A2erVrEslVGTXLr4SW5atjHUqTVI9QiuNzUIs4/Yhwef2lTRRF2D\nm3nxyNzjBh3jBQtXDcpuFna1aJ/ZyvINW4aFGm4KmRjz9Bm81wblC1c6qVZ4nLAKve09wTnKYZON\nceLfpRDLAYvIW4F/B/ZS1fe5KWHvUtUfJGqNi6r+liL9u1X1WuDaNI5vGOCMyqKcL1SP8wUnHv2T\nZRtpbWnmm7NnAHDhwlUlj4g9x1Uo8eiv0rtw4SqWb9jCle3TWbyyIzDUEJYFAc5I2CubTlNP2cvx\njZNm5pfTnHPSNC5auAp/dL8BEp+si5sF8UNgKbCX+/zPwAWJWmIYVUathB0K8Y/myrW/u7efnyzb\nGDrRpsBNyzYOjpCDQg1BWRB+tnb1Muen6YvZe2GNuNsCLN+whcKp1QF3eZLEdcCTVPU21wZUtQ+o\nrvaihpEwWfcHS5Lu3n4uWLgqVnVbuQlcijN6DXuftnb1Fs2UyKIR5m7NTXT1xBPx8Rx1WLZJElko\nfuI64O0isjvuBVVEjgJei36JYdQu1aofWypxRpdx5TLL3X+eNDUI23v6hokHNTc1DMs39md6hJ1W\n0qcb952/CFgCvF1EHsER5/lcsqYYRvWQlzZAHnRnlMecFZ5bbW1pZvwuowJH2RPHjWHBWYfS2tI8\nKPaeRz5w3EKMx0Xk74FpOOe3VlUrU/4wjComL22AkUZhr7pCWlua6QoYwfppbmpgl6bGIeXJniOd\nOvfuwNe86ArQ55kDDPGzIP6xYNFhIuK1qzeMuiMpzYWsqKZ0uFJYcNahLFi6NlATwwsJXFhE1H7i\nuDFDJDv9hH2OeWk/FBI3BPFO39/fAfOA01KyyTByJy9tgHKpRecLO/U2npt/CtfMnhEYEijmLKPu\nVsJU7arl840bghgS7xWRFuDWVCwyjCqgfWYrF9/xRN3FR5NiwtimyLBAHAoTNMJCAsW0J6IcdJje\nRN6hB49yK+G2A9aq3qhrzjx870QqsuqRUw7Zk7Z9J3L5XWsGHXFLcxPvP3TPYX3swvjwkVNiHctz\nlv5jecQZzcaJ9RaWQmdF3BjwXewMzzQABwG3pWWUYeSNV91lBHPLo89zZXtw1kDbvhO56LZVoWGR\nRhHOOXIfrmyfDoT3jfOWd3R20+gK/rQ0NyFC4IRbuQSVQmeFxJFPcDMgPPqADar6QmpWpUjbrrvq\n8sMPz9sMI01WuZM2M2aUvYvHN3bS05dsrdFBr6wD4I977J/ofvPiqP13D123/Lmt9A0Eh2/escd4\nJo0fA8DmbTtYt2k7Az4/1CDC5F3HsOmNHUOWFzJ6VCNTJjYP7svP5m072Lilm56+/sjtoPTPOuq8\nPeThh1eoalux7eLGgB+Os51h1AtJO9/6I7x+bvO2HaHOF2Ddpu0ATBo/ho1buoc52QFVXn59B8UK\nqXv6+nl203ae29xF38AAo0c1MmFsE5u27WDAN/zu6esfcsyg/eRFpAMWkTcIfhe8rshvScWqNJk2\nDR56KG8rjDQ59ljnfwWf86evuK/iSaZCbr15LgBn/8P8RPebF4WqaOCox8WJmwtwzewZZYkFlUtr\nS/Ngupo/vMFRpe3nOVdjOJKYDU4jHbCq7hprL4ZRZ1R5hW1VENRmPkyCshAF5ty+mpaQbIrGEJH3\nSvDS1YL0i/OipCJwEdlDRKZ4f2kZZRh581qFLd5HCoVt5ktxmb0DimqwYtroUcG94SrBy24IUm/L\ni1gOWEROE5GngfXAw8BzwC9StMswcmW35qa8TagZymkz79HZ3ctVZ0xnwtih73d37wD9CSqlCTuL\na6qpzDzuCPirOJGSP6vqVOB4YFlqVhlGzsQM4dUlpXYY9reZL4f2ma2MHT08GlqsBGbC2KbYtn74\nqCmDsepidkZpGCdNXAfcq6qvAg0i0qCqDwJFUywMo1ZJegKulhi/y6hYOsLgOOuoNvNxKXVU2trS\nzMqvnMiCDw5VNDv3qClDRtMtzU18c/aMwZzjMDu9s20UyTQ8EbcSrtNtEf9r4CYReQWnGs4w6pIk\nmlvWKiVdfHx+Om635CDCJuOCKOzQXFiI4Xe2QQSVJ886cHLsCr4kieuAT8dpfHkh8GFgN+CKtIwy\njLwZqc63VHr7lQsWrmLB0rWDKWn+bslhbeUL2fZmPOfbKJKKbu/PV7+Uy8RcXAf8KWChqnYAP0rR\nHsPIlcWNTOkaAAAgAElEQVQrOwabRRrx8VLSlm/YwoNPbRocWTY3NcQSNIqjedTc1JiI882z9LiQ\nuA54V+A+EdkCLARuV9WX0zPLMLJn8coO5vx0dSZ9yuqR7t7+IV2kOzq7aWqUolrFcds/jRlVfusk\nP5WkoR399omJ2OARtxT5cuByETkEmA08LCIvqOoJiVpjGDmyYOlac74VUvju9fZr0YySuHccnd29\nQwo/wggT9/EoNw3tgD3GcdMn31XWa8Mo9ZLyCvAX4FVgj0QtMYycqab80Hqi2DxcKZN+/sKPILzw\nQkdnN8rO0Ih/lB2WhjZhbNNgRkVLc9OwQpAXtr6ZeLPWuIUY/yIiDwG/AnYHPqmqhyRqiWHkTLW0\nqak3SkwrLkrUhTIovFDotMO6ZFx26sE8Mvc41s8/hXFjhjfzLOb8yyHuCHgf4AJVPVhV56nqH/0r\nRWRColYZRg7MOnBy3ibUJeW2SyqsjvNoEGHq3Ls5ev4Dw0akYc7Zv7x9ZitnHt46mOvcKMKZhw9N\nZwubmEt6wi6WA1bVi1U1Kp/kVwnZYxi5YALs1UVrSzOnHLJnoOhlv2rJ4QX/cu+z9nKU+1VZtKIj\n8fBCHJKZVowSBy13hyLvFZG1IvKMiMwNWD9GRBa66x8Vkf2StsEYOVSTQMtIoLWlmdYQZykwWBhR\nbPDc3dvPvCVrOHr+A0ydezddPX3DypML2xbFCVNkRVIOONGpYxFpBP4beB9O+6NzROSggs0+AWxV\n1XcA1wBfT9IGY2RhE3DZ8mJnd2hJ8IePmsKDT22KfUHs7O4dnHTb2tUL4kyiFXZX9h87iI7O7sCw\nRpqU25QzbY4AnlHVdQAicitONZ4/9nw6MM99/FPgWhERjdNjyTAKKKUU1qicvVqaIzsWT517d9n7\n7u1Xxo0ZxarLTgw9dlSM10t1y4KkHHDSIYhW4Hnf8xeAI8O2UdU+EXkNJ0Nj8zDjRM4HzgeYMsVk\njI2hLF7ZwbY3+/I2Y8TghRggvGNxlJOMQ9QdTbE291mGomKHIETkMBH5vIh8TkQOK1h9fMJ2JYqq\nXq+qbaraNnmyzXQbQ1mwdC295U7VGyWjUHTSqxJlNYhOKWyf2cpVZ0wPjUFnSdw84K/gaEDsDkwC\nbhSRS731qrolYbs6cFLfPPZ2lwVuIyKjcASCXk3YDmMEkKcWQK0ybnQjgpMq5sVbS6G7tz+yAs5z\nkmGpaFEUTrqF7f+Rucfl7oTjhiA+DByqqm8CiMh8YBVwZUp2/R44QESm4jjas4F/KNhmCfBR4HfA\nWcADFv81SmXxyo4RJT159Nsn8tyr3YMx1/12b+b/nt0S+/xbmptCY6tHz3+gpIvZ1q5eFq/sCC0r\n9sIT/tLilrFNbHuzb8gdS1ODMH6XUXR29QaWHkcRFI5obmrMLAwR1wG/COwCvOk+H8PwEWliuDHd\nzwJLgUbgBlVdIyJXAMtVdQnwA+B/ReQZYAuOkzaMkii1j1mts2zdVq7+0KFDHNTilR3MW7KGzhh9\n8Dq7w51mmDPbpakhdIJzwdK1RZ1lYZy4mNZDKYRNBMaV0ayUuA74NWCNiNyPM1h4D/CYiHwLQFU/\nn7RhqnoPcE/Bsq/4Hr8JfDDp4xq1R+EP8mfbdjBp/JhY29eD820QEBH6Y8Sx+1WHCdoUjjSLjWLD\nBHHCnBkQ6tDKSf8Lm7iLIsppB+2v2hzwne6fx0PJm2IYpROk7bpuk9Os5bcBPzqgalqSV4IXNml1\nzyvuCBZ2Fh0EOVC/mHrY+xT2+sJ9+AmzLwv9jaDvSBxVtSyIK0dpIuxGVRJU1TSgynOvdgX+6MaM\naqh55xsUh72wxBFbsZGn55jCRoKlTlzOO+3gwPBEscmyJIiqfAtzwGHzAknn2yZVCWcYuRDmSPr6\nBwJ/dHFHidVKc1Mj8047eNjyUkeScbZvn9ka2pwzbtNO/7681K+wCrW0iCPQU0hYMCfpkJU5YKOm\nGUkSkkGqXR5hEovnHjUlcHnckWdYU824zTY9kpw4K5U4Aj2FhKWnJZ22Zg7YqGmCHE+DCKMagr/a\nE8Y2VZTgnyd+1a7FKzsGBWiOnv8AwJDiAq+9+oNPbeLMw1vLHnkm4YjiiKSnSdjFKeoiVM5ryiEy\nBiwidxEx6lbV0xK1xjBKJGjmff/J44Dh+Zye6DaETwpVO14Bw5u9A8Pi21edMX1YKlhHZzeLVnSU\nfbs/56Rpw/rkNTVKSY6onBhskkRpToSNzNtntrJ8wxZuefR5+lUj7z4qodgk3H+4/88A3gb8xH1+\nDmBNOY2qYNjM+71OCtpVZ0wPve1dsHRtTTpgCG7h45dTTNzZFQ7BSgyElhODTZqg7Iyo7AggUDO4\nbd+JiTrhSAesqg8DiMjVqtrmW3WXiCxPzArDSIGofNF6lJ+MOqdyzzdIJ6N3QEty6GHCOnnH74vp\nAmcxao8bAx4nIvt7T9wS4XGJWWEYGZP3j78SwvIP9mppLmvCKYokRq9ZxVNLJercshq1x3XAFwIP\nichDIvIw8CBwQaKWGEaGBDmFwi64WVBOw0qFUIeWlLPzJvnCog2lOPQ8U9CiiLpYtYSIAIUtL5e4\nhRj3isgBwIHuoqdUdUeilhhGhoRNzFx426qibdSToqlBmH3EPiz8/fPDOvBG4VW/RaV1VZLyFVUF\nB+U59HLKh5MibKItTLvCqywMIunvhsQREBORscBFwL6q+knXGU9T1Z8na076tO26qy4//PC8zTBK\nZPO2HWzc0k1PXz+jRzUyZWJzuN7DKrd6a8aMko+zbF16iqYHvbIOgD/u4UTz3rHH+MFzCmJUYwP9\nA4r/N9ogwv6Tx0VqXVTK4xs7Q20q+t5XGZu37WDdpu0MhLyHYd+rqO/BUfvvXvS48vDDKwrmzQKJ\nqwVxI7ACeJf7vAO4Hag5B2zUHoU/op6+/kG9h6QdwajGBvr6BxLdZxANDcKk8WN45pVtodu07Tuh\ntAtPEeLuK8z5Ahw2paWsY6dJ1Hlt3NI9xPmCU6q+cUs3k8aPGfwrZPSoxsD3YfSoZHPI4zrgt6vq\nbBE5B0BVu0RKrEWsFqZNg4ceytsKowROD9GZbW1p5pG5xw1/wbHHOv9L/JwXr+xgzu2rU+uOcevN\nTnPvs/9hPgKsn38KXww5t0aRQdnISQkcOyis0NzUGBiLDbMp9P3OkWLndebcu0M1HdbPPyV0vxsj\n9ntYnFBKTPcYdxKuR0SacTMAReTtgMWAjUzIakY6y9ZE3lHCWu94spFJVYuV0oq9WrMWgih2XuVm\nhWQ1cRh3BDwPuBfYR0RuAo4GzkvUEsMIIe080rg6uEniidl4P+gv3rZ6mL5CknmnpVzEoirHqo1i\n5xU10VaMLCYO42ZB3CciK4CjcEbvX1DVYd2HDSMNKvkRFaPYjH9anHPkzpaH7TNbQ+Ukkxrll3oR\nyzNroRSKnVe1X0xiOWAR+ZWqHg/cHbDMMFIlzR9R0C1s2owb3ciV7dOHLEt7lJ/mRSwLykkl86jm\ni0kxMZ5dgLHAJBGZwM4inLcA1XlGRl1S2K1hwdK1XLhwVcXOOI+S5K6e4Q4/bQdZ7SPBKOJ0tKjF\n84LiI+BP4VS87YWThuY54NeBa1O0yzACCfoxXrhwFRcsXDVYoNBewv7CRp5pEjSqzcKRVPNIMIpi\namq1el5QXIznv4D/EpHPqeq3M7LJMEIJ+jF6U1feyOiYIk05/QSNPNOkQYZKOeYpVF4rVIOaWlrE\nzYIYEJEWVe0EcMMR56jq/6RnmmEMp9iPrru3fzDJPg6FI880k9C8IoEjXB3aS+58ku2+cEQ1NYus\nJqpVTS0J4uYBf9JzvgCquhX4ZDomGUY4cX50UZVcfjzBGS8D4ZrZMxJvOePxzdkzOGxKC5PGj3EK\nPn66eojz9QjLzR3J1FJecqnEdcCN/so3EWkERqdjkmGEE1a44CdOuWhYm5xZB05OyNKhFArlRInv\n1MOtdZJUq5paEsQNQdwLLBSR77rPP+UuM4xM8YcMOjq7h7UPb25ybvOLETaxc8eKFxK01qFwVF3M\nwdbDrXXS1PJEWxRxHfCXcJzup93n9wPfT8UiwyhCUEqafxJr0m+Kx3/DnGBXb7JCPEG3ysUyL+rh\n1tqIR9xKuAER+SHwgKqmGqASkQXAqUAP8Cxwnj/+7NvuOeANoB/oiyP9ZtQf5Y6MdmtuSq0nXKMI\n/ao70+IK7AtqdOlRjkC7UbvErYQ7DViAE/edKiIzgCtS6op8P3CxqvaJyNeBi3FG4EHMspJooxzS\n0vKLoxjmOeQg8fcBdTo21+PttjGcuJNwlwFHAJ0AqroKmJqGQap6n6r2uU+XAXuncRyjPvAyGabO\nvZuj5z8QWz2sM6CzcBLEnUBrn9ka2l24s7s3MRU0o7qJ64B7VfW1gmVZ6PZ9HPhFyDoF7hORFSJy\nftROROR8EVkuIss3bdqUuJFGPoRlMmzeVlwpNa2JrlL2G7WtpaKNDOI64DUi8g846WgHiMi3gf8r\n96Ai8ksR+UPA3+m+bS4B+oCbQnZzjKoeBrwP+IyIvDvseKp6vaq2qWrb5MnppBkZ2ROWybBxS/Ao\n1D9a7urpoynhgGupualR21oq2sggbhbE54BLcETYbwaWAleWe1BVPSFqvYh8DHg/cLyGNK1T1Q73\n/ysicidOiOTX5dpk1B5hTiqoEKNQQ2JrV2+iXZDDJtyiaJ/ZyuV3rWFrQDik2EjaSpjrg6IO2C26\nuEJV/xXHCaeKiLwX+H/A36tqV8g244AGVX3DfXwicEXathnVRVg6V1AhRtBouZROxFEIxHaAm7ft\n4PT5Dww6zlMO2ZNFKzpKUkGLow5m1AZFQxCq2g8ck4EtHtcCuwL3i8gqEbkOQET2EpF73G3eCvxW\nRFYDjwF3q6oVhowwwkpUgwox0rylV+LFbL3mov6Y9aIVHZx5eGtJVV6ltBcyqpu4IYiVIrIEpxPy\ndm+hqt6RtEGq+o6Q5S8CJ7uP1wGHJn3seqYeb1kLhXRaxjahCs+8so2NW7rZuLJjcJu0ZSfjOPig\nDr3dvf08+NSmkppd1rM62Egj7iTcLsCrwHE4RRKn4sRojRogLFugHlKd2me28sjc47hm9gze7B0Y\nLK7o6esfco5Bo+Ukp+AqEQkq1XGW22jSqD6KOmA3BvyEqp5X8PfxDOwzEmAk3LIWO0e/oAs41WpB\nEeBiQj9hxBHxCRMJKtVx1rM62Egjbgz4nAxsMVJiJNyyxjlHr4dYc1PjsA7E4IyID5uyW1kj40Ur\nOoreUUyZ2ExDQQleOY6zntXBRhpxY8CPiMi1wEKGxoAfT8UqI1HqWdDaI+wcG0SYOvfuwbh3VBNO\nBZat21pWhVGcFvKeSHxrS3PFsfh6VQcbacR1wDPc//5UL8WJCRtVThoNH6ttUi+stZA30vXi3sVa\nDwWNjOMS545i0vgxJU24GfVNXDW0WWkbYqRH0g0fqzEP1X+ODsMDCWn3fYtzR1GYB5z3hcvIl7hq\naLvhCPJ45b4P4xRnFOpDGFVKkresxbrUVgdZSJXsJM4dhT8PGHZ2dF6+YQtXtk/PwkyjyogbgrgB\n+APwIff5R4AbgTPSMMqobqpxUq9wVB4XT7vX+18KDQKqxBrJLl7Zwdte2U7hhUGBm5ZtpG3fibEv\nXtUW/jHKJ64Dfruqnul7frmIrErDIKP6KWVSLytnETW5FoYAz151MgBT595d8jEHFJ6bf0rR7byL\nw40ho3Kvki7O+5JU+MeceHUQtxCjW0QGy5FF5GigfnKYjJKIm4eaZQFIOaNv/wWj3IyQODrEcS4O\nce1PIqe7ngtzao24DvjTwH+LyHNuK6BrgX9OzSqjqombh5plAUiYA21taeabs2cUvWDE6bYcRBwH\nFse5xr0AJBH+GQmFObVC3CyIVcChIvIW9/nrqVplVD1xJvWyjBVHpdoVdlJuFBlWJedtc8HC8iJr\nUZOQxXQoSkkJTCKnuxpj+COVWCNgEfl3EWlR1ddV9XURmSAiZesBGyODLDULCkuNR49qHDIqD6qC\n87IQ9nPDCDC8hXwphDmwKB2KUqvYkihDNi2J6iHuJNz7VPXL3hNV3SoiJwOXpmOWUQ+kUQASxeBI\n9t7dATgsRkjEmxbznPHfvn1i2appYQ7Mc66jb2qkp6+/LPH2wn1VMoGW9edihBPXATeKyBhV3QEg\nIs3AmPTMMuqBpAtAKqXYLbYC//fsFkY3Cj0lirUXc2DtM1thSgtAxZVwleZ0V9vnMpKJ64BvAn4l\nIje6z88DfpSOSUY9UU2aBXE0gRUYN2YUuqMvdseMSka0eVFNn8tIJlYMWFW/jtMD7q/dv6+q6jfS\nNMwwkiZupkNnVy+z37lP0e2amxr55uwZPDL3OHNmRlnEHQHjtvyxtj9GzVKYDRHGXi3NLFrxQuA6\n8W1Ta6Neo/qI7YANox7w33pfuvhJblq2cUh9WlOD0NnVQ3fvQODrlXjVb4YRh7iFGIZRd1zZPp1r\nZs8YLChpaW4Cge090VVrxSrfDCMuNgI2RjT+EfHR8x8Y7CkXRTXIbxr1QVw5yicZru/3GrAcuFJV\nX03aMMNIizAhmlIqwbp7+7n8rjXmgI2KiDsC/gXQD9zsPj8bGAv8BfghTpdkw6h6otTESm1dv7Wr\nl8UrO8wJG2UT1wGfoKqH+Z4/KSKPq+phInJuGoYZRlwKR7Q/27ZjsP9aIVFCNHNOmlayFkR1idAb\ntUbcSbhGETnCeyIi7wS8hMq+xK0yjJgESSuu27Sdzdt2BG4fJUTTPrMVKbElsgnYGJUQ1wH/E/AD\nEVnvylH+APgnERkHXJWWcYZRjKAR7YAqG7cEO8ZiQjQfPnJK4PqmkF+KCdgYlRC3Eu73qjodpzvy\noap6iLtsu6relqRBIjJPRDpEZJX7d3LIdu8VkbUi8oyIzE3SBqN2CBuB9vQFp5IVUxO7sn065x41\nhUZ3KNwowrlHTWHBB4trChtGqZTVlFNE0m7KeY2q/keEPY3AfwPvAV4Afi8iS1T1jynZY1QpYRNn\no0cFlxzHEaK5sn16aJNME7AxkqRWm3IeATyjqusARORW4HTAHPAII0hasUGEKRPDQwNxhWiC0tUq\nVTIzDD/V2pTzsyLyjzh5xl9U1a0F61uB533PXwCOTNEeo0oJGtHuP3lcaBZEXJJqfmkYUcR1wN0i\ncoyq/hYqb8opIr8E3haw6hLgO8BXcQo/vgpcDXy83GO5xzsfOB9gypTgSZZ6ZSR0vx02or23cqnq\nsHS1eUus+MJIjrgO+NPAj9xYsABbgI+Ve1BVPSHOdiLyPeDnAas6AL9e4N7usrDjXQ9cD9DW1laa\n0nYNY6O48gmb3OvstuILIzniZkGsUtVDgUOA6ao6U1VXp2GQiOzpe/oBnNhzIb8HDhCRqSIyGqcy\nb0ka9tQy1v22fKLSy+z9M5IicgQsIheFLAdAVf8zBZu+ISIzcJX/gE+5x9wL+L6qnqyqfSLyWWAp\nTkHIDaq6JgVbapp66X5bShhl8coOpmzspKevny/Of6DskEtUVVytvX9G9VIsBLFrJlb4UNWPhCx/\nETjZ9/we4J6s7KpFkmhhnjelhFG8bW/sixdyiXLs7TNb+fIdT9AVoAtcS++fUd1EOmBVvTwrQ4zk\nqYfut1FhlEKnWsq2xRz74pUd9A4Mny5oapCaev+M6sYE2euY9pmtXHXG9EHB8daWZq46Y3pNTSDF\nCaMsXtnB0fMfCFUyC9pHsfj4gqVrA5ty9g4oC5auNUF2IxFMkL3OqfXut8XCKItXdjDnp6sjOxi3\njG0atizKsS9e2REpS2nZJEZS2AjYqGqCtBuaGoSunj6mzr2bi25bVbR9vAasDovj7tbcNOhco7Bs\nEiMJysqC8EgpC8IwBimsdNutuYntPX1s7XJaBwU510KC2gyFxcdFGBaaCMOyIYxKKTYC3tX9a8Mp\nxmh1//4ZOCzidYaRGO0zW3lk7nGsn38K48aMKjriLURgWMw2LD7e2VW8J5yHZUMYlRIrC0JEfg0c\npqpvuM/nAXenbp1hFFDOqFMJ7lwRFB9fsHRtrLZEtZZNYlQncWPAbwV6fM973GV1jTe7PnXu3daK\nvEood9QZ13EXc6rlZJN436Nl617l8Y2d9j0yBombBfFj4DERudN93g78KB2TqgPTUahOgmK3TY3C\nuNGjeK27l4aQnkJxHXf7zFbmLVkTGDdubWkuWY6y8HvU09dv3yNjkFgOWFW/JiK/AP7OXXSeqq5M\nz6z8KSWp3yiNShTaigmqL17ZQcPNwoBvdq65qZFZB07m6PkPDHlN2H7mnXbwMCcvwKwDJ5d8rvY9\nMqIoJQ94LPC6qt4oIpNFZKqqrk/LsLypFx2FaiOJO4uo3Ob2ma1snjyOjVu6EZyR76wDJ7NoRceQ\nYxbqPBTasXzDFm5athHPjSuwaEUHbftOLMlx2vfIiCJWDFhELgO+BFzsLmoCfpKWUdVAseaNRnlk\nodA2afwYDpvSwvr5p/DI3ON48KlNsVLLunv7ufyuNRw9/wF+4nO+ldhp3yMjiriTcB8ATgO2w6Aw\nTuZCPVlSrHmjUR5pjQj9E6aPb+wc0pa+lH1v7eqNzIIo1U77HhlRxHXAPaqqOHdiuO3o65p60FGo\nRtIYEXphjY7ObhRnomvdpu2D2QZJjjaDypqj8H+PwGkWat8jwyNuDPg2Efku0CIin8RpEfS99Myq\nDmpdR6EaSUOhLSisMaA6ONEVdMxyiVN5V8jg9+je3QE4zL5ThkvcLIj/EJH3AK8D04CvqOr9qVpm\n1CVx2sKXSrGwhrfvoPQywbmta21pZvuOvsD0Mz+vFVlvGKUQOwvCdbjmdI2KSfrOIo7wvHfMqBS4\nwgyNsGMZRlLEcsAicgbwdWAPnEGDAKqqb0nRNmMEkETX5qAQQ4MEC6cXS2EDuHDhqmEZEP5jGUZS\nxJ2E+wZwmqrupqpvUdVdzfkalVI4eebl4pZaqls4YTp6VCP7Tx5X1ii7fWZryRNthlEucUMQL6vq\nn1K1xKh7Cke7XT19iVWJDRnZLltQkZ1RimhWwWYkSVwHvFxEFgKLgcEES1W9IxWrjLojqAIujLyr\nxMJiypC/bUZ9EdcBvwXoAk70LVOgrh1wEvFJwyEoVSyMvCe65pw0LTQOnLdtRn0RNw3tvLQNqTZM\nDS1Z4o4cq6FKLEgLAqrDNqO+iJsF8a2Axa8By1X1Z8maVB2YilWyhN3WtzQ3MW7MqKq7y7iyfTpt\n+060OyAjVeKGIHYBDgRud5+fCawHDhWRWap6QRrG5YmpWCVLWAXcvNMOrlqnZpWQRtrEdcCHAEer\naj+AiHwH+A1wDFC8hWwNEie5fyRRaTw8jQo4w6h14jrgCcB4nLADwDhgoqr2i8iO8JfVLmloFtQq\nScXDbURpGEOJ64C/AawSkYdwquDeDfy7q4r2yyQNctPdPC/XAnSq6oyA7Z4D3gD6gT5VbUvSDhux\n7cTi4YaRDnGzIH4gIvcAR7iLvuxqAgPMSdIgVZ3tPRaRq9k56g5ilqpuTvL4fqptxJZXWpzFww0j\nHSJLkUXkQPf/YcCewPPu39vcZakhIgJ8CLglzePUCkmV7ZaDdXUwjHQopgVxkfv/6oC//0jRLnAa\ngL6sqk+HrFfgPhFZISLnR+1IRM4XkeUisnzTpk2JG5oFWbTyCcO6OhhGOkSGIFT1fPf/rCQPKiK/\nBN4WsOoSX17xOUSPfo9R1Q4R2QO4X0SeUtVfB22oqtcD1wO0tbWVIamdP3mGASwebhjpEOmAReSd\nwPOq+hf3+T/i5ABvAOap6pZyDqqqJxQ57ijgDODwiH10uP9fEZE7ceLTgQ64Hsg7LS7peLiVeRtG\n8RDEd4EeABF5NzAf+DHOxNj1Kdp1AvCUqr4QtFJExonIrt5jHI2KP6RoT+7UUxggz3i2YVQTxRxw\no2+UOxu4XlUXqeq/Ae9I0a6zKQg/iMhebiYGwFuB34rIauAx4G5VvTdFe3KnnpqE5hnPNoxqolga\nWqOIjFLVPuB4wD/ZFbudUamo6scClr0InOw+Xgccmtbxq5VqS4srl1pIa7MQiZEFxZzoLcDDIrIZ\n6MYpP0ZE3kF0fq5hhJJ3PLsYpoRnZEVkCEJVvwZ8EfghTtaBl0HQAHwuXdOMeqXa49kWIjGyomgY\nQVWXBSz7czrmGCOBak9rq4UQiVEfpBbHNYwoqjmeXe0hEqN+MAdslEzYBFW9TFyZEp6RFeaAjZII\nm6BavmELi1Z01MXEVbWHSIz6wRywURJhE1S3PPo8/arDlteqZGU1h0iM+sEcsFESYRNRhc632PZx\nqJeQhmGEUawSzjCGEDYR1ShS0vbFsHJlYyRgDtgoibAc3nOO3CfR3F7LxTVGAhaCMEoiaoIqyTbu\nlotrjATMARslEzZBleTEleXiGiMBC0EYVUm1lysbRhLYCNioSiwX1xgJmAM2qhbLxTXqHQtBGIZh\n5IQ5YMMwjJwwB2wYhpET5oANwzBywibhIjAtAsMw0sQccAjWFyw97MJmGA4WggjBtAjSwUR2DGMn\n5oBDMC2CdLALm2HsxBxwCGGaA6ZFUBl2YTOMnZgDDsG0CNLBLmyGsRNzwCG0z2zlqjOm09rSjACt\nLc1cdcZ0myyqELuwGcZOcsuCEJEPAvOAvwaOUNXlvnUXA58A+oHPq+rSgNdPBW4FdgdWAB9R1Z4k\nbTQtguQxkR3D2EmeaWh/AM4AvutfKCIHAWcDBwN7Ab8Ukb9S1f6C138duEZVbxWR63Ac9nfSN9uo\nFLuwGYZDbiEIVf2TqgZNfZ8O3KqqO1R1PfAMcIR/AxER4Djgp+6iHwHtadprGIaRNNUYA24Fnvc9\nf8Fd5md3oFNV+yK2GUREzheR5SKyfNOmTYkaaxiGUS6phiBE5JfA2wJWXaKqP0vz2H5U9XrgeoC2\ntrbg/umGYRgZk6oDVtUTynhZB7CP7/ne7jI/rwItIjLKHQUHbWPUCFaabIxUqjEEsQQ4W0TGuJkO\nB876a5sAAAtiSURBVACP+TdQVQUeBM5yF30UyGxEbSSHlSYbI5ncHLCIfEBEXgDeBdwtIksBVHUN\ncBvwR+Be4DNeBoSI3CMie7m7+BJwkYg8gxMT/kHW52BUjpUmGyOZ3NLQVPVO4M6QdV8Dvhaw/GTf\n43UUZEcYtYeVJhsjmWoMQRgjCCtNNkYy5oCNXLHSZGMkY4LsRq5YabIxkjEHbOSOlSYbIxULQRiG\nYeSEOWDDMIycMAdsGIaRE+aADcMwcsIcsGEYRk6YAzYMw8gJc8CGYRg5YQ7YMAwjJ8wBG4Zh5IQ4\n0rojBxHZBGwo8WWTgM0pmGPHt+Pb8evz+Puq6uRiG404B1wOIrJcVdvs+HZ8O74dP0ksBGEYhpET\n5oANwzBywhxwPK6349vx7fh2/KSxGLBhGEZO2AjYMAwjJ8wBG4Zh5IQ54BBEZIGIPCUiT4jInSLS\n4lt3sYg8IyJrReSklI7/QRFZIyIDItLmW76fiHSLyCr377osj++uS/38A+yZJyIdvvM+ufirKj7m\ne91zfEZE5qZ9vBAbnhORJ91zXp7B8W4QkVdE5A++ZRNF5H4Redr9PyHj42f22YvIPiLyoIj80f3+\nf8Fdns57oKr2F/AHnAiMch9/Hfi6+/ggYDUwBpgKPAs0pnD8vwamAQ8Bbb7l+wF/yOD8w46fyfkH\n2DMP+NcMP/9G99z2B0a753xQDt/D54BJGR7v3cBh/u8Y8A1grvt4rvdbyPD4mX32wJ7AYe7jXYE/\nu9/5VN4DGwGHoKr3qWqf+3QZsLf7+HTgVlXdoarrgWeAI1I4/p9UdW3S+03g+JmcfxVwBPCMqq5T\n1R7gVpxzr2tU9dfAloLFpwM/ch//CGjP+PiZoaovqerj7uM3gD8BraT0HpgDjsfHgV+4j1uB533r\nXnCXZclUEVkpIg+LyN9lfOw8z/+zbkjohjRvg12q4XMGUOA+EVkhIufncHyAt6rqS+7jvwBvzcGG\nLD97wAn3ATOBR0npPRjRXZFF5JfA2wJWXaKqP3O3uQToA27K4/gBvARMUdVXReRwYLGIHKyqr2d0\n/NSIsgf4DvBVHIf0VeBqnAtjvXOMqnaIyB7A/SLylDtKzAVVVRHJOnc1889eRMYDi4ALVPV1ERlc\nl+R7MKIdsKqeELVeRD4GvB84Xt3gD9AB7OPbbG93WeLHD3nNDmCH+3iFiDwL/BVQ8gRNOccnwfMv\n1x4R+R7w8ySOGUFq51kKqtrh/n9FRO7ECY1k7YBfFpE9VfUlEdkTeCXLg6vqy97jLD57EWnCcb43\nqeod7uJU3gMLQYQgIu8F/h9wmqp2+VYtAc4WkTEiMhU4AHgsQ7smi0ij+3h/9/jrsjo+OZ2/+6X3\n+ADwh7BtE+L3wAEiMlVERgNn45x7ZojIOBHZ1XuMMzGc9nkHsQT4qPv4o0Cmd0dZfvbiDHV/APxJ\nVf/Ttyqd9yCLmcVa/MOZXHoeWOX+XedbdwnODPla4H0pHf8DOHHHHcDLwFJ3+ZnAGtemx4FTszx+\nVucfYM//Ak8CT7g/hj0zOObJOLPgz+KEZbL+Du6Pk32x2v3MU7cBuAUnzNXrfv6fAHYHfgU8DfwS\nmJjx8TP77IFjcEIdT/h++yen9R5YKbJhGEZOWAjCMAwjJ8wBG4Zh5IQ5YMMwjJwwB2wYhpET5oAN\nwzBywhywYRhGTpgDrhNE5BJXPu8JV7LvSHf590XkoJSPPU9E/jXu8lpDRC4QkbElvuZYEUm7Wi/o\nuNsS2s88EVEReYdv2QXuMr886gx32XsLXj/s+yiOrOsqV97zNZ+85N8mYXMtMqJLkesFEXkXTsn0\nYaq6Q0Qm4Ugooqr/lKtxOSMio3Snqt2w5zG5APgJ0FVswzrjSZwKwCvd5x/EKQjxcw7wW/f/vRD+\nfVTVD7jrj8WRl3x/6mdQ5dgIuD7YE9isjk4EqrpZVV8EEJGHvBGLiHxCRP4sIo+JyPdE5Fp3+Q9F\n5Fsi8n8isk5EzvJ2LCJzROT37kjmct/yS9x9/RZHNzgSd6S0THYK3E9wl7/TN0paID4h7oLXf0kc\nYfLVIjK/yD4fEpFviiNg/gX3/K4TkUeBb7glvje478NKETndfV2jiPyHiPzB3efnROTzwF7AgyLy\noLvdiSLyOxF5XERuF0e4xRNwf0pEHgfOCDmPj4nIz1wbnxaRyyLes/ki8hnf83ki8q8iMl5EfuUe\n/0nP/oLXDhmBi8i14mibICKHi6Okt0JElsrQUl8/i3ElOEXk7cBrwGbfPgXHKX8MeI+I7OKuCv0+\nGkMxB1wf3Afs4zrE/xGRvy/cQET2Av4NOAo4GjiwYJM9ccow3w94Du5EHK2HI4AZwOEi8m5xVNjO\ndpedDLwzho0/Br6kqofgjKw8x3Mj8ClVnQH0B71QRN6H4wiOVNVDccSxo/YJzoirTVWvdp/vDfyt\nql6EU0r9gKoeAcwCFoijtXA+juD9DHefN6nqt4AXgVmqOssdzV0KnKCqh+GIIF3kOp/vAacChxOs\n6uZxBE5J+SHAB6Wg44iPhcCHfM8/5C57E/iAe/xZwNWuMyyKOEIz3wbOUtXDgRuAr4Vs/jrwvIj8\nDc7nvbBg/d8C61X1WRzh/lPc5UW/j4aDOeA6QFW34fzozwc2AQu90Y6PI4CHVXWLqvYCtxesX6yq\nA6r6R3ZqnZ7o/q3E0Z04EMch/x1wp6p2qSODGSlSIyK7AS2q+rC76EfAu8Vp87Srqv7OXX5zyC5O\nAG5UVxRJVbeE7dP3mkJncbuqeg7+RGCuiKzCcRy7AFPc43zXC1GoapAw+FE4HRIecV//UWBfnPdm\nvao+rU59/0/C3xHuV9VXVbUbuAPnwjcMVV0J7CEie4nIocBWVX0eEODfReQJHF2CVuLr004D/gZH\n2nIVzsVk74jtb8Vxvu3AnQXrznHXe9ud49od5/toYDHgusF1Lg8BD4nIkziO4Ycl7GKH77H4/l+l\nqt/1bygiF5RvaWZsj3guwJla0PEj7iASx4GeU/DaGSXYVijAEiXIcjtwFs6I2ruofBiYDByuqr0i\n8hzORcRPH0MHWN56Adao6rti2vpzYAGwXH26uOIo8p0JnC6OZrYAu4vIrqr6RgLfxxGBjYDrABGZ\nJiIH+BbNADYUbPZ74O9FZIKIjML58RRjKfBxX4yzVRxh8F8D7SLSLI5c4qlRO1HV14CtsrN7x0dw\nRuOdwBviZmzgjLSCuB84T9xMBBGZGLbPGOfkndfnvNt2EZnpO86n3PcHEZnoLn8Dpz8YOO2pjhY3\nO8CNJ/8V8BSwnxsrBXc0GMJ7xGny2IwzsnwkYtuFOO/LWey8a9kNeMV1vrNwRuCFbAAOEkc2tAU4\n3l2+FpgszkQZItIkIgeHHdy96/gSw8MUxwNPqOo+qrqfqu6Lo6H7gZjfRwMbAdcL44Fvuz+0Phwp\nzSHta9TpqvDvONq9W3AcxmtRO1XV+0Tkr4Hfub5qG3Cuqj4uIgtxZBJfwXHuxfgocJ3rRNcB57nL\nPwF8T0QGcBzoMJtU9V53hLlcRHqAe4AvR+yzGF8Fvgk8ISINwHqc2Pf3ccTtnxCRXpyY7rXA9cC9\nIvKiGwf+GHCLiIxx93epqv5ZnJZBd4tIF/AbdjrtQh7DcVZ7Az9R1VAxfVVd417kOnRnS5ybgLvc\nkeVynM+y8HXPi8htONq563HCSKhqjziTrN9ywzij3PeiMLvBv69bAxafw/CQxCLg0+6+Ir+PhoPJ\nUY4gRGS8qm5zR3h3AjeoauGPKBeb3MdzcbRev5CnTWniOu82Vf1s3rYY+WMj4JHFPBE5ASceeB9O\nmlHenCIiF+N8FzfgpDQZxojARsCGkTMi4nVbKOR4VX01a3uM7DAHbBiGkROWBWEYhpET5oANwzBy\nwhywYRhGTpgDNgzDyIn/DxErmXQiVFIxAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b6d68ad50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWAAAAFgCAYAAACFYaNMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJztnXucHGWV97+/mUxgAsgkJnIZCReXDS9sIIFZwIVVgi6o\nCEa8IKuut11cX3UX1LwbF18NiBKNiKu8K6Ki7IIYEIkgaEABL6yIIQmXKBEMEJgoJIQBQsZkZnLe\nP6p60tNT1V3dXVV9mfP9fPoz3U9dntOXOXXqPOciM8NxHMfJn45GC+A4jjNRcQXsOI7TIFwBO47j\nNAhXwI7jOA3CFbDjOE6DcAXsOI7TIFwBO47jNAhXwI7jOA3CFbDjOE6DmNRoAfJm+vTpdsABBzRa\nDCdL1q4N/s6a1Vg5SmlWuZzUueeeezaZ2YxK+004BXzAAQewYsWKRovhZMkJJwR/77ijkVKMp1nl\nclJH0mNJ9nMXhOM4ToNwBew4jtMgGq6AJV0u6SlJDxSNTZN0q6SHwr9TY459V7jPQ5LelZ/UjuM4\n9dNwBQx8G3hNydhC4KdmdjDw0/D1GCRNAz4FHAMcDXwqTlE7juM0Iw1XwGb2c2BzyfAbgCvC51cA\n8yMOPRm41cw2m9kzwK2MV+SO4zhNS8MVcAx7mdkfw+d/AvaK2KcXeLzo9RPh2DgknSVphaQVGzdu\nTFdSx3GcGmlWBTyKBS076mrbYWaXmVmfmfXNmFExNM9xHCcXmlUBPylpH4Dw71MR+/QD+xW9fmk4\n5jiO0xI0qwK+AShENbwL+EHEPsuBkyRNDRffTgrHHMdxWoKGZ8JJuho4AZgu6QmCyIbFwDWS3gc8\nBrw13LcP+Gcz+0cz2yzp08BvwlOdb2ali3mO4zg1sWxVP0uWr2XDwCD79nSz4ORZzJ8bucxUMw1X\nwGZ2ZsymV0XsuwL4x6LXlwOXZySa4zgTlGWr+vn49+9ncGgEgP6BQT7+/fsBUlXCzeqCcBzHaRhL\nlq8dVb4FBodGWLJ8barzuAJ2HMcpYcPAYFXjteIK2HEcp4R9e7qrGq8VV8COUyfLVvVz3OLbOHDh\nTRy3+DaWrfJoyFZnwcmz6O7qHDPW3dXJgpPTreXc8EU4x2ll8lqscfKl8N21fRSE47Qy5RZrXAG3\nNvPn9mb+HboLwnHqIK/FGqc9cQXsOHWQ12KN0564AnacGJIsruW1WOO0J+4DdpwIki6u5bVY47Qn\nroAdJ4JqFtfyWKxx2hN3QThOBL645uSBK2DHicAX15w8cAXsOBH44pqTB+4DdpwIfHHNyQNXwI4T\ngy+uOVnjLgjHcZwG4QrYcRynQbgCdhzHaRDuA3Ycx4lgQjTldBzHaTbyqvPsCthxIsjD+nGal7zq\nPLsCdpwSvMuF4005HadB5NWSvJVp9z543pTTcRpEnJXT74V4gJ13CP0Dgxg77xDaSQnnlYrelApY\n0ixJq4sez0k6u2SfEyQ9W7TPJxslr9NexFk5grZSMrUyEe4Q5s/t5cLTZ9Pb042A3p5uLjx99sSI\ngjCztcAcAEmdQD9wfcSuvzCz1+cpm9P+LDh5FucsXY2VjBt4s00mTqlOb8oZ8CrgD2b2WKMFcSYG\n8+f2jlO+BVpJyWTlp/VSnenRCgr4bcDVMdteLuleST+SdFieQjntTW+LK5ks/bReqjM9mloBS5oM\nnAZcG7F5JbC/mR0BfAVYVuY8Z0laIWnFxo0bsxHWaStaXclk6afNyz86EWhKH3ARrwVWmtmTpRvM\n7Lmi5zdL+k9J081sU8S+lwGXAfT19cXdXTrOKK1eDzhrP62X6kyHZlfAZxLjfpC0N/CkmZmkowms\n+afzFM5pb1pZyezb0x0ZNtcqLpRmYELXgpC0G/B3wPuLxv4ZwMwuBd4MfEDSMDAIvM3M3Lp1GkKj\nUpfj5l1w8qwx2XzQWi6URjPha0GY2QvAi0vGLi16fglwSd5yOU4pjUpdTjJvq7pQGo3XgnCcFiGP\nf9YoS7fSvK3sQmk0ecU6uwJ2nDrJ+p81ztItVb5pzzuRycuH7grYcSL4xLL7uerX6ymsKkzp6uCz\npx8eaVFm/c8aZ+l2SoxELHv4Qlv9LDh5FguuvZehHTs/364ONaYWhKTZqc7qOE3MJ5bdz5V37VS+\nAFuHdvCRa1ZHJjJkHTMcZ9GOmLV0rHLTowqvUyBpIsZ/Srpb0v+WtGf6YjhO83D1rx+PHN9hRCYy\npJmYEJU+HGfRFubxhIj0WbJ8LUMjY+8uhkYs9YJDiVwQZva3kg4G3gvcI+lu4Ftmdmuq0jhOExB1\nW1+g1BotXRy7+Iw5NSvAOF/vm47q5bp7+iNDynyhLRuariC7mT0EfAL4N+CVwJclPSjp9FQlcpwm\nptgaTbveQpyv9/YHN7qlmzN5FRxKZAFLOhx4D3AKcCtwqpmtlLQv8Cvg+6lK5ThNSsG/umxVPx+9\n5t5x1nI94WdxBd83DAy6pZsz8w6ZwZV3rY8cT5OkURBfAb4B/LuZjf5KzGyDpE+kKpHjNDkFyzfO\nVVHLbeqyVf0IIstgelRD/tz+YHTRrrjxWknqA35lmW3/nZ44jtN4dpvcyQvbo2NsC4swcTG4UJvC\nXLJ8baTyFXhUQwOIuxtJuy1VUhfEwcCFwKHAroVxMzsoVWkcpwno6uwAaktyqDUMLO68hndibgQS\nRN3gKOVQtKSLcN8CvgoMA/OA/wKuTFcUx2kOnh0cit22b093rIXbKdW8OFYu1MzJn7hAmLTLfSX1\nAXeb2U8lKWwNtEjSPYA3wnTajrjMtmJ3QFSlsXoiE5JWL6um6lqjKrQ1mwzNTFIFvE1SB/CQpA8R\nNMncPTuxHKdxRClDAW8/duYY5ZGmYklSvayaqmuNqtDWbDI0O0kV8L8CU4B/AT5N4IZ4V1ZCOU4j\nmT+3lxWPbeaqu9aPLowZcOVd67nyrvWj/sGe7i56pnSxYWBwdHGuXiVc7vhqqq7lVU6xHM0gQ610\ndcDQjujxNKmogMO28GeY2ceALQTxwI7Ttixb1c/Sux+P7Yxc8AMOFPmK+wcGOWfpas5euprejG61\nq8nOaobW8c0gQ60Mx3z5ceO1UlEBm9mIpOPTndZxmpcly9eOqYKVlMIRWd1qV1N1rRlaEiWVoRn9\nxHktwiU1qFdJukHSOyWdXnikK4rjNAdpxHqm1YG4mKiqayI6OyurCm1RxYKqkbdUhrTTuVuNpD7g\nXQkaXp5YNGZ4CrLThsTV2a2WDQODY6y769YPMHNaN9NrPF+cb/q6e/rp23/aGKsxi5ZE1S6qJZGh\nlf3EaZDUB3yfmV2cgzyO03DSUL4Ae3Z3jVFY24dHWLfxBX65qr9m5XL7gxvH+abjFFba9SNqUZaV\nZGhlP3EaJPUBnwm4AnYmBEkt4J7uLiR4ZuvQuDoO3V2dSONTlneY1WXd5dH+KM5izWLuJH7igkz9\nA4Oj303cQmcz+pPLkdQFcaekS4ClwAuFQTNbmYlUjtNAKilfwbi6v1H/+OcsXR15fNYKq1YquRiy\nmLtSAkqpTIXvJsr90Ypxx0kX4eYAhwHnAxeFjy9kJZTjNJJK6f7G+M4Y8+f2cufCE3lk8SncufDE\nUYUVd3ylBaw4smx/VM7FkNXclbqJRMkUJVsS+ZuRpNXQ5mUtiOM0C0k8wEms2CjrrkCt1lkWi2sF\nKrkYspq7nJ+40udcvL0V/cmJuyJLOoXACi6uhnZ+FkI5TrOT5La7WGFF3brXutpfUFgFt8c5S1ez\nZPnampVh4TxxF57i95p3Yfg4t0fx9kr79kzpYu75t/DM1iBxpqe7i0WnHdYUbomkXZEvBc4APkxw\nh/YWYP8M5XKchrHb5M6K+yS97S64JuLcGrVaZ2nFzxafJ4quTvHCtuFEcb/l5kgaO1xKlNujQKn7\nI2rfrk7x7ODQqPKFIINxwbX3lpUj7jeQ5LdRDUl9wH9jZv8APGNm5wEvB/4yVUkikPSopPslrZa0\nImK7JH1Z0sOS7pN0ZNYyOe3PjgqLcD3dXVVbT3t2d1U1Xom0/J3lfKxTp3SBBQqrViVf74Wi2EcM\nQYQKRPfFi/In7zZ5ElFJjUM7ync4/swbZ9PZMfay2dkhPvPG2YnkTkpSF0Th8rg17AP3NLBPqpLE\nM8/MNsVsey1wcPg4hqBm8TE5yeW0KYNRVVhCOoBFpx1W9TnjCnnXWuA7LX9n3P4CpkyeNMZyhOrd\nJmkkWlTj9ijd98CFN8XuW+6zKiS9XP3rxxkxo1PizKP3S91tkVQB/1BSD7AEWEmwTvH1VCWpjTcA\n/2VmBtwlqUfSPmb2x0YL5rQnX6yx7fxAiSKrNF6JtELCyp0nDSWfxjmqje0t3r+jTEx3uc9q2ap+\nlv7m8dFjR8xY+pvHx2Uc1ksiF4SZfdrMBszsOgLf7yFmlkcxdgNukXSPpLMitvcCjxe9fiIcG4Ok\nsyStkLRi48Z0m+o57cfUKcndAkn9m2m3Oa83JKwgd//A4Dj/dOE8achc7zmqdWGU7h+nfLs6VPaz\nOu/GNQyNjD12aMQ478Y1ieROSi3VLb9iZs+mKkU8x5vZkQSuhg9KekUtJzGzy8ysz8z6ZsxIt620\n03586tR4F0Ox37Aa5bDg5Fl0lPgb6omhrRQ/W47ShTdjZ+xz8XnSiPuNWxhLurBXra87zqdd/NH3\ndHex5C1HlP2sSl0vlcZrJXEYWhF9qUpQBjPrD/8+Jel64Gjg50W79AP7Fb1+aTjmODUzf24vZyfI\nYqvGvzl/bi+bZuzG+s2BxZlWF41ajo+S2wiU750Ld9bbSiPut/QcPVO62PLn4dFaypXioat1YcS6\nNgweXXzK6MvCHUCjU5ZrUcBPpS5FBJJ2AzrM7Pnw+UkEmXjF3AB8SNJ3CRbfnnX/r1Mvy1b1j6vt\nUKCnyD1RrRKYvvsuTN99Fx4pUgSNoBq5o5R8tT7Z4nMct/i2qhb2qvV1J60t0Swpy1W5ICTtDrw5\nI1lK2Qv4paR7gbuBm8zsx5L+WdI/h/vcDKwDHiZYFPzfOcnmtDHlkhK2/Hl49JY5bb9uXtQjdzVu\nlyj/eLUXrWrdIJX2X7aqn49ec2/TpCwnTcSYLWkVsAZYEy6K/VWWgpnZOjM7InwcZmafCccvNbNL\nw+dmZh80s5eZ2WwzGxcr7DjVUm6Fvjh+NMu6DFlSj9xJfbJxijou7jlO+Vfr6y63f0GmuIW5RqQs\nJ3VBfA34iJndDiDpBOAy4G8ykstxGkal9NfCtizrMmRJPXIntWDjFPWuXR10d3XGVj+Lk7dav3PU\n/uWSTqAxdy5JFfBuBeULYGZ3hH5Zx2k75h0ygyvvWh+7vbNoST3tugx5UesCXlKfbJyiHtg6xMVn\nzGnIRauchduoO5ekCnidpP8L/Hf4+h0EvlfHaTuuX1k+kGbEbLS4S6FAePGiXSMWddIoRJ7kHJXq\n9xYop6jzLuhTSaZOKXEIX9okXYR7LzCDoAfcdcD0cMxx2o4XtsffphYorOQX/IlxbYLyII3CPEnP\nkdQn2wj/eKWkmAUnz6KrpL5DV4e46K3jY4LzKsaTtCfcuWb2L6nO7DhtThaLOlFWapy/ddENaxJb\nxdXGNFeyFvP2jycOLStN+4upxbF9OPoiHDdeK0l7wh2f6qyO08TExQBXS62VzuKIUzJxC0sDg0OZ\nJTwkIU9XQ5ILyJLlayPTi6MuMnH1mMrUaaqJpC6IVZJukPROSacXHumK4jjNwd+8bFoq53lh+3BV\nboBKxCmZpAwOjfDRa6Lr4LZqTHOBJBeQZuyYkVQB70pQgvJE4NTw8fqshHKcRnLXus2pnKdgXUVR\nTZHy4sI59TJixjlLV3NAybxxPtt5h8youZh6niS5gFRzkemIcU3EjddK0p5w70l3WsdpXkbS8D+E\n9A8MjtYcuG79ADOndfPLCFfCOUtXc/bS1fT2dDPvkBnc/uBGNgwMsmd3Fy9sHx536xxFp8QOM/bs\n7uK5Pw9FFiKHsdEaC669F4j22c47ZAbX3dPfFCm7lUgSnZE0ggPg5QdN484/jL8Qv/ygdO6OCiTN\nhLsirAdceD1V0uWpSuI4bYhgNLJg+/AIDz+1hbOXro4shkO475V3rR89ZmBwKJHyhaCTx8VnzGHb\n8I5Y5VvK0A5j0Q1BicXSzs63P7ixaVJ2K5EkOqOarLo1G56PnCduvFaSxgEfbmYDhRdm9oykualK\n4jhtSIrGdEX27emumO0VRWGhrpRm9JmWI2l0RhLrPe4ziRuvlaQ+4A5JUwsvJE2jtkpqjtPWpOwi\nTEzhVjpN5djqC3OtQFIFfBHwK0mflnQB8D/A57MTy3Fak7ws3t0md0beSscpx84yzefiOoBkkUxR\nT4fkdiRpS6L/Ak4HngT+BJxuZv9d/ijHcZJQi9W8dfsIdy48kYvPmAPAOUtXc9zi25h3yIxIpXnR\nW4/gS2fMiez0G9cBpJ6uG1GkkbHXbpR1I0h6kZk9F7oc/gR8p2jbNDNLJ17HcSYgXZ1it8mTGBjc\nWVOit0IltgL79nRHJmZcd08/bzqqdzSKojgDbdmqfjqAYg9xJQsszWSKNDok54UEUVUra+1iHUcl\nP+53COJ972Fn66jivwelK47jTAymlrTmGTEbvb1f8djmstXYCvvFKbTbH9w4prVQgSXL1zJUEh5R\nqG/cyGpkzbio9/ZjZkZ+B28/Zmaq85RVwGb2+vDvganO6jgTmN7QTxvXmqegPK/+9eOjldamTO5k\n6/aRMRbtOQn61tU6nqQyWrUV2KptL9RILpg/G9j5HXRKnHnMfqPjaZG4JZGkwyWd5qnIjlM7HVLZ\naIUNA4MsW9XP7Q9uZEfoknj7sTPpmTJ53L7VRikkHU/iq63Fn1vvol7eC3h9+09j7z13RcDee+5K\n3/7pJmFAwlCyMOnicIKWRIVyFEZQntJxnDIU/HWTJ3Uyc1o3GwgUcVRrnD27u8b5dYtvhYuz0aIy\nuwqJH8ctvm1UsRWs1D27u+jq1JjEjigFmLSwTbX+3HoqpOXdSDOv+ZLG8h5rZoemNqvjTCAKLd+P\nnNnDpi3bYvuSdXd1IlUusFPqqliyfC39A4PjisIv+N69YIz6fQcGh+jqEFOndDGwdShWAdZb2Kac\na6LWRb28F/Dymi+pC+JXklwBO06NFBTW+s2DkQq20JVhYGuyTKvC+Qrpw7093eNikIdGLHLRbcrk\nSePC14pv5+spbNMzpSuTULO8F/Dymi+pAv4vAiW8VtJ9ku6XdF+qkjhOG2PAyvUDsQW9C1XKOhLG\nOSXtwRZFQSnGKckkvtq4fczGW/Bp1I/IOysvrpZz2jWekyrgbwLvBF7DzlKUp6YqieO0OZW6KRjE\ntkwvptjPW1Ca1SiiTim2gwbUV9jm2SrrSiQl7xZHTdMRI2Sjmd2Q6syOM0Gpp+NGXPPPBSfP4uyY\nsLRiSlvCFzMwOMSyVf2jftpaCtsU/NGl1Gup5t3iaGtM64u48VqppiPGdySd6WFojlMfBvR0d9WU\nghzX/HP+3F56Ym6PO6UxVmpvGWVYr6sgS0u1tFxms2XP1UJSC7gb2AacVDTmYWiOUyPbhndw8Rlz\nYi3Gaijc3i867bDIguNR9RvirOXiELZaFFycpQqMdvUoTrvO0optBZqyI4ak/QgW/vYiUPSXmdl/\nlOxzAvAD4JFw6Ptmdn6ecjpOrRT6syXx+Vai+PZ+l0kdowp46pQuPnXqYZFdjc+9/n5e2B7tiqg3\n5rXYNbFsVT/n3bhmTNZf4T03c4eNnu6uyNq/cXcZtZI4Ey5nhoGPhrHHxwIfjAmD+4WZzQkfrnyd\nlqKS8k3ioijc3hcSB4qVxp/L+Cu7Osv/6w8OjXB2RIhaNRRkKk25Lp2nGTtsLDrtMLpKKsd1dYhF\np0VXjquVplTAZvZHM1sZPn8e+B3QXJdIx8mYi8+Yw6OLTylby7fgXiiXOBBFXLRCKfXE8SbtztGM\nxXjmz+1lyVuOGBPlseQtR6RuqTd9VwtJBwBzgV9HbH65pHuBDcDHzGxNzDnOAs4CmDkz3WpGjpMF\nvT3do//sZx6zX2RlrnccO3N0n2oTB+IK40RRawZYUsXajMV4IN1SnHEkbcq5l6RvSvpR+PpQSe/L\nVLJgnt2B64Czzey5ks0rgf3N7AjgK8CyuPOY2WVm1mdmfTNmzMhOYMdJgdKogQvmz+Ydx84ctYQ7\nJd5x7MwxlbmqTVSIilYoRy1WahLFmmUsbyuQ1AXxbWA5sG/4+vfA2VkIVEBSF4HyvcrMxkVbmNlz\nZrYlfH4z0CVpepYyOU4eHDlzT5YsXzum6tcF82fzhwtfx6OLT+Gitx7B7Q9uHLO92vCv0kSKnu6u\n2NZEUJuVGqfkCw6VejtstANJXRDTzewaSR8HMLNhSemmhBQhSQTZd78zsy/G7LM38KSZmaSjCS4m\nT2clk+PkxZ1/2NlopjRSIK5K14Wnz+bC02ePCf+ad8gMlixfyzlLV0cmLkTdYpeeH2q3UvNOnmhF\nkirgFyS9mDAOXNKxwLOZSQXHEaQ+3y+pELD478BMADO7FHgz8AFJw8Ag8DazFGJ6HKfJGBwa4bwb\n11RcbCtOTohS1OcsXc2KxzaXLSqettLMw49aC9UWk8+KpAr4I8ANwMsk3QnMIFCAmWBmv6RCFI6Z\nXQJckpUMjtNMPLM1SBOOWzgrHY9S1AZcddd6+vaf1pRKMS/yri1cjqRdkVcCrwT+Bng/cJiZeTU0\nx8mRJcvXxoaklY7HLZoZ5dONJ0Ln4mpD9rIkaUeMfygZOlJSoV2947QNeSuaKV0dDA7tSFScZ0Oo\nFKMoTeooF2ZWLqKh2TsXp+E6aKbmoEldEH9d9HxX4FUEYWCugJ22Ik8rqKtD7NLVmbjCViESIUqx\nlhbYWXDyLM5ZujpSYZeLaEhDOdWiJJM2AU3DddBMzUGTuiA+XPT4J+BIYPdsRXOc/MnLCprS1QEa\n3xk5jkIkQtJws/lze3n7sTPHLaRUimiot/B5LS6MpMek5TpI+hnm0QS01ky4FwBvVe+0HdVkiNXD\ntmFLXIgnqmpYVLjZ2UtXj6s01rf/tKqs0ahGn9WEodXiwkh6TFqugySRHk3VlFPSjewsRdoBHApc\nk5oUjtMkJC1sXi+VlG9PdxeLThtfyQzGVxsrVhSllcYuPH32aPPOJNQbhlaLkkx6TJqug0rhcXn5\nwpNawF8oej4MPGZmT6QmheM0CfPn9uaigDtj2tL39nRz58ITR32icUkUBcoVvClWGNX4ZeuJ3a1F\nSSY9pl7rvBqaqimnmf2s6HGnK1/HqY9jD5o6zg9Z6PV2wMKbOHvp6kR+1EoKodAmPq/Qslo6YlTj\n167Uqy4t8moCWtYClvQ80e2rBJiZvShVaRynBSlYs1OndCVeVFu5/lnedFQvtz+4kf6BwYp94uJu\nfyv5rPft6a7rdrraiIZaXBjVHJNXZt28Q2ZEVqCbd0i6xbzKKmAz2yPV2RynyanFKiy4Eg7dZw8e\nfXow0SLe4NAItz+4kTsXnjjaqqcSUdZu1G15gYIVeU6MS6WS9VzrQlQtSrLZUpZvf3BjVeO1omrK\nJ0h6CUEcMABmNv4S0eT07bGHrTjqqEaL4WTJ6lDhzJlT1WGbtmxj3cYX2JFRSZFDn1oHwG9fctDo\n2LEHvZi71iWvITV5Uueopb19eGTc60Lf5MmTOpk5rZvpu+/CyvUDke3UJ0/q5MiZPbFz1XpcO1Du\nOzn2oBdXPF4/+9k9ZtZXab+kURCnARcRlKN8CtifoEtFuv05HKeBrN88mJnyjWJSRwebtmyr6pjt\nwyM8+dzImNcbn9/BQTN2Y/ruu4zbf9OWbezYMf49dUjMnFbenxmlfIvHN23ZxvrNg6MXgoLCbwcm\nT+qMvfikSdIoiE8T9Gb7iZnNlTQPeEeqkuTFrFlwxx2NlsLJkhNOCP5W+T2/aeFNiVKCa+W731kI\nwNv+fjEQNM2cMnlSKnHHnRIXvfWIsrGsBQrNOo+ucMv/0RjXSCHGOGkH5lZkfUxZzgtPn82RSd5f\nmTZSxSQtyD5kZk8DHZI6zOx2oKJ57TitRN6pqM9sHUotrGnEbFxkQ1yI2pTJkxIpyXLRCc1U0CYL\n8oq4SGoBD4TtgX4OXCXpKYJsOMdpG/JKwiggoKeKyIlKlEY21BvLWi46odaFvVaiaXrCAW8AtgLn\nAD8G/gCcmpVQjtMI5s/tTdQKPi2MwApOc85iBRhn0Rskrm0wf24vdy48kUcWnzJq+R648CY6Ym6x\nm7XBZrOSVAG/H9jHzIbN7Aoz+3LoknCctuLtx+bfNdvY2X2gXmVcrADLNd6sNhmjNJkjKotvojfY\nrIWkCngP4BZJv5D0IUl7ZSmU4zSKvv2nNWReI/AzXnzGnNii65UoVYDFfswoqvHZxvmTO6XMs9La\nmUQ+YDM7DzhP0uHAGcDPJD1hZq/OVDrHyYFCtlceVdDKsWFgsKx/NYpCBl1UxTTY6cc8MCbCI6nP\nNm6/HWY8sviUxPI6Y6m2HOVTwJ8Iug+/JH1xHCdf4kK1GkFP2BY+aUnMOKUbRb2VxJqpiHk7kcgF\nIel/S7oD+CnwYuCfzOzwLAVznDwoV00sb7b8eZhlq/rL+m4hcDV86Yw5Y7ogV6KWIjlpHu9Ek9QC\n3g8428wi740kTTWzZ9ITy3HyodFuh2KGdthoe3nYGf61Z3cXEgxsHaq5D1q9dX7TblffCuTRuj6p\nD/jjFXb5KUGbIsdpGZat6q9YhSxvCr7WLGJQ6z1nsxXMyZK8OmIkjYKoRJ7hk46TCkuWr22I8p06\npSs2MiHOp5pHfzJnJ3ll+qWlgJvJiHCcRDQia6tD4pTD92Hr9uHI7f0Dg8w9/5YxCjbPgupOQFN1\nxGgEkl4jaa2khyUtjNi+i6Sl4fZfSzogfymdVqYhK/iCpXc/Xjb9+JmtQyz43r2jCjYtaywvK7od\nrPW8OmI0pQtCUifw/4DXEjQAPVPSoSW7vQ94xsz+ArgY+FyaMjjtzwEvzl8B79hhDEWUhyxlaMRG\nFWwa1litVnS1yrRdrPW8oj4SxwFLOhI4nsDdcKeZrSza/KpUpYKjgYfNbF0493cJ6lH8tmifNwCL\nwuffAy7bZU2jAAAgAElEQVSRJKumwrzTEOpdXa7l+NJj5h0ygzv/sLnet5IpBQWbRgxuLW2JalmI\nyqubcNbkFfWRtCD7J4G3AN8Ph74l6VozuwDAzNL+JfcCjxe9fgI4Jm4fMxuW9CxBjPKmlGVxUqTe\n1eVajo86JqrfV7NRULBpdAOuxYquRZnm5TvNg2aqhvZ24K/N7FNm9imC4uzvzE6sdJF0lqQVklZs\n3JhuTyenOur1Z55345qqj2+mZIukdHVqVMHWU5u24EKIuy0sZ0XXokzz8p22C0ldEBsIesH9OXy9\nC5ClU6efIPmjwEsj5ivs84SkScCeBCnS4zCzy4DLAPr6+txF0UDqsZCWreqPXbwqd3wzJVsU6Onu\n4tnBoVF3yA/v/SMDg8F7K3SsKFawtVhjldKsK1nRtbg+0rDWJxJJFfCzwBpJtxL4gP8OuFvSlwHM\n7F9Slus3wMGSDiRQtG8D/r5knxuAdwG/At4M3Ob+3+anHn9mOSu3XPxssyVbdHd1jCrfgl/xgvmz\nU5+nnOWfpI5ELcp0ImbM1UNSBXx9+ChwR/qi7CT06X4IWA50Apeb2RpJ5wMrzOwG4JvAf0t6GNhM\noKSdJqceC6mclRt3fKOSLaLokDBgcGgHkF12VYG4z0swmu5cjlqV6UTKmKuXpKnIV2QtSMScNwM3\nl4x9suj5nwkWBjMjj1zwiUY9FlKc9dzT3VX1olAWdHUGtXG3j0QVK++I7LicZYRAGtETrkyzpdpy\nlBOGvHLBW4G0L0S1/lPHWc+LTjss9pikpR2rpXALH/W5FH9ePVO62PLn4VGrN4qsLhLuj21+XAHH\n0C7xjPXSTBeiWqznBSfP4pylq1N1Q3R1aHTeqLmLx49bfFvFpptZRQi4P7b5cQUcQzvFM9ZDs12I\nqrWe58/tTb3T8ZK3HJFYhkq/l6wtUnchNDdlFbCkGymzgGxmp6UuUZPgHQAC2uFCNDXF1u9QneVf\nzgVSTUcLpz2pZAF/Ifx7OrA3cGX4+kzgyayEagZa0X+WxaJhK12ICu//onVPM3lSJ+tX9bPisc2p\nKt9qifsdeQNLByooYDP7GYCki8ysr2jTjZJWZCpZg2k1/1lWvtpWuRCVvv/twyMs+N69DEVEJORJ\n8e8IYPIkV77OTpL6gHeTdFBRcZwDgd2yE6s5aCX/WaUU33ZvRRP1/hutfAuM/o5+/GIAjmyyz85p\nHEkV8DnAHZLWEcRx7w+8PzOpnKqJ88kWLOF6LONWuBC1kk/acQokKsZjZj8GDgb+FfgXYJaZLc9S\nMKc64nyynVIurVUaTV4+6U559y0nPZK2pZ8CLAA+ZGb3AjMlvT5TyZyqiCsgPRJTHqPdLMao99/V\nKbo60lWYZx6zX+WdHCchSctRfgvYDrw8fN0PXJCJRE5NxJUsrLb5Y700qh1N8fuHYLFryZuPYMlb\njmDqlK7U5rnunidarruD07woSQExSSvMrE/SKjObG47da2ZHZC5hyvT19dmKFW0dwDGGqJKEWYVB\n5TlXWU44Ifh7xx1jZFt0w5rRko/1UKiuVnUcb4RcTnsi6Z6SyLFIki7CbZfUTZiUIellwLY65Gsc\na9fu/EeowKYt21i/eZDtwyNMntTJzGndTN99l2zlS5n5wPFR7+MX6b+PmesH+Nbw+PKHk6/qhJk9\nqc8Xy+ow8+2EE0a/w72HR7g0g6k6vio2zdgt2e+iSC7HgeQKeBHwY2A/SVcBxwHvyUqoZmDTlm2s\n2/jCaAWr7cMjrNv4AkDLKeHpu++Si8zbI5RvufGsKf0Os2CHGes3D7bcb8JpDhK5IAAkvZigFZGA\nu8ysJXuvJXVBHLf4tsgMsN6e7kS1VCciTfOZhRbmca/5ZC7dMAQ8sviUyju6C2LCkNQFkTQK4qdm\n9rSZ3WRmPzSzTZJ+Wr+YzUs71EDIm7xaeVdi05ZtrFw/kFsromZMy3Zag0rFeHYFpgDTJU0luNgD\nvIigK3Hb0ko1EJqFZsiaW7aqn30zdjsU04xp2U7rUMkH/H7gbGBf4B52KuDngEsylKvhtEoNhGYj\n76y50gJEW7cP89UyyrcQwdApxcZIV6K3pzuXC4x3ZGl/KhXj+Q/gPyR92My+kpNMTUEzWHNZ0g7/\n3FEFiMpRHDa2bFV/TXWC8/JnN1MhfCc7kkZB7JDUY2YDAKE74kwz+8/sRGs8rVADoRba5Z+7XNff\nUkoV5/y5vTXFBed1B9RshfCdbEiaCfdPBeULYGbPAP+UjUhO1lSqnNYqJF0QjXMdLTrtMKpNVM5L\n+fki8MQgqQLulHZWIZHUCUzORiQna9rlnztuQXRSRweTJ3WOScmO6932Ny+bVtWceaUhx703XwRu\nL5Iq4B8DSyW9StKrgKvDMacFaZd/7riwtwOmT+HImT08svgU7lx4YqzVumxVP3ete6aqOfO6S2iW\nkD4nW5L6gP+NICLiA+HrW4FvZCKRkzlZRnjkubgXt1CaJM264AevNhIir7uEdl8EdgISKWAz2yHp\n28BtZtZajkJnHFn9czdica/WhdJqFvCKyfMuoV0XgZ2dJFLAkk4DlhD4fQ+UNAc4v527Irc7Wfxz\n17Nyn3dYXK2WrLsAnDRJ6oL4FHA0cAeAma0O+8KljqQlwKkE9Yf/ALynOAKjaL9HgeeBEWA4Sd61\nky21Lu5lYTlv2rKNNyy+jQ0Dg+zZ3YUEA1uHRpV7uXbx5XCL1EmTpItwQ2b2bMlYVrmetwJ/ZWaH\nA78HPl5m33lmNicr5duo4uKtSq2Le2mHxRWqoPUPDGLAwOAQz2wdwtip3OcdMmPcIpfj5E1SBbxG\n0t8ThKMdLOkrwP9kIZCZ3WJmw+HLu4CXZjFPJQpWWeGfuPCP60o4nlpX7ss1FK3lArh+82DZWhCD\nQyPc/uDGMR1EvNeb0wiSKuAPA4cRFGH/DvAsQY2IrHkv8KOYbQbcIukeSWeVO4mksyStkLRi48aN\niSZul2SFPIlri1Tptj3OQhbUdAFMUn94w8Ag8+f2cufCE3lk8Sm5Fe9xnGIq+oDDpIvzzexjwLlp\nTCrpJ8DeEZvONbMfhPucCwwDV8Wc5ngz65f0EuBWSQ+a2c+jdjSzy4DLIKgHnETGdklWyJtaFvei\nwuIKRXOKSbqgN3lSZ0UlXKr0a/UJO049VFTAZjYi6fg0JzWzV5fbLundwOuBV1lMxXgz6w//PiXp\neoJFwkgFXAtejjIbykU7FI/HKcMkF8CZ07pHu5dE0dmhcW6RqIuA42RN0iiIVZJuAK4FRn/ZZvb9\ntAWS9Brg/wCvNLOtMfvsBnSY2fPh85OA89OUw8tRpk+laIdiyzauu0aSC2Cl9kA7dtg4K7rw+pxr\nVlPOG7FsVb9HQjipkdQHvCvwNHAiQYjYqQQWahZcAuxB4FZYLelSAEn7Sro53Gcv4JeS7gXuBm4y\ns1RTo2v1ZzrxVONXrzcVt5wSjtOv8+f2llW+kF8qsjMxSOoDvs/MLs5BHszsL2LGNwCvC5+vA47I\nWpYk/sx2qKubF5WiHSq5Jar9bOOKrkdFPBS+x1rfg+PUQlIf8JlALgq4lWiXurp5EefbLUQ7QHm3\nRLWcecx+XHnX+sjxYkq/x3L4GoCTJkl9wHdKugRYylgf8MpMpGoRvGh2dVQb7QD1WcAXzJ8NwNW/\nfpwRMzolzjxmP/r2nzbG4t66fTjx4puvAThpklQBzwn/Fi90GYFPeMLioWrVUU20Q8ESrvfu4oL5\ns0cVMVTfxijuPThOGiSthjYva0FaEQ9Vq55SJVzOT5vF3UWtVdAcJwsSRUFI2lPSFwvZZJIukrRn\n1sI1O7Wu1E/kGhOlKd5Ryre7qzO2Tm+9dxd+d+I0E0ldEJcDDwBvDV+/E/gWcHoWQrUKtazUT/SF\nuzgLtFNih9noZ7hk+dq67y6iIlQ8481pJpIq4JeZ2ZuKXp8nqfqe3m1ItSv1zbJwl3X4XNz54yzQ\nHWY8sviUMWP1JMJs2rIt8kL3pqN6ue6efndDOE1B0kSMweJ0ZEnHAW5G1EAzLNxlXemt3PmTlqys\nNxFm/ebByAtdcRU0x2k0SS3gDwBXFPl9nwHenYlETUQWVmIzLNxlbYWXO381Kd71xAHHFeMpVEED\nOHtp9TdxhYuUJ984aZA0CmI1cISkF4Wvn8tUqiYgK19tM9SYyNoKL3f+vJpNTp4UXWy9cKGrNaX4\nvBvX8OehHRPWh++kS9KecJ8FPl9oDSRpKvBRM/tElsI1kqysxGbodpu1FV7p/Hk0m5w5rZvurs7Y\nC12tF5tntg6NG/PkG6dWkvqAX1vcl83MniGsy9CuZGklFhcCv3Phibn/49Zb6KbR508Sxjd9913K\n+pDTdvl4eJtTC0l9wJ2SdjGzbQCSuoHyNf9anGbw1WZF1lZ4luevxjVUztKutf5vT3cXA4PjreB2\n+F04+ZNUAV8F/FTSt8LX7wGuyEak5qAZfLVZkrUbIKvzp+UaKr1IJG1ItOi0w9r6d+HkS9JFuM+F\ntXcLnSw+bWbLsxOr8TSDr9YZT5quoeKLRFwB+KhjwH8XTjoktYAJC56nWvS82cljscipjjRdQ8Vh\nhnt2d9HVKYZGKtvC/rtw0iLpIpzjNAVpLfCVJosMDA6BwdQpXSlK6zjlSWwBO04zkJYLIMqXPLTD\nmDJ5UmSomeNkgStgp+VIwwXQDCnhjpM0EeN+xjcueBZYAVxgZk+nLZjjZEmcL3nPmDAzx8mCpD7g\nHwE3AW8PHzcSKN8/Ad/ORDLHyZA4X/LQyI4GSeRMRJK6IF5tZkcWvb5f0kozO1LSO7IQzHGypNiX\n3B925vASlU7eJLWAOyUdXXgh6a+BgvkwnLpUjpMD8+f2jlrCcR04HCdLklrA/whcLml3gka2zwHv\nk7QbcGFWwjkOZFs83nvEOY0kaSbcb4DZhXrAZvZs0eZrshDMcSD7Fk7VRD10d3nYvJMuVTXlBH5K\nUBPCm3I6uVCu9kMaJM2g6wAuPP3wVOZ0nAJJL+mXA88TNOV8K4EL4ltlj6gRSYsk9UtaHT4iy15K\neo2ktZIelrQwC1mcxpN1vG5UNITCv50KnvX2dPPFM+Z4+rGTOs3alPNiM/tC3EZJncD/A/4OeAL4\njaQbzOy3GcrkNICktR+K/cTXrR9g5rRupic4vxfXcRpJUgU8KOl4M/slNEVTzqOBh81sXSjPd4E3\nAK6A24wkZUFL/cTbh0dYt/EFfrmqP5Ei9eI6TqOopSmngM1k25TzQ5L+gSDZ46NhB45ieoHHi14/\nARwTdzJJZwFnAcycOTNlUZ0o0opcSGKhRvmJd5jV1SYoy8gLxynQkKackn4C7B2x6Vzgq8CnCVKf\nPw1cBLy3nvnM7DLgMoC+vj4P+MyYtCMXKlmoafuJs468cJwCZRWwpI/EjANgZl+sZVIze3XlvUDS\n14EfRmzqB/Yrev3ScMxpArJue19K2u2j8pbfmbhUioLYo8IjdSTtU/TyjcADEbv9BjhY0oGSJgNv\nA27IQh6nevKuNBYVydAhjasRnKSZZzk5vVKakzZlLWAzOy8vQYr4vKQ5BC6IR4H3A0jaF/iGmb3O\nzIYlfQhYTpASfbmZrWmArBOCav2heTc0LfUTT57Uycxp3RxdJGM5t0Lxsfv2dNPd1cHWofFFeXq8\nWLuTMk1XD9jM3hkzvgF4XdHrm4Gb85JrIrJsVT+LblgzpjxjEn9oIxqajvET37Vk3PY4t8KiG9aw\nbXjHGMUch5eLcNKm6RRwq9Juq+alFmMxlfyhzRhbG+c+qKb277NeJ9hJGVfAKdCOq+aVitRU8ofG\nRS406kIV5xapBndBOGlTUxREgVqjINqNdlw1r6Rga+1C3KgL1YKTZ3HO0tXj2rpUg7sgnLRJGgXR\nR5CM0Rs+/hk4ssxxE4p2XDUvp2Br9efWW1gnaRRDFPPn9talfMFdEE76lFXAZnZeGAnxUuBIM/uo\nmX0UOArwlLKQOGWV1ap/HkSFdkHQtv3C02fXZLHWc6EqbSNfsJ6rUcK9dX4frfx9Os1J0mpoewHb\ni15vD8cc4vuLZbnqnzXz5/Zy4emz6e3pRgTK60tnzGHVJ0+q2V1Qz4UqjbKUcReVao53nDRJugj3\nX8Ddkq4PX88HrshGpOYh6YJRM676p0HaRWrqCU9Lw81T+j1V45KYOqWr5b9Pp/lIWgviM5J+BPxt\nOPQeM1uVnViNp9oFI6+oVZl6LlRpJXcUf0/HLb4tcWTEKYfvU3knx6mSasLQpgDPmdm3JM2QdKCZ\nPZKVYI2mHSMbmoFaL1TVWM+btmxj/eZB3rTwpnFKvviupmdKF10dYmhHZVv49gc3Vi2z41QikQKW\n9CmCSIhZBJ0wuoArgeOyE62xtGNkQyuTxHpetqqf825cw1ef2gIwZrGuQLESf2brEF2dorurg8GI\n1ONi6o0hdpwoklrAbwTmAishSAuWlEkxnmYh73oGTmXKWc9JMvcKz4sZGjF2lNe9AEiV93Gcakka\nBbHdzIzAqCBsR9/WtGNkQzuTJHMv7u5lJEGGhRlVhbw5ThKSKuBrJH0N6JH0T8BPgK9nJ1bjiQrD\nqjX+1cmeJJl7cXcvnQnN20U3eME9J12SRkF8QdLfEXRDngV80sxuzVSyJsAjG1qHcrUeOrQzhjdq\nIe9NR/Vy3T39ZS1oqK5wj+MkIXEURKhw217pOq3JgpNncfbS6EbdxtjQwaiFvL79p9UUH+w49ZA0\nCuJ04HPASwiacgowM3tRhrI5TmLmz+2NV8AGBxaFpN258MTI4wtK+oCFN0Wep8MX4pyUSWoBfx44\n1cx+l6UwzUa71fhtdzql2AW10pC0uO/xE8vujxwHSBAu7DhVkXQR7smJqHzrLf7i5MuZx+xXcZ9y\n9SOWrernqrvWxx5bbzEfxyklqQW8QtJSYBmwrTBoZt/PRKomwDPhWo8L5s8OnnxHUMaTGxcxsWT5\n2rL+Xw9BdNImqQJ+EbAVOKlozIC2VcCeCdeaXDB/NnxpGgA93V2RkQt7dkd3tij33XoxHicLkoah\nvSdrQZoNz4RrfeLCe+PG475zAZ869bD0BHOckKRREF+OGH4WWGFmP0hXpOZg3iEzuDLCHzjvkBkN\nkMaphYGt0XG7ceNRBX8Adu1KulTiONWR9Je1KzAHeCh8HE7QJeN9kr6UkWwNJa76lVfFah2qLQBf\nyH6cWtJ8c3Bohy/AOpmQVAEfDswzs6+Y2VeAVwOHEBTpOanskS2K+4Bbn1rqecyf28uUyeNvDKvt\nvuE4SUiqgKcCuxe93g2YZmYjFEVFtBPt2OdtolFrPQ+/+Dp5UU0ixmpJdxCsSbwC+GxYFe0nGcnW\nUNwH3B7UUs/DF2CdvEgaBfFNSTcDR4dD/25mG8LnC9IUKIw3Ltwj9gADZjYnYr9HgeeBEWDYzPrS\nlMN9wBOXenrXOU41lFXAkg4xswclHRkOPR7+3VvS3ma2Mm2BzOyMovkvIoi2iGOemW1KWwbw29CJ\nTLs2WXWaj0oW8EeAs4CLIrYZML6qSUpIEvDWLOcoR8+ULp6JCFfqmRIdxO+0F16K1MmDsgrYzM4K\n/87LR5wx/C1BDYqHYrYbcIskA75mZpfFnUjSWQQXEmbOnJlo8rgmCQmaJziO4ySibBSEpL+WtHfR\n63+Q9ANJX5Y0rdZJJf1E0gMRjzcU7XYmcHWZ0xxvZkcCrwU+KOkVcTua2WVm1mdmfTNmJFtEezam\n+HbcuOM4TrVUckF8jSDml1DBLQY+TJCUcRnw5lomNbNXl9suaRJwOnBUmXP0h3+fknQ9wQLhz2uR\nJ4pWXQn3EpqO0zpUigPuNLPN4fMzgMvM7Doz+7/AX2Qo16uBB83siaiNknYrdGUOQ+FOAh5IU4BW\nbMrpJTQdp7WoqIBDaxTgVcBtRdsStzOqgbdR4n6QtG8YCgewF/BLSfcCdwM3mdmP0xSgFZtyliuh\n6VTHslX9HLf4Ng5ceBPHLb7NL2JOJlRSolcDP5O0CRgEfgEg6S8oHx5WF2b27oixDcDrwufrgCOy\nmr9Aq62Ee+hcOhTuJAoXsySdNBynFspawGb2GeCjwLcJFr0KMQAdBL5gp4nw9Ol08DsJJy8q1oIw\ns7vM7Hoze6Fo7PdZJGE49dGKfutmxO8knLzwQqdtRCv6rZsRv5Nw8iLLhTSnAZT6rQuLSR6Wlhyv\nBeHkhSvgJiLtGF5fTKoNrwXh5IUr4CYhC2XpnZ1rp9UiYJzWxBVwk5CFsmz1xSTP6nPaHV+EaxKy\nUJatvJjkWX3ORMAVcJOQhbJs5bA0j8V1JgKugJuELJRlK4eltbr7xHGS4D7gJiGrlfdWXUxq1Wp0\njlMNroCbiFZVllngsbjORMAVsNOUeCyuMxFwBew0LX5H4LQ7vgjnOI7TIFwBO47jNAhXwI7jOA3C\nFbDjOE6DcAXsOI7TIDwKogxeDKY6/PNynOpwBRyD19KtDv+8HKd63AURgxeDqQ7/vBynelwBx+DF\nYKrDPy/HqR5XwDG0ci3dRuCfl+NUjyvgGFq5lm4j8M/LcaqnYQpY0lskrZG0Q1JfybaPS3pY0lpJ\nJ8ccf6CkX4f7LZU0OU35WrmWbiPwz8txqqeRURAPAKcDXyselHQo8DbgMGBf4CeS/tLMRkqO/xxw\nsZl9V9KlwPuAr6YpoBeDqQ7/vBynOhpmAZvZ78wsaon8DcB3zWybmT0CPAwcXbyDJAEnAt8Lh64A\n5mcpr+M4Tto0ow+4F3i86PUT4VgxLwYGzGy4zD6jSDpL0gpJKzZu3JiqsI7jOLWSqQtC0k+AvSM2\nnWtmP8hy7mLM7DLgMoC+vj7La17HcZxyZKqAzezVNRzWD+xX9Pql4VgxTwM9kiaFVnDUPk4L42nN\nzkSgGV0QNwBvk7SLpAOBg4G7i3cwMwNuB94cDr0LyM2idrKlkNbcPzCIsTOtedkqv8Y67UUjw9De\nKOkJ4OXATZKWA5jZGuAa4LfAj4EPFiIgJN0sad/wFP8GfETSwwQ+4W/m/R6cbPC0Zmei0LAwNDO7\nHrg+ZttngM9EjL+u6Pk6SqIjnPbA05qdiUIzuiCcCY6nNTsTBVfATtPhac3ORMHrATtNRyHawaMg\nnHbHFbDTlHhaszMRcBeE4zhOg3AF7DiO0yBcATuO4zQIV8CO4zgNwhWw4zhOg3AF7DiO0yBcATuO\n4zQIV8CO4zgNwhWw4zhOg1BQWnfiIGkj8FiVh00HNmUgjs/v8/v87Tn//mY2o9JOE04B14KkFWbW\n5/P7/D6/z58m7oJwHMdpEK6AHcdxGoQr4GRc5vP7/D6/z5827gN2HMdpEG4BO47jNAhXwI7jOA3C\nFXAMkpZIelDSfZKul9RTtO3jkh6WtFbSyRnN/xZJayTtkNRXNH6ApEFJq8PHpXnOH27L/P1HyLNI\nUn/R+35d5aPqnvM14Xt8WNLCrOeLkeFRSfeH73lFDvNdLukpSQ8UjU2TdKukh8K/U3OeP7fvXtJ+\nkm6X9Nvw9/+v4Xg2n4GZ+SPiAZwETAqffw74XPj8UOBeYBfgQOAPQGcG8/8vYBZwB9BXNH4A8EAO\n7z9u/lzef4Q8i4CP5fj9d4bv7SBgcvieD23A7/BRYHqO870COLL4NwZ8HlgYPl9Y+F/Icf7cvntg\nH+DI8PkewO/D33wmn4FbwDGY2S1mNhy+vAt4afj8DcB3zWybmT0CPAwcncH8vzOztWmfN4X5c3n/\nTcDRwMNmts7MtgPfJXjvbY2Z/RzYXDL8BuCK8PkVwPyc588NM/ujma0Mnz8P/A7oJaPPwBVwMt4L\n/Ch83gs8XrTtiXAsTw6UtErSzyT9bc5zN/L9fyh0CV2e5W1wSDN8zwAG3CLpHklnNWB+gL3M7I/h\n8z8BezVAhjy/eyBw9wFzgV+T0WcwobsiS/oJsHfEpnPN7AfhPucCw8BVjZg/gj8CM83saUlHAcsk\nHWZmz+U0f2aUkwf4KvBpAoX0aeAiggtju3O8mfVLeglwq6QHQyuxIZiZSco7djX3717S7sB1wNlm\n9pyk0W1pfgYTWgGb2avLbZf0buD1wKssdP4A/cB+Rbu9NBxLff6YY7YB28Ln90j6A/CXQNULNLXM\nT4rvv1Z5JH0d+GEac5Yhs/dZDWbWH/59StL1BK6RvBXwk5L2MbM/StoHeCrPyc3sycLzPL57SV0E\nyvcqM/t+OJzJZ+AuiBgkvQb4P8BpZra1aNMNwNsk7SLpQOBg4O4c5ZohqTN8flA4/7q85qdB7z/8\n0Rd4I/BA3L4p8RvgYEkHSpoMvI3gveeGpN0k7VF4TrAwnPX7juIG4F3h83cBud4d5fndKzB1vwn8\nzsy+WLQpm88gj5XFVnwQLC49DqwOH5cWbTuXYIV8LfDajOZ/I4HfcRvwJLA8HH8TsCaUaSVwap7z\n5/X+I+T5b+B+4L7wn2GfHOZ8HcEq+B8I3DJ5/wYPIoi+uDf8zjOXAbiawM01FH7/7wNeDPwUeAj4\nCTAt5/lz++6B4wlcHfcV/e+/LqvPwFORHcdxGoS7IBzHcRqEK2DHcZwG4QrYcRynQbgCdhzHaRCu\ngB3HcRqEK2DHcZwG4Qq4hZF0blgy776wTN8x4fg3JB2a8dyLJH0s6XirIelsSVOqPOYESVln6EXN\nuyWl8xSXfXxI0veLf0eS7gjLcxbKQn4vHJ8Vblst6XeSGt1CqGWY0KnIrYyklxOkSR9pZtskTSco\nm4iZ/WNDhWswkibZzkp2414n5GzgSmBrpR3bjIvN7AsAks4AbpM028w2htvfbmalae9fDo8r1E+Z\nnZ+4rY1bwK3LPsAmC2pDYGabzGwDjFoqfeHz90n6vaS7JX1d0iXh+LclfVnS/0haJ+nNhRNLWiDp\nN6FlfV7R+LnhuX5JUCu4LJLmSLpLO4vaTw3H/7rIal+iouLbJcf/m4Ji5PdKWlzhnHdI+pKCouX/\nGr6/SyX9Gvh8mNZ7efg5rJL0hvC4TklfkPRAeM4PS/oXYF/gdkm3h/udJOlXklZKulZBsZZC0fYH\nJeuAeDQAAAWoSURBVK0ETo95H++W9INQxockfarMZ7ZY0geLXi+S9DFJu0v6aTj//QX5S44dY4FL\nukRBPRMkHaWget49kpZrbHpvJGa2FLgF+PsKu+5DkLVWOO7+cM7O8Pst/JbeH44rlG2tpJ9Iurn4\n9zehyDu90h+ppUzuTpAm+XvgP4FXFm27A+gjUCKPAtOALuAXwCXhPt8GriW4CB9KUPsWgnoDlwEK\nt/2QoEj2UQTpoFOAFxGkao8rkk1R8WyCdM5Xhs/PB74UPn8AeHn4fDERBeaB1wL/A0wJX0+rcM47\ngP8sOv7boeyd4evPAu8In/eEn9tuwAeA77Gz+H5hnkcJC6ED0wkK4OwWvv434JPArgTp6geHn9c1\nwA8j3su7CdJrXwx0h++/r3S/cN+5wM+KXv+WoCjQJOBFRfI8zM6mulvCvycUzw9cEs7dFX6WM8Lx\nM4DLy313RWNnA18t+ozXsjNFd0k4/h7gWYKSrecAPeH4WcAnwue7EBSMOpDgQnUrQdH7fYEB4M2N\n/p9qxMNdEC2KmW1RUI7yb4F5wFJJC83s20W7HU3wz7wZQNK1BJXTCiwzsx3AbyUV6pueFD5Wha93\nJ1AwewDXW1iYSFLZwjSS9iT4R/xZOHQFcK2C1k57mNmvwvHvELhSSnk18K3CfGa2Oe6cRccsLTnH\ntWY2UvS+TtNO//SuwMxwnkstdFEUPqsSjiW4SN2poCzhZOBXwCHAI2b2UPieryRQOlHcamZPh/t9\nn6DmwLgKdma2StJLJO0LzACeMbPHFVTo+qykVwA7CGoT70VQm7YSs4C/IihnCYHi+2PZI3aiktfj\nXBBm9i1Jy4HXEBQuf7+kIwg+88OLrNs9CX5LrwCuDr+bDZJuSyhL2+EKuIUJf8B3AHdIup+gStO3\nqzjFtqLnKvp7oZl9rXhHSWfXLmluvFDmtYA3WUmXD6lUv0QiAgV6Zsmxc6qQrbToSrkiLNcCbyao\njVy4qLydQCEfZWZDkh4luIgUM8xYt2Jhu4A1ZvbyKuQtMJcEpU4tcH9dDlweupT+Kpz3w2a2vHhf\n5dDPr1VwH3CLomDl+eCioTnAYyW7/QZ4paSpkiYRVFKrxHLgvUU+zl4FxcB/DsyX1K2gROKp5U5i\nZs8Cz2hnx453EljjA8DzCiM2CMo8RnEr8B6FkQiSpsWdM8F7KryvDyvUuJLmFs3z/vDzQdK0cPx5\nAqsfgpZUx0n6i3Cf3ST9JfAgcICkl4X7jVHQJfydgsaO3QTtbO4ss+9Sgs/lzey08PcEngqV7zxg\n/4jjHgMOVVAqtAd4VTi+FpihYOEWSV2SDiszP+F+byKwYq+usN9rQgsdSXsTuFr6CT7zDxRt+0sF\nZTV/DpwR+oj3IbiDm5C4Bdy67A58JfxHGybwCY65/bWgk8JnCer1biZQGM+WO6mZ3SLpfwG/CnXV\nFgLf6UpJSwlKIz5FoNwr8S7g0lCJriPwFUJQYvDrknYQKNBxMpnZj0MLc4Wk7cDNwL+XOWclPg18\nCbhPUgfwCIHr4xsEbpn7JA0BXyfwnV4G/FjSBjObFy5mXS1pl/B8nzCz3ytoE3STpK0EPvY9iOZu\ngiLfLwWuLL2NL3nva8KLXL/tbINzFXBjeKezguC7LD3ucUnXEPiYHyF0I5nZ9tAN8OXQjTMp/CzW\nREx/jqR3EPjHHwBOtJ0REABXSRoMn2+yoIj+ScB/SPpzOL7AzP4k6RsETWRXhhe+jQQXn+uBEwn8\n2+sJ3DkTEi9H2eZI2j30F08i+OFfbmbXN4NM4fOFBPVd/7WRMmVJqLz7zOxDjZalGZH0bYLFw+81\nWpa8cQu4/Vkk6dUE/sBbgGUNlgfgFEkfJ/j9PUawUu84Ew63gB2nAUgqdFgo5VWFaAmn/XEF7DiO\n0yA8CsJxHKdBuAJ2HMdpEK6AHcdxGoQrYMdxnAbx/wHDyIqwrfS8VQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b7d20ef90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWAAAAFgCAYAAACFYaNMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b7d159b90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "alg_list = [\"MAST\", \"DESeq\", \"edgeR\"]\n",
    "estim_list = [mast_, deseq_2, edger_]\n",
    "for t in range(3):\n",
    "    alg = alg_list[t]\n",
    "    estim = estim_list[t]\n",
    "    p, fc = estim.fit(return_fc=True)\n",
    "\n",
    "    p = np.array(stats.p_adjust(FloatVector(p), method = 'BH'))\n",
    "    signed_p = - np.log10(p) * np.sign(fc)\n",
    "\n",
    "    signed_p[signed_p >= 20] = 20\n",
    "    signed_p[signed_p <= -20] = -20\n",
    "\n",
    "    plt.figure(figsize=(5, 5))\n",
    "    plt.scatter(signed_p, signed_p_value)\n",
    "    plt.axvline(x=3, c=\"r\")\n",
    "    plt.axvline(x=-3, c=\"r\")\n",
    "    plt.axhline(y=np.log10(0.05), c=\"r\")\n",
    "    plt.axhline(y=-np.log10(0.05), c=\"r\")\n",
    "    plt.xlabel(\"Signed log corrected p_value \"+alg)\n",
    "    plt.ylabel(\"Signed log corrected p_value micro-array\")\n",
    "    plt.tight_layout()\n",
    "    plt.savefig(interest + \"_\"+ alg + \"_vs_micro_array.pdf\", transparency=True)\n",
    "    print interest + \"_\" + \"Detected \", np.sum(p <= 0.05), \" genes with \" + alg\n",
    "    print interest + \"_\" + \"Detected \", np.sum(p_value <= 0.05), \" genes with microarray\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "1\n",
      "2\n",
      "3\n",
      "4\n",
      "5\n",
      "6\n",
      "7\n",
      "8\n",
      "9\n",
      "10\n",
      "11\n",
      "12\n",
      "13\n",
      "14\n",
      "15\n",
      "16\n",
      "17\n",
      "18\n",
      "19\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[[0.073530264912539028, 0.76195418663262859, 0.77699152011362205],\n",
       " [0.068250986446227457, 0.76407072488814065, 0.75826893353941272],\n",
       " [0.077765533677118415, 0.77260147889449005, 0.7276271634849687],\n",
       " [0.06571610300534475, 0.74687195787500604, 0.72894300792291511],\n",
       " [0.066476238389557463, 0.81145682788121143, 0.72523775707701532],\n",
       " [0.073137387809561, 0.73231267047962778, 0.75063842822729998],\n",
       " [0.067199978491508036, 0.82443267559563582, 0.7312196276647589],\n",
       " [0.076060841551785857, 0.73331833113736489, 0.74386148126488161],\n",
       " [0.066649895844789767, 0.7419593462360633, 0.75332442179428261],\n",
       " [0.070325053965708775, 0.75922082702337057, 0.70693150646782799],\n",
       " [0.070269363881499861, 0.77580140511537032, 0.75295125109653716],\n",
       " [0.066838485681943213, 0.78232651181128521, 0.73260509350153857],\n",
       " [0.072138465222251119, 0.75967116402180879, 0.72413077683556826],\n",
       " [0.070912261946707938, 0.77283260186826508, 0.7374312488686523],\n",
       " [0.077054115195049844, 0.76841790311523128, 0.76475486305470852],\n",
       " [0.067760362192793133, 0.77206408028342466, 0.73434284361641389],\n",
       " [0.07224165860596303, 0.77594926120694474, 0.74223373296016271],\n",
       " [0.072709432948611694, 0.78839349006976422, 0.71921132879401817],\n",
       " [0.069727937232613127, 0.80565268488927322, 0.75851539329128215],\n",
       " [0.06013452783126641, 0.71649494666524693, 0.75402759792249752]]"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "boostrap_scores_de(p_value, edger_, 20, p_prior=p_prior, rank_auc=rank_auc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### deseq"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/romain/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:6: RuntimeWarning: invalid value encountered in log\n",
      "  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1\n",
      "2\n",
      "3\n",
      "4\n",
      "5\n",
      "6\n",
      "7\n",
      "8\n",
      "9\n",
      "10\n",
      "11\n",
      "12\n",
      "13\n",
      "14\n",
      "15\n",
      "16\n",
      "17\n",
      "18\n",
      "19\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[[0.032902867101084761, 0.75064093569227797, 0.52526938374252241],\n",
       " [0.036789116382926083, 0.72017810096583035, 0.5179735461816265],\n",
       " [0.056749138334733717, 0.56656014843078162, 0.52498740987703107],\n",
       " [0.053730686712023967, 0.53090684455984949, 0.5306070519810977],\n",
       " [0.039129326918783583, 0.67061776463314104, 0.56007875281260056],\n",
       " [0.034052745141134398, 0.69409934011357677, 0.52993442061155505],\n",
       " [0.053500138973116597, 0.60034740095641459, 0.52857115763920492],\n",
       " [0.036578018526089592, 0.71090287459174029, 0.52722321624760649],\n",
       " [0.033767152313007941, 0.70978620792624969, 0.51776167095316028],\n",
       " [0.037461152311075947, 0.70510829261397867, 0.52464612260895827],\n",
       " [0.032786172923720942, 0.71062023825762077, 0.53715575721464059],\n",
       " [0.036776903122752322, 0.69476703943609541, 0.50803282689608897],\n",
       " [0.031149835542971429, 0.73337895785256335, 0.51878387978316931],\n",
       " [0.031469050849262284, 0.76241826912790078, 0.52070611885915308],\n",
       " [0.032477264713363083, 0.76937458923521596, 0.53382288143281109],\n",
       " [0.037800571718959758, 0.69876076492377492, 0.52646699039922895],\n",
       " [0.03941311672865884, 0.70365038370668653, 0.53003174925384489],\n",
       " [0.037672128241556378, 0.73211051514709979, 0.53377095631641092],\n",
       " [0.033399164282163565, 0.69160867743950627, 0.51726138265245813],\n",
       " [0.039219063521729651, 0.74222275590963471, 0.51529748873958892]]"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "boostrap_scores_de(p_value, deseq_2, 20, p_prior=p_prior, rank_auc=rank_auc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### MAST"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "1\n",
      "2\n",
      "3\n",
      "4\n",
      "5\n",
      "6\n",
      "7\n",
      "8\n",
      "9\n",
      "10\n",
      "11\n",
      "12\n",
      "13\n",
      "14\n",
      "15\n",
      "16\n",
      "17\n",
      "18\n",
      "19\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[[0.065781157888356173, 0.76990326605871884, 0.7609410028266288],\n",
       " [0.062784048122332473, 0.80917994940195381, 0.71204310957015748],\n",
       " [0.063971421212487903, 0.77001789394647091, 0.72957517022432006],\n",
       " [0.060448377538310706, 0.8125881475527672, 0.72636840859406548],\n",
       " [0.066768861752931272, 0.7854925099161234, 0.70587186877758745],\n",
       " [0.063340469095832436, 0.81813090452909309, 0.72191960120862753],\n",
       " [0.061036225462394837, 0.82093187501924691, 0.75659941239539386],\n",
       " [0.070572252155508608, 0.77022310055222787, 0.74184663798265049],\n",
       " [0.063235704755484873, 0.81860553895462995, 0.72209226227773371],\n",
       " [0.066928372070680031, 0.82596621361863576, 0.72422685436595791],\n",
       " [0.07967038190921473, 0.79923288463129472, 0.7388919058161717],\n",
       " [0.063989746041377762, 0.78559024397455945, 0.72696993747998384],\n",
       " [0.067434143767636387, 0.75430709654858719, 0.74708773688680941],\n",
       " [0.059395332998462259, 0.71476823798946998, 0.74906916189760087],\n",
       " [0.063526875290721915, 0.79988289017266589, 0.75284681899828731],\n",
       " [0.066169561989585887, 0.79532265352750287, 0.73440828773131717],\n",
       " [0.062650195078347529, 0.76764944984080585, 0.72528649205619833],\n",
       " [0.069350135570590304, 0.7682962580970627, 0.73030619491206816],\n",
       " [0.067181831562086514, 0.80766849220846715, 0.71682052995808787],\n",
       " [0.062791723026266705, 0.7960548046500352, 0.71742762855591291]]"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "boostrap_scores_de(p_value, mast_, 20, p_prior=p_prior, rank_auc=rank_auc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "### DCA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "idr = IDR.IDR()\n",
    "\n",
    "deseq_2_DCA = deseq.DESeq2(100, 100, \\\n",
    "                 data_path + \"de/dca_data_train.npy\",\\\n",
    "                 data_path + \"label_train\", \\\n",
    "                    couple_celltypes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:6: RuntimeWarning: invalid value encountered in log10\n",
      "  \n",
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:6: RuntimeWarning: invalid value encountered in sign\n",
      "  \n",
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:6: RuntimeWarning: invalid value encountered in multiply\n",
      "  \n",
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:8: RuntimeWarning: invalid value encountered in greater_equal\n",
      "  \n",
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:9: RuntimeWarning: invalid value encountered in less_equal\n",
      "  if __name__ == '__main__':\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "BDC_Detected  320  genes with DCA\n",
      "BDC_Detected  879  genes with microarray\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:21: RuntimeWarning: invalid value encountered in less_equal\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWAAAAFgCAYAAACFYaNMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJztnXucHGWV97+/mUzCJCCTSFQYQdBlwyvLJZBFfNEVxAVv\nYARUXN3Fy8rqu7qLunmFxVeClyW7kdVV1kVcUVREQCBGUQNyU1lRQy5AlAgGCIwoA2ECJGMyMznv\nH1WV1PRUVVf3VHX3dJ/v59Of7n6qup7TXd2nT53nXGRmOI7jOI2nq9kCOI7jdCqugB3HcZqEK2DH\ncZwm4QrYcRynSbgCdhzHaRKugB3HcZqEK2DHcZwm4QrYcRynSbgCdhzHaRLTmi1Ao9lrr71s//33\nb7YYTpmsXx/cz5vXXDkqaVW5nMK58847HzezudX26zgFvP/++7Ny5cpmi+GUybHHBve33tpMKSbS\nqnI5hSPpoTz7uQvCcRynSbgCdhzHaRJNV8CSLpX0mKR7YmNzJN0o6b7wfnbKa88I97lP0hmNk9px\nHGfyNF0BA18FXl0xdjZwk5kdCNwUPh+HpDnAecBLgKOA89IUteM4TivSdAVsZj8GNlUMvwG4LHx8\nGbAw4aUnAjea2SYzexK4kYmK3HEcp2VpugJO4blm9mj4+PfAcxP26Qcejj1/JBybgKQzJa2UtHJw\ncLBYSR3HceqkVRXwTixo2TGpth1mdomZLTCzBXPnVg3NcxzHaQitqoD/IGlvgPD+sYR9BoB9Y8+f\nH445juNMCVpVAS8HoqiGM4DvJOyzAjhB0uxw8e2EcMxxHGdK0PRMOElXAMcCe0l6hCCyYQlwlaR3\nAw8Bbw73XQC818z+1sw2SfoE8MvwUB83s8rFPMdxnLpYtnqApSvW87uhYfbp62XRifNYOD9xmalu\nmq6AzeytKZuOT9h3JfC3seeXApeWJJrjOB3KstUDnHPt3QyPjAEwMDTMOdfeDVCoEm5VF4TjOE7T\nWLpi/U7lGzE8MsbSFesLnccVsOM4TgW/GxquabxeXAE7juNUsE9fb03j9eIK2HEyWLZ6gGOW3MwB\nZ1/PMUtuZtlqj3TsBBadOI/enu5xY7093Sw6sdhazk1fhHOcVqVRCzFO6xGd37aPgnCcViVrIcYV\ncPuzcH5/6efZXRCOk0KjFmKczsUVsOOk0KiFGKdzcQXsOAksWz3Alm2jE8bLWIhxOhf3ATtOBZWL\nbxGzZ/Zw3kkHu//XKQy3gB2ngqTFN4CZ06e58nUKxRWw41Tgi29Oo3AF7DgV+OKb0yhcATtOBY3K\ngnIcX4RznAoalQXlOK6AHSeBRmRBOY67IBzHcZqEK2DHcZwm4QrYcRynSbgP2HEcp4JGNOQEV8CO\n4zjjaGQdaFfAjpNAoywgp/VoZB1oV8COU4F3wuhsGpmK7otwjlNBo1qSl4X3sZscjUxFdwXsOBVM\n5WI8kfU+MDSMsct6dyWcn0amorekApY0T9Ka2O0pSWdV7HOspM2xfT7WLHmd9mLP3p6axluJqW69\ntwIL5/dzwSmH0N/Xi4D+vl4uOOWQzomCMLP1wOEAkrqBAeC6hF1/Ymavb6RsTvsj1TbeSkxl672V\naFQqeksq4AqOB35rZg81WxCnMxjaOlLTeCuxT18vAwnK1ktp1k4jImFa0gVRwenAFSnbXippraQf\nSDq4kUI57ctUrgfspTSLoVG+9JZWwJKmAycDVydsXgW8wMwOAz4PLMs4zpmSVkpaOTg4WI6wTtsw\nlZVYI/2X7UyjfOmt7oJ4DbDKzP5QucHMnoo9/r6kL0jay8weT9j3EuASgAULFliZAjtTn6leD9hL\naU6eRvnSW10Bv5UU94Ok5wF/MDOTdBSBNf9EI4Vz2hdXYp1No3zpLauAJc0C/hL4u9jYewHM7GLg\nNOB9kkaBYeB0M3Pr1imETk1F7tT3XclxB83lG3dsTBwvkpZVwGa2BXh2xdjFsccXARc1Wi6n/enU\nVOROfd9J3HJv8lpR2ni9tPQinOM0g8kswEzlNGBP4thFo3zAroAdp4J6f3xTPQ3Ykzh20ahQRFfA\njlNBvT++qW5BTuX456JZdOI8errGpz72dKnwUMRcCljSIYXO6jgtTNpCS7UFmKluQU7l+OdSqEw9\nLyEVPe8i3BckzQC+ClxuZpuLF8VxWoNaFmDiUQNdEmMJgThTxYKc6vHPRbJ0xXpGxsafy5ExK7wo\ney4FbGYvl3Qg8C7gTkm/AL5iZjcWJonjtAh5LdllqwdY9O21O3+oScp3qlmQHv8c0HKLcGZ2H/BR\n4CPAK4DPSbpX0imFSuQ4TSZvOcrzv7tugpUE0CU8DXiK0yh/eC4LWNKhwDuB1wE3AieZ2SpJ+wA/\nA64tVCrHaSJpZSc3/3GEZasHdirUJ1Oqo+0weHDJ68oSz2kArZaI8Xngv4F/NrOdNriZ/U7SRwuV\nyHGaTJpiNaNjExM6jZZKxDCzV5jZ1+PKN7bt64VK5DgtTDysrC/FVZE27kwdkupAZI3XS94wtAMl\nfVvSryRtiG6FSuI4U4RoIWbxyQcnxoouPtlLU091GtUVJa8L4ivAecBngOMI/MGexOF0JNFCjIdt\ntS9pZb2KLveVVwH3mtlNkhS2Blos6U7AG2E6bUdvTxfDIztSto0PKys6bCseV7xnbw9S0ArJlXt7\nkteK3SapC7hP0vslvRHYvUS5HKdpXHDKoYk/jNkze0oNK6usJTE0PMKTW0emZF0JJx95FfA/AjOB\nfwCOBN4OnFGWUI7TTBbO72f3GeNTcp81o5vVHzuhVAs0qZZEnKlUV8LJR1UFHLaFf4uZPWNmj5jZ\nO83sVDO7owHyOU7DecmnbuSpbeMV4VPbxnjJp8pN/MyTZTVV6ko4+ajqAzazMUkva4QwjtMK/OHp\n7TWNF0VaG5zKfWrFu1y0LnkX4VZLWk7QnXhLNGhmngHnOBnEld81G4fYb04ve6Xsu+jEeeM6UlRS\nT10J73LR2uRVwLsRNLx8ZWzM8BRkx0mlUvltHx1jw+AWfhpLZ45TGdZWRBREWo3ixcvXuVXcAlRV\nwKEP+C4z+0wD5HGctiFJ+e2w7JKGRYe1pfmMh4ZHGBoOUq7dKm4eVRfhzGyMoD284zg10AoF2vP6\njIdHxvjwVWunZC+7qUzeMLTbJV0k6eWSjohupUrmOE2iqBoPrdDiJ6nLRRpjZh5z3GDyKuDDgYOB\njwMXhrdPlyWU4zST7aPJi2BDwyN8dNnduY+TpPy6VHxfsSwWzu/nglMOob+vd2eN4tkzq/+ReMxx\nY8jbEeO4sgVxnFZha0oaMrCzRuwnF1Zvk1i5qDZ9Wjf7zenlqAb7WSv9ypWLg2l4zHH55I2CQNLr\nCKzg3aIxM/t4GUI5Titzxc8fzqWAoUL53bG0kPknG9db+ccw1XvZ1UL02Q0MDdMdvu/+JkaB5O2I\ncTFBKvJxBIXZTwN+UaJcjtOyJCmrPDz+zDY2bhrm1LOvrzv0q6i43vgfQ5JFPNV62eWh8n1G5zHp\nMxRBnG0lRTdGzusD/t9m9jfAk2Z2PvBS4E8LlmUCkh6UdLekNZJWJmyXpM9Jul/SXb4w6DSC7jqK\nwi5bPcCGwS1sHx2b1EJXWlzvZPy1SX7iduxll1Vro/Iz7OlOPsdp4/WS1wUROYO2hn3gngD2LlSS\ndI4zs8dTtr0GODC8vQT4r/Decepm1vRutmxP94++9SX71nzMpSvWc2GF5Rz96GtRdGWFtnVCN+Rq\nn1F8+/aEZqtZ4/WSVwF/T1IfsBRYRWCdf6lQSerjDcDXzMyAOyT1SdrbzB5ttmDO1GVrhvJ9+9H7\n5fb/xilKcabVi2gXf22ZdSuq1dpoxmeYtyfcJ8xsyMyuAV4AHGRmjSjGbsANku6UdGbC9n7g4djz\nR8KxcUg6U9JKSSsHB4ttque0H2lt6QUseMGcuo5ZVExwUmhbq/trl60e4JglN1dN8qish1x0PHJW\nTHSzPsN62gp93sw2Fy5JMi8zsyMIXA1/L+kv6jmImV1iZgvMbMHcucW2lXbajzQXr0HdCmHRifPo\nqjhwPT/6qeavrUWpluHfjhP/7GCXL7+Zn2HuMLQYCwqXIgUzGwjvH5N0HXAU8OPYLgNA3CH3/HDM\ncepmKKUtPdTnt4Xgx//43Fls3DSMYFKX11PJX5umVD981VpgfORGI1K3W+2zq0cBP1a4FAlImgV0\nmdnT4eMTCDLx4iwH3i/pWwSLb5vd/+tMlj17e3YWqkmiXoWw1+4z2Gv3GTyw5HX1ijblSPusxswm\nhH61u387iZpcEJJ2J4gBbgTPBX4qaS1BzPH1ZvZDSe+V9N5wn+8DG4D7CRYF/0+DZHPamJGx9Ew4\naE2FkNfP2miyPqtK98JU9G9PlryJGIcAXwPmhM8fB84ws3vKEszMNgCHJYxfHHtswN+XJYPTmWSF\noLWiQmjlouvViszHLeTKDL1OqFOc1wXxReBDZnYLgKRjgUuA/12SXI7TkrTiglfW4lWzZY3m//BV\na3OlO7eaj7Zs8rogZkXKF8DMbgVmlSKR47QwragcWqHucBYL5/dz4ZsP6zj3Qh7yWsAbJP0/4Ovh\n87cT+F4dp6P46LK760rEKJOpsHjVyu6FpOSPRpFXAb8LOJ+gB5wBPwnHHKejuPyOjSx4wZyWUBwR\nSX7WyViXZWWjtaJ7Ic1/3iXYkZB1PK2rwbUgwp5w55rZPxQ6s+NMQYzAn/nBK9dkKqdEJVaSTEVa\nl628oFcGaf7zNEaTtPIkqKqAzWxM0ssKndVxpjBZZQwhXYm97Jlt7LX7jEnNnWadFmVdtvKCXhk0\n20+e1wWxWtJy4GpgSzRoZt6W3ulokpRTmhLbuCn4sb9hyc11WaqNsE5bfUGvaKoV6CmbvFEQuxGU\noHwlcFJ4e31ZQjnOVGJgaHhcAkTaD3r76BgbBrfUXWym7FoJMPmiQa2aEJLGohPn0VPh1618XiZ5\ne8K9s2xBHGcqE1eoaQs4ADsmURO4EdbpZBb0arXQ06IPGh4pUalvBc+a0c1T2yb6gg98TrHRt7ks\nYEmXhfWAo+ezJV1aqCSO0wYMj4ylKt808irQRrS5n0y1tVos9KQqaYu+vZZFV68trRxlmswjFUXW\nR8YsUfkCDD69vdD58/qADzWzoeiJmT0paX6hkjhOh5JXgdZjnaYt2mWFmtW7oJf2RzIwNMwxFX7v\nJGVdqQih/AXAWq8esoo01UNeBdwlabaZPQkgaU4Nr3UcJ4O88bq1hpuluQRWPrSJa+4cKHwxL21B\nS+Ec8bmyQr0qKXMBsNmLcHmV6IXAzyRdTfB5ngZ8qjSpHKdD6OvtqbubcTXSXAJX/PzhCXUZirA0\nkyz0pO7CwyNjO1vC56HMjL68MpdF3pZEXwNOAf4A/B44xcy+nv0qx3GqNVCWKM3HmVWLt5b985Lk\nP05TZGNmE2pD9HRrQgRC2fUikmR+29H7lTZfJZkWsKRnmdlTocvh98A3Y9vmmNmmsgV0nKlKb083\nF5wS1I1ICxV7cutIaZlmaZfXaVEafTOTe+FlkeRLvv3sV+7cfsySmxNl6I/5gpsdBZF0VfGNOzaW\nOmdENRfENwnife8ksMpVcf/CUqVznClMPHpg4fx++OGzWbVxaMJ+ZS00pS3agTE8MrHofE6PwE7y\nhJ1lLRymuVPaMeMujUwFbGavD+8PaIw4jtMedEuJimT7aPXC5EWRtmj3wSvXJO6/ucoKf6W1u2Xb\naNW05WoLh2W2oZ8M/SlXD/0F+6NzRzJIOhTYP/4aT0V2nGTS/KzTpyW3Ra9caFq2eoDFy9ftDHua\nPbOH8046uK5moJWvWbpifYprQixbPZCaNFFp7aZR+WeSZum2cuGfRSfO40NXriF+ndBF/oiVvORt\nSXQpcCiwDnbKZATlKR2nY0lbMe/v651g3X3nmW3sN6eX3p7uzFjeZasHWHT1WkZijtont46w6NsT\nOwnXQ1qboKRGmRFJERVp5I1aaEThn3ot7JUPbaLSSbMjHC/yzyFvLYijzWyBmZ1hZu8Mb14P2Olo\nohXzpE4Pxx00d0Km14bBoI5VtUyzpSvWj1O+ESNjVkjdh2jlvzshRCMtcy2vi6SWqIWyU6uTsu3y\nZtZd8fOHaxqvl7wuiJ9JerGZ/arQ2R1nitLX27NztX/BC+ZMsLKSrLsdZmzcNFw1ljdLARWlnBbO\n70/1BQ8MDe90RUQWZNr63OyZPcycPq0uH27ZnTzSLOwPX1X9SiLNhZQ3djkveRXw1wiU8O+BbYRX\nXmZ2aKHSOM4UYWh4hAPOvn6n0omHVMV9t5VsHx2bkJZbqQiysrNqVU5Zl+BZ80QZc1f+8uHEFGEI\nrN16/NIRRXfyqCQrDrpVfM15FfCXgb8G7oYJrhHH6Uiiy9oPXbmG7m7tVFTV6gVUpuXCeEVw3EFz\nE+NQe7pVk3KqtsiV1TJ+eGSMb/58Y2phof4CIhbK7hOX9QfTKkXm8yrgQTNbXqokjjNF2QHsSLES\nq1GpCJatHuCaOyf6KGdN7+ZTb8xXlSyi2iJXdKyzUlwRWVXd4skWkyFPanW9C2lZfzDQGkXma+mI\n8U3guwQuCMDD0BynCOKKIC3aYMv2XYtj1aqZJR03bXzh/P7UsLR6qEVZ5tl3MqFq0fYPX7U20Xfb\nCl2j8yrgXgLFe0JszMPQHCcn/X29TJ/WnZiIEVcEWVZZrdXM0lORx8f71pox19ebnLJci7LMu281\nKz5Puc09e3vYsn10nC+77BoTeclbjOedCbfSwtAk7SvpFkm/krRO0j8m7HOspM2S1oS3j5Ulj+NM\nhujHvt+cXrqUXWymmlUWVTPLU/h80YnzJoTIwa5FqCgcKwpLmx2rBTFjWhenHvn8xHY9i08+OFG2\nWgqy5903y4pPCzP76LK7xxV2HxoeYWzMmD2zp+Yi82XTqjV9R4EPm9kqSXsAd0q6MSEM7idRurTj\nNJNjXjSHB58Y3mlxSTC0dWT8pXXYEbm/rzf1srua3xLyVzPLugRPWoT6Y8zaHRoe4Zo7B3jLUfty\ny72DuVwKtcT15t03K1QtTYknLWDuIKh18cCS1yXO2yxaUgGb2aPAo+HjpyX9GugHPA7ZaUlWbdyc\ny6raa/cZmQtY8ciANL+slFw4J8l6zor3reZ7Hh4Z45Z7B3MvuNUS15t336xQtbT3lcbQ8EjL1Z7I\nmwnXNCTtD8wHfp6w+aWS1kr6gaTk66LgGGdKWilp5eDgYEmSOp1MVu+zqEvwqo1DPP7MtoRXj2fh\n/H5uP/uVfPYth09wIXSRrHyzQtTy9JIrIistyeWR5mvNu29Wj7p6FtHqzYwri7xNOZ8r6cuSfhA+\nf7Gkd5crGkjaHbgGOMvMnqrYvAp4gZkdBnweWJZ2HDO7JEylXjB37tzyBHY6mkplVemjjNrS5/3B\nVyqfvt6e1CD8WdOnpVpyeZRdEQ0/I3nji3S79SSrmFqaf0Z/SA8seR23n/3KcYuHST7uNLpEbh/1\nrOnJx00br5e8LoivAl8Bzg2f/wa4kiBBoxQk9RAo38uTwt3iCtnMvi/pC5L2MrPHy5LJcbKoVFZp\n6chnXbmGpSvW57r8jcfJHrPk5tQkj8rMvPhx0xIeomP+bmiYvpk99HRpXA2KeiMFto3u+pvIKjhf\nb/PP+OshPcwsTk8sUaaSJCv/U288hA9fvZax2OfR3SU+9cZD6pY3ibwKeC8zu0rSOQBmNiopf1e9\nGpEkAuX+azP795R9ngf8wcxM0lEE1vwTZcnkOFkkKas8IWVQPZ418ltWi9WNX1ZXHrdS2X102d1c\nfsfGnTUentw6Qk+36OvtYfPwSN3+0fO/uy61/sIHr1xTuN81Ok6ln7inW8yaPm3ce0n7DNN851B+\nd468CniLpGcTVt6TdDSwuVBJxnMMYeqzpMjT/s/AfgBmdjFBY9D3SRoFhoHTzQqulOE4KST9wCvj\nT7uqNJ5MyoJLslJr7SIcv6xOi5GNK9+IkTFj1oxprDnvhIkHzcGy1QM8uTXZQo8+h3pr/mYtntWi\nLMusPVEPeRXwh4DlwIsk3Q7MJVCApWBmPyUo+JO1z0XARWXJ4DhpCFh62mFVkwvyVM6KLLK0xITd\nerpqUr7x46YlOmRVN5tMem7eUpnxP4g8RefzJG3kcWfUoqgbVSw+lwIO43FfAcwj+P6tN7PsiiOO\n06YYwY+48pI6LY04qwV7VJM3LQwsS/n29/WydftootXZLaUuOGUp2cmk59aivAeGhnMXnS+ycHte\nv3MjisVD/o4Yf1MxdISkqF2943QcSRXN0hTQjgxLOFLMtVqe/X293H72KydYasCEjhtxIssvyRcq\nam+5U4vLJU63lFl0Pq7kii7cXlQdjSLI64L489jj3YDjCcLAXAE7HU9kGe3Z25MYpbBnbw+zZiT/\n1Lol9j/7+prmi/stF87vZ+VDm7ji5w8zZka3xKlH9nPLvYOpdSAGhoYntFIS8Laj96vZL1uryyUi\na99asuFqJcu1AFT9Mym6gE9eF8QH4s8l9QHfKlQSx5nCRKFcSUiBZdn1X5pgDdfaYaGyDm9UvjI6\nzpgZ19w5wKlH9o8r2FM5n7Grn129tX2zXC47zFKVWNRFI2/R+SILt6e5Fs7/7jr+OLKj6p/JcQcV\nm0dQbybcFsBb1TtOyD59vQylRABE412xwjYJ7dhyEU9EgOwU4niiQ1L/t0j5Vh4znr13zJKbUxNH\nslwuDyx5HRe++bDEBJDzTjqYRSfOm1DoB5Iz+mpJ2qhGmsxPbh3Jtdh5/V2P1jxnFnl9wN9l1xVL\nF/Bi4KpCJXGcKUxWnGnfzB7OufZuvjK2K0GhnoDJpDKQeX2VeQv41LL6X801kCfqIE8URHSsIha/\nsrpk5CEtzK5e8vqAPx17PAo8ZGaPFCqJ40xRZs/sSU0I6O3pxmxiCmytpJWBTFMoe/b25IofzpO9\nl7b6n8c1kKU4o/FICT+5dYTzv7tu3LaiyVNtrpHkrQd8W+x2uytfxwmILqmjlfXhkbGdl/vRpfLm\nKj3iko759qP3G3fJvfRNE+OOIb3OgxLqHiTNkzd7LxqPuyeWrljPqUf21+0aWLZ6gEVXrx23cBmF\nopVVICdyZ9RLWjH6esm0gCU9DYkx21FX5GcVKo3jTDFmTOua0KFizGyccktbjOoLoyMGhoZ3xgrX\nuiCWdpmf1ucNgh9vrR2Z9+nrTXRPXHPnQN3+2KUr1ucORSuSrDZM0QJib08XWxO6gbz+sL0LlSVT\nAZvZHoXO5jhtxtDwSGJa7/DIGIuXr2PLttFU/+vrD9ubTy6cfHGXpMv8tAI13RK/veC1qcfKcivk\ncU9Ui7GNb89yg5fdMDPtfUZ/Joeff0OiAv7e2kcLOWcRqqV8gqTnEMQBA2BmE0vPtzgL9tjDVh55\nZLPFcMpkTWj9HX54XS+/Y0M5NZ1e/NgGAH71nBcCMK2ri64usX10jOnTutlvTi97hV0zJkvWe4h6\n08XnfPyZbWzcNMz20TGmdXeBweiOHeP2yTrm0S98No8/s40Ng1vGhdpJortLjI7tYFpXF2Nm1KJz\niv5c4sTfc+U81d5rNXTbbXea2YJq++WNgjgZuBDYB3gMeAHwayC1CLrjONmM7thBVOA3qhUMTFrZ\nBEXfK1MtdhE1Bo3mfPqPoww+vW2n4hwd24EkpnV3sX10jI2bAms0rano9GmBD3rjpuEJcc5mxmhY\nBnJ0R1o143SK/Fwq2Wv3GaUo9lrIGwXxCeBo4EdmNl/SccDbyxOrRObNg1tvbbYUTpkce2xwX+d5\nPr3GzLS8fOubZwfH/6slqftEcbn1Ei1sJflW08iqVRHR29OdmNwRXbYfMb+fU8++PtOtMBm6JS58\nc/JCZBm86//9INEFMbOni1994jXVD5Az0DtvIsaImT0BdEnqMrNbgKrmteM4tTFZ3+fi5etqUr6Q\nLxsvKbmjMuphz4IjBCplbGT7oBkpnTbSxuslrwU8FLYH+jFwuaTHCLLhHMcpkCgut97mkWkdM7LI\nYwFDEPWQJVO92X15KaMaWRrVshqLIq8CfgNB0fMPAm8D9gQ+XqgkjtPhRNEGadloKx/alLtFfC0c\n/cLZrNq4uWrcsEiuAhfJULRySqLs6IiIIgsAZZHXBfF3wN5mNmpml5nZ50KXhOM4BdDX28NuPV18\n8Mo1fPiqtYnhXpffsTGzo2/W5fnMnq7UBparNm4el1DR19tDT/d4czZpSa+yoWWacooSU/IYyCLI\nLEzbt2gFmMb+z06eJ228XvIq4D2AGyT9RNL7JT23UCkcp02IsthqQQSNLJ/cOoKR7pOtpgCzOlKc\ncuTzueCUQxKL8kT+3ajz8JrzTmDpaYeN8/Xm6aCRlpV34ZsPyzxGRH9fLw8seR2rP3YCn3nL4blb\n3JfBHRuerGm8XvKmIp9vZgcDfw/sDdwm6UeFSuI4U5xuiSP225Mrfv5wTa8z6q8VEVeAWZfnt9w7\nyML5/anF4StfW9kKvj9H2/q0qmVA1QI4STUkiqqAVg9pf4K1lg+tRl4fcMRjwO8Jug8/p1BJHGeK\nM2bG7b/d1NA544t2WR0pIgVbr28zb03eyqy8yJ+dRVr6dVEV0OohbWEy6QpiMuSygCX9H0m3AjcB\nzwbeY2aHFiqJ4zg1EdXOjZRclnUWhYilFRSvVmi8Xos0rWg7BAr8s285fEI94lbgrS/Zt6bxeslr\nAe8LnGVmiRU+JM02s2KdI47TZkj11QFOY9b0aSyc388xS26u6sIYGh7hmCU3s3X7aOL2W+4dzHx9\nvWFxWW6RRroUauWTCw/hgcFnxl3RHPOiOYXWgYD8PuBz0pRvyE0FyeM4bUvB7kOGhkdYtnogd2jW\nwNBwakHxrGNEFnZWBEYaaa6N/r7ellW+ELznVRs3jxtbtXFz4Ykg9bYkqqTkEGzHcZI459q7mTl9\n8tlZWT7gtCpoH75qbdW2RWmREY2KZqiXrMpvRVLrIlwaZaWAO46TQRGdHaopxDTrOPI5Z7UtytOW\nqBVptbb0DUfSq4H/ALqB/zazJRXbZwBfA44kiMp4i5k92Gg5HWcqU9mHLcnXm6ePWlaacDOjGeLU\n4sdutUzw88vzAAAgAElEQVS4ahTqgpDUDfwn8BqCBqBvlfTiit3eDTxpZn8CfAb41yJlcKYeUbuc\nOzY8waqNQ7n8dXk7ALcrM8OFPICPLrubD165ZoKv97iD5qZm0cVpVJpwPdTqx26U6yS3ApZ0hKR/\nkPQBSUdUbD6+UKngKOB+M9tgZtuBbxHUo4jzBuCy8PG3geOlssuBtDdTWRnFf2AQ1JGttlCU9KPM\nauUzleiWEJDQ+X0c0ee1bPVAamePeBW0LBqVJlwPtfp0G5UIkrcg+8eANwHXhkNfkXS1mX0SwMyK\njj7vB+LpRI8AL0nbx8xGJW0miFF+vGBZOoJa2pG3IrV08816TTsQb61zQJXaxmLXpXlWunHkRjgg\no+ZvKy+s1ePTbYTrJK8F/Dbgz83sPDM7j6A4+1+XJ1axSDpT0kpJKwcHs+MdO5VGrfqWRZqPcmBo\nONWab+VL5nrp6+0ZZ6lVs0qNXQtkacSPkXa82TN7WvqPOk3uZlvteRXw74j1ggNmAGVenw4QJH9E\nPD9hvp37SJpGUCIzsUKbmV1iZgvMbMHcudkZP51Ko1Z9J0Oai2TZ6oHMRYg0f1+zf3xF0tfbw2ff\ncjhrzjthnCJM8mVWEi1KJSHGW7ZpvtHzTmrt7mStGg6XVwFvBtZJ+qqkrwD3EBRp/5ykz5Ug1y+B\nAyUdIGk6cDqwvGKf5cAZ4ePTgJutlm5/zjha1UKIyFpEybp8jkiy5hedOK8hAezdFY7YPAtaeYhS\neR8MK5gtnN8/4U8KqOq/jSICKuUS8Laj9xun0JtdJKdeWlXuvGFo14W3iFuLF2UXoU/3/cAKgjC0\nS81snaSPAyvNbDnwZeDrku4HNhEoaadO8hZbaRZZLpK8VnrSfr09XYm9v4rgmBfN4fL3vHTnnwQE\nDSwvOOUQlq5Yn+g2ifc+y/K3JhWwSfPjX3DKIdx+9isnbIdd57iWeN1WCSurlVaUO5cCNrPLqu9V\nLGb2feD7FWMfiz3+I8HCYGnUm/8+FWn1gPksF0meOFUILL3onA4MDWf0Dc5Pf18vvxsaZs/eHqSg\nK0TlZ7fzh//DoJ35EeF4kjKs9N8mva+0xp3VFiKrneMsBdVJv4VG0rKJGM1mqkcF1EMrWggRWYHx\ni06cx6Jvr2VkLF2d9vZ0c9xBc8ed06S9a1HKXaLuDsZ5/vBqvSrJ48ev5xx34m+hURSViNF2TPWo\ngHYjaxFl4fx+Zk1PtyUif98t9w5WDTurxSKusfnwBCqLniel8dbityzLj++/hfJwCziFqRAV0ElU\nsxg3p3QDFrus1A/mSLLoz+nOiPYtm1os1rL8+P5bKI9MBSzpu2QYBWZ2cuEStQiNygV38pOljPKc\nr2q+4khZ5cmGa6UFyoiy/Pj+WyiPai6ITwMXAg8QtKX/Unh7BvhtuaI1l0UnzpvQGTbqQOC0Hnni\nPLNiYvt6ezj1yP7cl9WtEMKURDW3Rj20agxtO5BpAZvZbQCSLjSzBbFN35W0slTJWoFK29+jjJtK\n1kp83PoDmNbVtbPN+9IV68ftu3j5OoYqXBZbto1y5S8fzlzIi6jM+mr3CIFWj5CZyuT1Ac+S9EIz\n2wAg6QBgVnliNZ+lK9YzUrHKMrLDMmsLOOWRZyU+clE8/u3d2TC4ZWf3h8p9l65YP0EBV57rLOLp\nPp0SIdDKETJTmbxREB8EbpV0q6TbgFuAs8oTq/lk1RZwGk8tK/EbNw1PaL8e33eyi0dx5e0RAs5k\nyJuI8UNJBwIHhUP3mtm28sRqPo1qS+3sIutSvpaV+O2jyaFm1Vqz5yX+HfAIAWcy5G1LPxNYBLzf\nzNYC+0l6famSNZm0Ft9Zrb+d+qlWMLuWGNfp05IX2qJ9kxaVero0YdE1jfh3oNVraDitTV4XxFeA\n7cBLw+cDwCdLkahFSIvxbETsZydS7VK+lpX4/eb00qWJBXCifZMSHJa+6TCWnnZYrvPb19uzs+DN\nlm2jExR3q0QITOUC+51C3kW4F5nZWyS9FcDMtrZ794njDprLN+7YmDjuFE+1S/laVuL32n0GsKtO\nQ9K+aYtKC+f3s39GEfOeLrFl++hOP/DQ8AhdAilYnOuWOPXI5i9Ydcri4FRHeSo4SvofgrZDt5vZ\nEZJeBFxhZkeVLWDRLNhjD1t55JFV91u1cSjRlzh9WjdH7NdXhmhtx+PPbGPjpmG2j44xfVo3+83p\n3akcKyn0814TJlIcfnitIgNwx4bEstIATOvuYnQsu3pal8QL586a+F4nKVct+Pe3uei22+6sCN1N\nJK8LYjHwQ2BfSZcDNwEfqV+81idtISdt3BnP489sY8Pglp2f1/bRMTYMbuHxZ5LXbpPcBl0S+81p\nLZdPNeULsMOMjZuauwjn39+pQS4LGEDSswlaEQm4w8ymZO+1BQsW2MqV1XNIXnTO91OjIH57wWvL\nEK2tOGbJzTWVUoQCExqOPTa4v/XW2l8LmS6ItOiYSgQ8sOR1hcpVC/V8/k5xSMplAedtynmTmR0P\nXJ8w1pZ4FMTkaNUmiJMl7/lvdhREqxfYdwIyXRCSdpM0B9hL0mxJc8Lb/gRdidsWj4KYHFM1PCtv\npEDU9n32zB56EloONVvRtWoLHmc81SzgvyPIeNsHuBN2ttB6CrioRLmajlsQk2Oqfn55M9h2mO10\nMbRqLYipcEXRyjTivFYrxvMfwH9I+oCZfb7QmVscL0AyOVrl86v1R5Q3gy1uybuiaz8aFcaXNw54\nh6Q+MxsCkDQbeKuZfaEwSVoQ/2FNjmZ/fvX8iPKkKU8FS96ZHNX66xVF3jC090TKF8DMngTeU5gU\njlMC9RTKqaZY3ZfaGTSqxkdeBdwdz3yT1A1ML1QSxymYeiMxslh04jyWrljv6b1tTqMWkfMq4B8C\nV0o6XtLxwBXhmOO0LGX8iLIKBjntQ6O6gORVwB8hqAH8vvB2E/B/C5XEcQqmjLZSXvu3M2hUGF/e\nesA7JH0VuNnM/NvmTB0a0FbKa/+2J41YRM6bCXcysJTA73uApMOBj7dzV2Rn6pO3rVRlqFqttHpy\nidO65A1DOw84CrgVwMzWhH3hCkfSUuAkgvrDvwXeGY/AiO33IPA0MAaM5sm7djqLPItwSaFqWfR0\na1zjTg9JcyZDXh/wiJltrhgrqyjCjcCfmdmhwG+AczL2Pc7MDnflO/Upo3h4nkW4pFC1LGZNn+bp\nvU5h5LWA10n6K4JwtAOBfwD+pwyBzOyG2NM7gNPKmMdpHcrKOsqTDl2r/3bz8AhrzjuhbpkcJ05e\nC/gDwMHANuCbwGYa0xX5XcAPUrYZcIOkOyWdmXUQSWdKWilp5eDgYOFCOpOjrM7CeVaya/Xfur/X\nKZKqFnCYdPFxM/sn4NwiJpX0I+B5CZvONbPvhPucC4wCl6cc5mVmNiDpOcCNku41sx8n7WhmlwCX\nQFAPeNJvwCmUMrOOqq1kJ1nJWbi/1ymSqgrYzMYkvazISc3sVVnbJb0DeD1wvKVUjDezgfD+MUnX\nESwSJipgp7VJq7/QCGszqWjQZFrWO04t5PUBr5a0HLga2BINmtm1RQsk6dUESR6vMLOtKfvMArrM\n7Onw8QnAx4uWxWkMzS5dWWklZ3XEWHT12p2vcZzJktcHvBvwBPBKghCxkwgs1DK4CNiDwK2wRtLF\nAJL2kfT9cJ/nAj+VtBb4BXC9mXlq9BRlKhUPH9lhLF6+rtliOG1CXh/wXWb2mQbIg5n9Scr474DX\nho83AIeVLUurFtpuR1qhdGV0rqsRtaR3nMlS1QI2szHgrQ2QpaWIQqO88Er7U3muHadR5HVB3C7p\nIkkvl3REdCtVsiZTVmiU03rUmowxe2ZPidI4nUTeRbjDw/v4QpcR+ITbkkYVZHaaTy3ntKdbnHfS\nwSVK43QSeauhHVe2IK1GM0OjphLt4CevJfRs6WmHTbn357QuuVwQkvaU9O9RNpmkCyXtWbZwzaRR\nBZmnMu3iJz/uoLmo+m6Ah585xZLXB3wpQeWxN4e3p4CvlCVUKzCVQqOaRTv4yZetHuCaOwd88c1p\nCnl9wC8ys1Njz8+XtKYMgVqJZodGtTqt7ifP4x5ZvHxd7gW4nrzmiuPkJO9XajiejizpGKA1fmVO\n02hU48J6ePyZbVXdI8tWD9QU0zu6owRBnY4mrwX8PuCymN/3SeAdpUjUQrTDAlOZNDuFOIuNm4ZT\n3SPx+g+1UKSbwr9bDuSPglgDHCbpWeHzp0qVqgUoq0ZtO5FUyKZVFMn20WS3Qtw90ixXiX+3nIi8\nPeH+Bfi3qDWQpNnAh83so2UK10yyFpj8R7KLVvWTT5/WnTged480q/KZf7eciLw+4NfE+7KZ2ZOE\ndRnalVZfYHKy2W9Ob9UwwqRQw0bg3y0nIq8C7pY0I3oiqReYkbH/lKeVF5ic6uy1+4yqYYRJoYZv\nP3q/1GNO784bLZyNf7eciLyLcJcDN0mKYn/fCVxWjkitQSsvMDn5yOMeSdrne2sfTYyOmDk9788l\nG/9uORF5F+H+Nay9G3Wy+ISZrShPrObTygtMTrlsTglNSxuvFf9uORG5/9LDgucdVfS8VReYnHJp\nRB0Q/245kN8H7Dgdg9cBcRpFMU4tx2kj3EXgNApXwI6TgLsInEaQNxHjbiZmYm4GVgKfNLMnihbM\ncRyn3clrAf8AGAO+GT4/HZgJ/B74KkGXZMdpG7xWg9MI8irgV5lZvAfc3ZJWmdkRkt5ehmCO0yyW\nrR5g0dVrGdkRXPQNDA2z6Oq1gNdqcIqllky4o6Inkv4ciJaJRwuXynGayOLl63Yq34iRHcbi5eua\nJJHTruS1gP8WuFTS7oAIOmK8W9Is4IKyhHOcZpBWI7iW2sGOk4e8mXC/BA6J6gGb2ebY5qvKEMxx\nHKfdqakpJ3ATQU2Itm/K6XQuXSk1d9LGHadeWq4pp6TFkgYkrQlviWUvJb1a0npJ90s6uwxZnM5k\nR0rri7Rxx6mXVm3K+Rkz+3TaRkndwH8Cfwk8AvxS0nIz+1WJMjkdQn9KLYh+LxfpFMxUbcp5FHC/\nmW0ws+3At4A3NFEep4047qC5NY07Tr3kVcDvA/5T0oOSHgIuAt5bnli8X9Jdki4N2x9V0g88HHv+\nSDiWiKQzJa2UtHJwcLBoWZ02YtnqAa74+cOJ22651787TrE0pSmnpB8Bz0vYdC7wX8AnCFKfPwFc\nCLxrMvOZ2SXAJQALFixwT56TSNQsc8ySvyLeMsgpmkwFLOlDKeMAmNm/1zOpmb2q+l4g6UvA9xI2\nDQD7xp4/PxxznLpJapYZx1sGOUVTzQLeoyFSxJC0t5k9Gj59I3BPwm6/BA6UdACB4j0d+KsGiei0\nKVkWbk+3vB6wUziZCtjMzm+UIDH+TdLhBC6IB4G/A5C0D/DfZvZaMxuV9H5gBUFK9KVm5nmiBdDJ\nRWiy2tSPjbnnyimeluuIYWZ/bWaHmNmhZnZyZA2b2e/M7LWx/b5vZn9qZi8ys081T+L2IfKBDgwN\nYwRFaM659m6Wre4M705Wm/od4LUgnMLxguwdQh7LNskHOjwyxtIV6zvCCo7e41lXJoe4ey0Ip2ha\nzgJ2iievZZvmA+2k1f9O+KNxWoe6oiAi6o2CcBpLXsu2Ed2AHcfZRTULeI/wtoAgGaM/vL0XOCLj\ndU4Lkdey9W7AjtNYckVBSPoxcISZPR0+XwxcX7p0TiHktWy9G7DjNJa8i3DPBbbHnm8Px5wpwKIT\n53HOtXePc0OkWbbeDdhxGkdeBfw14BeSrgufLwQuK0ek1qFdYmLdsnWc1iRvLYhPSfoB8PJw6J1m\ntro8sZpPFDkQWY1R5ABMzZVyt2zz09fbkxpytmz1gH+OTmHUEoY2E3jKzP4DeCRMA25bsiIHnPZm\n8ckHp27z8+8USd6WROcBHwHOCYd6gG+UJVQr4DGxThJ+/p0iyWsBvxE4GdgCQVowTSjU00jSYl89\nJrb9Of+76SnHfv6dIsmrgLebmREUyCFsR9/WeExs5/Lk1vSUYz//TpHkjYK4StIXgT5J7yEokP6l\n8sRqPh454CTh598pkrxREJ+W9JcE3ZDnAR8zsxtLlawF8MgBx3HKJHc1tFDhtr3SdRwR+toSxh2n\nSPJGQZwi6T5JmyU9JelpSZPqC+c4rUpa6XUvye4UTV4L+N+Ak8zs12UK02q0SyacUxtpiRh9vT1N\nkMZpZ/Iq4D90ovJtp0w4Jz8jYztqGnecesmrgFdKuhJYBmyLBs3s2lKkagE6vTtEJ7Nle3Jn5LRx\nx6mXvAr4WcBW4ITYmAFtq4A9E85xnLLJG4b2zrIFaTW8O0TnIoElrLjJwyCcgsmlgCV9LmF4M7DS\nzL5TrEitwXEHzeUbd2xMHHfamyTlmzXuOPWSNxV5N+Bw4L7wdijwfODdkj5bkmxN5ZZ7B2sad9qH\n/pSrnLRxx6mXvAr4UOA4M/u8mX0eeBVwEEGRnhMyXzlFcR9w5+J1QJxGkXcRbjawO4HbAWAWMMfM\nxiRtS3/Z1MV9wLvotHhorwPiNIpaEjHWSLqVICPzL4B/Caui/agk2ZpKLX3U2plOjYf2OiBOI8gb\nBfFlSd8HjgqH/jmsCQywqEiBwnjjSMv1AUNmdnjCfg8CTwNjwKiZLShSDreCAjwe2nHKI1MBSzrI\nzO6VdEQ49HB4/zxJzzOzVUULZGZvic1/IbvcHkkcZ2aPFy2Dswv3hTtOeVSzgD8EnAlcmLDNgFcW\nLlGIJAFvLnOOLDr10rsS94U7TnlkRkGY2Znh/XEJt7IV48sJalDclyYecIOkOyWdmXUgSWdKWilp\n5eBgvjAyb8oZ4BEBjlMe1VwQfw48bGa/D5//DXAq8BCw2Mw21TOppB8Bz0vYdG4sseOtwBUZh3mZ\nmQ1Ieg5wo6R7zezHSTua2SXAJQALFizIFU7vl94B7gt3nPKo5oL4IkHML5L+AlgCfIAgKeMS4LR6\nJjWzV2VtlzQNOAU4MuMYA+H9Y5KuI1ggTFTA9eCX3rvwiADHKYdqiRjdMSv3LcAlZnaNmf0/4E9K\nlOtVwL1m9kjSRkmzJO0RPSZIBrmnSAHSUo49FdlxnKKoqoBDaxTgeODm2Lbc7Yzq4HQq3A+S9glD\n4QCeC/xU0lrgF8D1ZvbDIgXwVGTHccqmmhK9ArhN0uPAMPATAEl/QnZ42KQws3ckjP0OeG34eANw\nWFnzg/uAHccpn0wFbGafknQTsDdwg9nOelBdBL7gtsV9wI7jlE3VYjxmdoeZXWdmW2JjvykjCaOV\n8PArx3HKpkw/7pTGw68cxykbV8AZePhVZ9Jp1d+c5uEK2HHYpXQHhoYRQZoldG4KutMY8hZkd5y2\nJar7ES26VqZKdmIKutMYXAE7HU9S3Y9KPPzQKQNXwE7Hk0e5evihUwaugJ2Op5py9fBDpyxcATsd\nT1LMt8L7/r5eLjjlEF+Ac0rBoyCcjsdjvp1m4QrYcfCYb6c5uAvCcRynSbgCdhzHaRKugB3HcZqE\nK2DHcZwm4QrYcRynSbgCdhzHaRKugB3HcZqExwFn4HVhOxM/706jcAWcQlSiMKqS5XVhOwM/704j\ncRdECkklCr0ubPvj591pJK6AU/C29J2Jn3enkbgCTiGtRKHXhW1v/Lw7jcQVcArelr4z8fPuNJKm\nKWBJb5K0TtIOSQsqtp0j6X5J6yWdmPL6AyT9PNzvSknTi5Rv4fx+LjjlEPr7ehFeF7ZT8PPuNJJm\nRkHcA5wCfDE+KOnFwOnAwcA+wI8k/amZVTbt+lfgM2b2LUkXA+8G/qtIAb1EYWfi591pFE2zgM3s\n12aWtLT8BuBbZrbNzB4A7geOiu8gScArgW+HQ5cBC8uU13Ecp2ha0QfcDzwce/5IOBbn2cCQmY1m\n7LMTSWdKWilp5eDgYKHCOo7j1EupLghJPwKel7DpXDP7TplzxzGzS4BLABYsWGCNmtdxHCeLUhWw\nmb2qjpcNAPvGnj8/HIvzBNAnaVpoBSft4zh14+nITiNoRRfEcuB0STMkHQAcCPwivoOZGXALcFo4\ndAbQMIvaaW+idOSBoWGMXenIy1b7f7xTLM0MQ3ujpEeAlwLXS1oBYGbrgKuAXwE/BP4+ioCQ9H1J\n+4SH+AjwIUn3E/iEv9zo9+C0J56O7DSKpoWhmdl1wHUp2z4FfCph/LWxxxuoiI5wnCLwdGSnUbSi\nC8JxmoqnIzuNwhWw41Tg6chOo/B6wI5TQRTt4FEQTtm4AnacBDwd2WkE7oJwHMdpEq6AHcdxmoQr\nYMdxnCbhCthxHKdJuAJ2HMdpEq6AHcdxmoQrYMdxnCbhCthxHKdJuAJ2HMdpEgpK63YOkgaBh2p8\n2V7A4yWI4/P7/D5/e87/AjObW22njlPA9SBppZkt8Pl9fp/f5y8Sd0E4juM0CVfAjuM4TcIVcD4u\n8fl9fp/f5y8a9wE7juM0CbeAHcdxmoQrYMdxnCbhCjgFSUsl3SvpLknXSeqLbTtH0v2S1ks6saT5\n3yRpnaQdkhbExveXNCxpTXi7uJHzh9tKf/8J8iyWNBB736+t/qpJz/nq8D3eL+nssudLkeFBSXeH\n73llA+a7VNJjku6Jjc2RdKOk+8L72Q2ev2HnXtK+km6R9Kvw+/+P4Xg5n4GZ+S3hBpwATAsf/yvw\nr+HjFwNrgRnAAcBvge4S5v9fwDzgVmBBbHx/4J4GvP+0+Rvy/hPkWQz8UwPPf3f43l4ITA/f84ub\n8D18ENirgfP9BXBE/DsG/Btwdvj47Oi30MD5G3bugb2BI8LHewC/Cb/zpXwGbgGnYGY3mNlo+PQO\n4Pnh4zcA3zKzbWb2AHA/cFQJ8//azNYXfdwC5m/I+28BjgLuN7MNZrYd+BbBe29rzOzHwKaK4TcA\nl4WPLwMWNnj+hmFmj5rZqvDx08CvgX5K+gxcAefjXcAPwsf9wMOxbY+EY43kAEmrJd0m6eUNnruZ\n7//9oUvo0jIvg0Na4TwDGHCDpDslndmE+QGea2aPho9/Dzy3CTI08twDgbsPmA/8nJI+g47uiizp\nR8DzEjada2bfCfc5FxgFLm/G/Ak8CuxnZk9IOhJYJulgM3uqQfOXRpY8wH8BnyBQSJ8ALiT4Y2x3\nXmZmA5KeA9wo6d7QSmwKZmaSGh272vBzL2l34BrgLDN7StLObUV+Bh2tgM3sVVnbJb0DeD1wvIXO\nH2AA2De22/PDscLnT3nNNmBb+PhOSb8F/hSoeYGmnvkp8P3XK4+kLwHfK2LODEp7n7VgZgPh/WOS\nriNwjTRaAf9B0t5m9qikvYHHGjm5mf0hetyIcy+ph0D5Xm5m14bDpXwG7oJIQdKrgf8LnGxmW2Ob\nlgOnS5oh6QDgQOAXDZRrrqTu8PELw/k3NGp+mvT+wy99xBuBe9L2LYhfAgdKOkDSdOB0gvfeMCTN\nkrRH9JhgYbjs953EcuCM8PEZQEOvjhp57hWYul8Gfm1m/x7bVM5n0IiVxal4I1hcehhYE94ujm07\nl2CFfD3wmpLmfyOB33Eb8AdgRTh+KrAulGkVcFIj52/U+0+Q5+vA3cBd4Y9h7wbM+VqCVfDfErhl\nGv0dfCFB9MXa8JyXLgNwBYGbayQ8/+8Gng3cBNwH/AiY0+D5G3bugZcRuDruiv32X1vWZ+CpyI7j\nOE3CXRCO4zhNwhWw4zhOk3AF7DiO0yRcATuO4zQJV8CO4zhNwhWw4zhOk3AF3IZIOjcspXdXWL7v\nJeH4f0t6cclzL5b0T3nHpxqSzpI0s8bXHCup7My9pHmfKeg48XKQ90m6Nv49ktQjaUm4bZWkn0l6\nTWz74ZIsTG5yYrgCbjMkvZQgffoIMzsUeBVhURkz+1sz+1Uz5WsmkqZlPc/JWUBNCrhN+IyZHW5m\nBwJXAjdLmhtu+wRBGcc/M7MjCCqF7RF77VuBn4b3TgxXwO3H3sDjFtSMwMweN7PfAUi6VWFxdUnv\nlvQbSb+Q9CVJF4XjX5X0OUn/I2mDpNOiA0taJOmXoWV9fmz83PBYPyWoIZxJaBHdoV3F7meH438e\ns9qXKlaUu+L1H1FQpHytpCVVjnmrpM8qKGb+j+H7u1jSz4F/C9N9Lw0/h9WS3hC+rlvSpyXdEx7z\nA5L+AdgHuEXSLeF+J4QW3ypJV4dFXKJi7vdKWgWckvI+3iHpO6GM90k6L+MzWyLp72PPF0v6J0m7\nS7opnP/uSP6K146zwCVdpKDOCZKOVFBV705JKyrSfhMxsyuBG4C/Cq8G3gN8IPad+4OZXRUeX8Cb\ngHcAfylpt2rH7yganV7pt3JvwO4E6ZO/Ab4AvCK27VZgAYESeRCYA/QAPwEuCvf5KnA1wZ/ziwlq\n4kJQh+ASQOG27xEUzz6SIE10JvAsghTuCcWziRXVJkjzfEX4+OPAZ8PH9wAvDR8vIaHwPPAa4H+A\nmeHzOVWOeSvwhdjrvxrK3h0+/xfg7eHjvvBzmwW8D/g2u4ryR/M8SFggHdiLoDDOrPD5R4CPAbsR\nXHUcGH5eVwHfS3gv7yBIu3020Bu+/wWV+4X7zgduiz3/FUGxoGnAs2Ly3M+uZrvPhPfHxucHLgrn\n7gk/y7nh+FuAS7POXWzsLIIqZYcCqzO+j8cAN4WPvwmc2uzfSCvd3AJuM8zsGQKleCYwCFwZWTsx\njiL4MW8ysxEChRtnmZntsMBdEdU9PSG8rSaoQXEQgYJ5OXCdmW21oCRmZsEaSXsCfWZ2Wzh0GfAX\nClo+7WFmPwvHv5lyiFcBX7GwQJKZbUo7Zuw1V1Yc42ozG4u9r7MlrSFQ1rsB+4XzfNHCovxmllQk\n/GiCP6nbw9efAbyA4LN5wMzus0DzfCP9E+FGM3vCzIaBawlqEUzAzFYDz5G0j6TDgCfN7GECBf8v\nku4iqFHQT/5atfOAPyMoc7kG+Ci7Gg9UQ9V3AQK3w7fCx9/C3RDj6OhylO1KqFxuBW6VdDeBYvhq\nDYfYFnus2P0FZvbF+I6Szqpf0oaxJeO5CKyycd0/pFz6RQQKdJxSkXR4DbJVFmPJKs5yNXAaQc3k\n6McaWGIAAAK0SURBVE/lbcBc4EgzG5H0IMGfSJxRxrsbo+0C1pnZS2uQN2I+QQnU+4H9JD3LKmpS\nK6jadyrwBgV1tQU8W9IeFnSb6HjcAm4zJM2TdGBs6HDgoYrdfgm8QtLscCHq1ByHXgG8K+bj7FdQ\nJPzHwEJJvQpKJ56UdRAz2ww8qV2dPP6awBofAp5WGLFBUP4xiRuBd4a+RyTNSTtmjvcUva8PhL5K\nJM2PzfN34eeDpDnh+NPsWmC6AzhG0p+E+8yS9KfAvcD+kl4U7pdl9f2lgoaPvQSLV7dn7Hslwedy\nGruuWvYEHguV73EEFnglDwEvVlBCtA84PhxfD8xVsHAbRTMcnDE/4X6nElw5XBFeiXwZ+A8FZTuj\nkqlvCue5y8z2NbP9zewFBHV231htjk7BLeD2Y3fg8+EPbZTAQhnXysaCDgv/QlDHdxOBwticdVAz\nu0HS/wJ+FuqqZwh8p6skXUlQMvExAuVejTOAi0MlugF4Zzj+buBLknYQKNAJMpnZD0MLc6Wk7cD3\ngX/OOGY1PgF8FrhLUhfwAEEUyX8TFLq/S9II8CUC3+klwA8l/c7MjgvdO1dImhEe76Nm9hsF7YOu\nl7SVwMe+B8n8gkApPR/4hpmlFtY3s3Xhn9yA7WqPcznw3fBKZyXBuax83cOSriLwMT9A4EbCzLYr\nWGT9XOjGmRZ+FusSpv+gpLcT+MfvAV5pZoPRewY+CfxK0h8JrjA+RvDHc13Fca4h8K9/Le19dhJe\njrJDkbS7mT0TWnjXESy+VP5YmiJT+Phsgrqv/9hMmcokVN4LzOz9zZbFaQ5uAXcuiyW9isAfeAOw\nrMnyALxO0jkE38uHCFbqHadtcQvYcVoISVHnhUqON7MnGi2PUy6ugB3HcZqER0E4juM0CVfAjuM4\nTcIVsOM4TpNwBew4jtMk/j+WwTZQW79wUQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fd4d1488610>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "alg = \"DCA\"\n",
    "estim = deseq_2_DCA\n",
    "p, fc = estim.fit(return_fc=True)\n",
    "\n",
    "p = np.array(stats.p_adjust(FloatVector(p), method = 'BH'))\n",
    "signed_p = - np.log10(p) * np.sign(fc)\n",
    "\n",
    "signed_p[signed_p >= 20] = 20\n",
    "signed_p[signed_p <= -20] = -20\n",
    "\n",
    "plt.figure(figsize=(5, 5))\n",
    "plt.scatter(signed_p, signed_p_value)\n",
    "plt.axvline(x=3, c=\"r\")\n",
    "plt.axvline(x=-3, c=\"r\")\n",
    "plt.axhline(y=np.log10(0.05), c=\"r\")\n",
    "plt.axhline(y=-np.log10(0.05), c=\"r\")\n",
    "plt.xlabel(\"Signed log corrected p_value \"+alg)\n",
    "plt.ylabel(\"Signed log corrected p_value micro-array\")\n",
    "plt.tight_layout()\n",
    "plt.savefig(interest + \"_\"+ alg + \"_vs_micro_array.pdf\", dpi=300)\n",
    "print interest + \"_\" + \"Detected \", np.sum(p <= 0.05), \" genes with \" + alg\n",
    "print interest + \"_\" + \"Detected \", np.sum(p_value <= 0.05), \" genes with microarray\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ubuntu/tensorflow/lib/python2.7/site-packages/ipykernel_launcher.py:6: RuntimeWarning: invalid value encountered in log\n",
      "  \n",
      "benchmarking.py:87: RuntimeWarning: invalid value encountered in log\n",
      "  estimated_score = -np.log(estimated_score)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1\n",
      "2\n",
      "3\n",
      "4\n",
      "5\n",
      "6\n",
      "7\n",
      "8\n",
      "9\n",
      "10\n",
      "11\n",
      "12\n",
      "13\n",
      "14\n",
      "15\n",
      "16\n",
      "17\n",
      "18\n",
      "19\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[[0.20579102816569819, 0.50700714976269023, 0.67157251709257715],\n",
       " [0.18058755047714981, 0.49782412849795693, 0.66170187368282396],\n",
       " [0.18205883823373051, 0.46225563725478108, 0.66578333686665248],\n",
       " [0.2278194853744466, 0.50342213375738987, 0.67223089285162407],\n",
       " [0.22812672997825545, 0.54917835812372995, 0.66843008987057928],\n",
       " [0.22062963275894731, 0.43697002174118799, 0.69418578460456637],\n",
       " [0.21857668810339298, 0.66953554518025582, 0.65516347380074835],\n",
       " [0.1866262645253581, 0.5108688836529699, 0.6555767614511836],\n",
       " [0.22643630937751896, 0.60042451069729108, 0.6646308859017338],\n",
       " [0.2168311220627111, 0.50005428379919303, 0.66589880188473105],\n",
       " [0.24242353043242065, 0.56622430811054847, 0.69005335669347778],\n",
       " [0.17019556764117594, 0.44406928853146871, 0.65805407155577145],\n",
       " [0.22190486950121596, 0.61340128907657154, 0.68259697791406981],\n",
       " [0.20105116498883821, 0.5409368547746497, 0.65572889619756503],\n",
       " [0.18637336295159221, 0.44811322631735973, 0.66199251735432885],\n",
       " [0.17266588007687542, 0.47935498169189567, 0.65158790170132319],\n",
       " [0.22436243141291032, 0.5844902379268978, 0.67046648176560841],\n",
       " [0.18481191001782554, 0.50832858270150993, 0.66454407421942852],\n",
       " [0.2036333998475498, 0.57119740037876809, 0.65944604128560225],\n",
       " [0.21106741286959002, 0.57123566295546579, 0.67129713231833132]]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "boostrap_scores_de(p_value, deseq_2_DCA, 20, p_prior=p_prior, rank_auc=rank_auc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}