notebooks/demo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "1bf658eb-8bdf-465d-ab25-29dd59f235e5",
   "metadata": {},
   "source": [
    "# Load data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "35d86e2c-7ded-48a9-9c1a-35708b1bfc49",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "0211937f-81df-4b5f-8719-f439731de063",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load dataframes\n",
    "df_raw = pd.read_csv('../data/v1_raw.csv', header=0, index_col=0, delimiter=',')\n",
    "df_cpm = pd.read_csv('../data/v1_cpm.csv', header=0, index_col=0, delimiter=',')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "0133a1cc-bdc0-4845-9e07-bbd78b89c0e7",
   "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>0610007P14Rik</th>\n",
       "      <th>0610009B22Rik</th>\n",
       "      <th>0610010F05Rik</th>\n",
       "      <th>0610010K14Rik</th>\n",
       "      <th>0610011F06Rik</th>\n",
       "      <th>0610012G03Rik</th>\n",
       "      <th>0610030E20Rik</th>\n",
       "      <th>0610037L13Rik</th>\n",
       "      <th>1110001J03Rik</th>\n",
       "      <th>1110004E09Rik</th>\n",
       "      <th>...</th>\n",
       "      <th>Zxdb</th>\n",
       "      <th>Zxdc</th>\n",
       "      <th>Zyg11b</th>\n",
       "      <th>Zyx</th>\n",
       "      <th>Zzef1</th>\n",
       "      <th>Zzz3</th>\n",
       "      <th>l7Rn6</th>\n",
       "      <th>cell_types_98</th>\n",
       "      <th>cell_types_50</th>\n",
       "      <th>cell_types_25</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>LS-15006_S09_E1-50</th>\n",
       "      <td>79.0</td>\n",
       "      <td>140.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>114.0</td>\n",
       "      <td>575.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>108.0</td>\n",
       "      <td>206.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>78.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>36.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>190.0</td>\n",
       "      <td>Vip Arhgap36 Hmcn1</td>\n",
       "      <td>n70</td>\n",
       "      <td>n66</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LS-15006_S10_E1-50</th>\n",
       "      <td>121.0</td>\n",
       "      <td>175.0</td>\n",
       "      <td>181.0</td>\n",
       "      <td>196.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>318.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>202.0</td>\n",
       "      <td>48.0</td>\n",
       "      <td>75.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>29.0</td>\n",
       "      <td>178.0</td>\n",
       "      <td>Lamp5 Lsp1</td>\n",
       "      <td>Lamp5 Lsp1</td>\n",
       "      <td>n78</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LS-15006_S11_E1-50</th>\n",
       "      <td>89.0</td>\n",
       "      <td>68.0</td>\n",
       "      <td>100.0</td>\n",
       "      <td>75.0</td>\n",
       "      <td>14.0</td>\n",
       "      <td>170.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>101.0</td>\n",
       "      <td>88.0</td>\n",
       "      <td>35.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>153.0</td>\n",
       "      <td>36.0</td>\n",
       "      <td>104.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>153.0</td>\n",
       "      <td>Lamp5 Lsp1</td>\n",
       "      <td>Lamp5 Lsp1</td>\n",
       "      <td>n78</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LS-15006_S12_E1-50</th>\n",
       "      <td>115.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>35.0</td>\n",
       "      <td>17.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>193.0</td>\n",
       "      <td>90.0</td>\n",
       "      <td>189.0</td>\n",
       "      <td>117.0</td>\n",
       "      <td>44.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>31.0</td>\n",
       "      <td>162.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>110.0</td>\n",
       "      <td>Vip Crispld2 Htr2c</td>\n",
       "      <td>n65</td>\n",
       "      <td>n64</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LS-15006_S13_E1-50</th>\n",
       "      <td>390.0</td>\n",
       "      <td>177.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>42.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>71.0</td>\n",
       "      <td>52.0</td>\n",
       "      <td>162.0</td>\n",
       "      <td>129.0</td>\n",
       "      <td>...</td>\n",
       "      <td>20.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>85.0</td>\n",
       "      <td>44.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>335.0</td>\n",
       "      <td>Lamp5 Plch2 Dock5</td>\n",
       "      <td>n82</td>\n",
       "      <td>n79</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 10003 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                    0610007P14Rik  0610009B22Rik  0610010F05Rik  \\\n",
       "LS-15006_S09_E1-50           79.0          140.0            0.0   \n",
       "LS-15006_S10_E1-50          121.0          175.0          181.0   \n",
       "LS-15006_S11_E1-50           89.0           68.0          100.0   \n",
       "LS-15006_S12_E1-50          115.0            0.0           35.0   \n",
       "LS-15006_S13_E1-50          390.0          177.0            0.0   \n",
       "\n",
       "                    0610010K14Rik  0610011F06Rik  0610012G03Rik  \\\n",
       "LS-15006_S09_E1-50           30.0          114.0          575.0   \n",
       "LS-15006_S10_E1-50          196.0           30.0          318.0   \n",
       "LS-15006_S11_E1-50           75.0           14.0          170.0   \n",
       "LS-15006_S12_E1-50           17.0            0.0          193.0   \n",
       "LS-15006_S13_E1-50           42.0            0.0           23.0   \n",
       "\n",
       "                    0610030E20Rik  0610037L13Rik  1110001J03Rik  \\\n",
       "LS-15006_S09_E1-50            7.0          108.0          206.0   \n",
       "LS-15006_S10_E1-50            3.0          202.0           48.0   \n",
       "LS-15006_S11_E1-50            0.0          101.0           88.0   \n",
       "LS-15006_S12_E1-50           90.0          189.0          117.0   \n",
       "LS-15006_S13_E1-50           71.0           52.0          162.0   \n",
       "\n",
       "                    1110004E09Rik  ...  Zxdb  Zxdc  Zyg11b    Zyx  Zzef1  \\\n",
       "LS-15006_S09_E1-50            0.0  ...   0.0   0.0    78.0    0.0   36.0   \n",
       "LS-15006_S10_E1-50           75.0  ...   0.0   0.0    28.0    0.0    1.0   \n",
       "LS-15006_S11_E1-50           35.0  ...   0.0   0.0   153.0   36.0  104.0   \n",
       "LS-15006_S12_E1-50           44.0  ...   0.0   0.0    31.0  162.0    0.0   \n",
       "LS-15006_S13_E1-50          129.0  ...  20.0   0.0    85.0   44.0    0.0   \n",
       "\n",
       "                    Zzz3  l7Rn6       cell_types_98  cell_types_50  \\\n",
       "LS-15006_S09_E1-50   0.0  190.0  Vip Arhgap36 Hmcn1            n70   \n",
       "LS-15006_S10_E1-50  29.0  178.0          Lamp5 Lsp1     Lamp5 Lsp1   \n",
       "LS-15006_S11_E1-50  21.0  153.0          Lamp5 Lsp1     Lamp5 Lsp1   \n",
       "LS-15006_S12_E1-50  30.0  110.0  Vip Crispld2 Htr2c            n65   \n",
       "LS-15006_S13_E1-50   9.0  335.0   Lamp5 Plch2 Dock5            n82   \n",
       "\n",
       "                    cell_types_25  \n",
       "LS-15006_S09_E1-50            n66  \n",
       "LS-15006_S10_E1-50            n78  \n",
       "LS-15006_S11_E1-50            n78  \n",
       "LS-15006_S12_E1-50            n64  \n",
       "LS-15006_S13_E1-50            n79  \n",
       "\n",
       "[5 rows x 10003 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Preview raw\n",
    "df_raw.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "642c911e-da5c-4702-bf4c-7c17f5e58fee",
   "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>0610007P14Rik</th>\n",
       "      <th>0610009B22Rik</th>\n",
       "      <th>0610010F05Rik</th>\n",
       "      <th>0610010K14Rik</th>\n",
       "      <th>0610011F06Rik</th>\n",
       "      <th>0610012G03Rik</th>\n",
       "      <th>0610030E20Rik</th>\n",
       "      <th>0610037L13Rik</th>\n",
       "      <th>1110001J03Rik</th>\n",
       "      <th>1110004E09Rik</th>\n",
       "      <th>...</th>\n",
       "      <th>Zxdb</th>\n",
       "      <th>Zxdc</th>\n",
       "      <th>Zyg11b</th>\n",
       "      <th>Zyx</th>\n",
       "      <th>Zzef1</th>\n",
       "      <th>Zzz3</th>\n",
       "      <th>l7Rn6</th>\n",
       "      <th>cell_types_98</th>\n",
       "      <th>cell_types_50</th>\n",
       "      <th>cell_types_25</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>LS-15006_S09_E1-50</th>\n",
       "      <td>45.645770</td>\n",
       "      <td>80.891235</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>17.333836</td>\n",
       "      <td>65.868576</td>\n",
       "      <td>332.231870</td>\n",
       "      <td>4.044562</td>\n",
       "      <td>62.401810</td>\n",
       "      <td>119.025670</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>45.067974</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>20.800604</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>109.780960</td>\n",
       "      <td>Vip Arhgap36 Hmcn1</td>\n",
       "      <td>n70</td>\n",
       "      <td>n66</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LS-15006_S10_E1-50</th>\n",
       "      <td>63.363390</td>\n",
       "      <td>91.641266</td>\n",
       "      <td>94.783260</td>\n",
       "      <td>102.638220</td>\n",
       "      <td>15.709931</td>\n",
       "      <td>166.525280</td>\n",
       "      <td>1.570993</td>\n",
       "      <td>105.780205</td>\n",
       "      <td>25.135891</td>\n",
       "      <td>39.274830</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>14.662603</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.523664</td>\n",
       "      <td>15.186268</td>\n",
       "      <td>93.212265</td>\n",
       "      <td>Lamp5 Lsp1</td>\n",
       "      <td>Lamp5 Lsp1</td>\n",
       "      <td>n78</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LS-15006_S11_E1-50</th>\n",
       "      <td>44.837242</td>\n",
       "      <td>34.257668</td>\n",
       "      <td>50.378925</td>\n",
       "      <td>37.784195</td>\n",
       "      <td>7.053050</td>\n",
       "      <td>85.644170</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>50.882713</td>\n",
       "      <td>44.333454</td>\n",
       "      <td>17.632624</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>77.079760</td>\n",
       "      <td>18.136414</td>\n",
       "      <td>52.394080</td>\n",
       "      <td>10.579575</td>\n",
       "      <td>77.079760</td>\n",
       "      <td>Lamp5 Lsp1</td>\n",
       "      <td>Lamp5 Lsp1</td>\n",
       "      <td>n78</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LS-15006_S12_E1-50</th>\n",
       "      <td>50.184307</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>15.273485</td>\n",
       "      <td>7.418550</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>84.222360</td>\n",
       "      <td>39.274673</td>\n",
       "      <td>82.476820</td>\n",
       "      <td>51.057080</td>\n",
       "      <td>19.200953</td>\n",
       "      <td>...</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>13.527944</td>\n",
       "      <td>70.694410</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>13.091558</td>\n",
       "      <td>48.002380</td>\n",
       "      <td>Vip Crispld2 Htr2c</td>\n",
       "      <td>n65</td>\n",
       "      <td>n64</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LS-15006_S13_E1-50</th>\n",
       "      <td>221.910780</td>\n",
       "      <td>100.713356</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>23.898085</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>13.087047</td>\n",
       "      <td>40.399143</td>\n",
       "      <td>29.588104</td>\n",
       "      <td>92.178330</td>\n",
       "      <td>73.401260</td>\n",
       "      <td>...</td>\n",
       "      <td>11.38004</td>\n",
       "      <td>0.0</td>\n",
       "      <td>48.365173</td>\n",
       "      <td>25.036089</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>5.121018</td>\n",
       "      <td>190.615680</td>\n",
       "      <td>Lamp5 Plch2 Dock5</td>\n",
       "      <td>n82</td>\n",
       "      <td>n79</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 10003 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                    0610007P14Rik  0610009B22Rik  0610010F05Rik  \\\n",
       "LS-15006_S09_E1-50      45.645770      80.891235       0.000000   \n",
       "LS-15006_S10_E1-50      63.363390      91.641266      94.783260   \n",
       "LS-15006_S11_E1-50      44.837242      34.257668      50.378925   \n",
       "LS-15006_S12_E1-50      50.184307       0.000000      15.273485   \n",
       "LS-15006_S13_E1-50     221.910780     100.713356       0.000000   \n",
       "\n",
       "                    0610010K14Rik  0610011F06Rik  0610012G03Rik  \\\n",
       "LS-15006_S09_E1-50      17.333836      65.868576     332.231870   \n",
       "LS-15006_S10_E1-50     102.638220      15.709931     166.525280   \n",
       "LS-15006_S11_E1-50      37.784195       7.053050      85.644170   \n",
       "LS-15006_S12_E1-50       7.418550       0.000000      84.222360   \n",
       "LS-15006_S13_E1-50      23.898085       0.000000      13.087047   \n",
       "\n",
       "                    0610030E20Rik  0610037L13Rik  1110001J03Rik  \\\n",
       "LS-15006_S09_E1-50       4.044562      62.401810     119.025670   \n",
       "LS-15006_S10_E1-50       1.570993     105.780205      25.135891   \n",
       "LS-15006_S11_E1-50       0.000000      50.882713      44.333454   \n",
       "LS-15006_S12_E1-50      39.274673      82.476820      51.057080   \n",
       "LS-15006_S13_E1-50      40.399143      29.588104      92.178330   \n",
       "\n",
       "                    1110004E09Rik  ...      Zxdb  Zxdc     Zyg11b        Zyx  \\\n",
       "LS-15006_S09_E1-50       0.000000  ...   0.00000   0.0  45.067974   0.000000   \n",
       "LS-15006_S10_E1-50      39.274830  ...   0.00000   0.0  14.662603   0.000000   \n",
       "LS-15006_S11_E1-50      17.632624  ...   0.00000   0.0  77.079760  18.136414   \n",
       "LS-15006_S12_E1-50      19.200953  ...   0.00000   0.0  13.527944  70.694410   \n",
       "LS-15006_S13_E1-50      73.401260  ...  11.38004   0.0  48.365173  25.036089   \n",
       "\n",
       "                        Zzef1       Zzz3       l7Rn6       cell_types_98  \\\n",
       "LS-15006_S09_E1-50  20.800604   0.000000  109.780960  Vip Arhgap36 Hmcn1   \n",
       "LS-15006_S10_E1-50   0.523664  15.186268   93.212265          Lamp5 Lsp1   \n",
       "LS-15006_S11_E1-50  52.394080  10.579575   77.079760          Lamp5 Lsp1   \n",
       "LS-15006_S12_E1-50   0.000000  13.091558   48.002380  Vip Crispld2 Htr2c   \n",
       "LS-15006_S13_E1-50   0.000000   5.121018  190.615680   Lamp5 Plch2 Dock5   \n",
       "\n",
       "                    cell_types_50  cell_types_25  \n",
       "LS-15006_S09_E1-50            n70            n66  \n",
       "LS-15006_S10_E1-50     Lamp5 Lsp1            n78  \n",
       "LS-15006_S11_E1-50     Lamp5 Lsp1            n78  \n",
       "LS-15006_S12_E1-50            n65            n64  \n",
       "LS-15006_S13_E1-50            n82            n79  \n",
       "\n",
       "[5 rows x 10003 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Preview CPM\n",
    "df_cpm.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "2a02758d-6a6a-404b-93c3-1396db953885",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Extract cell types\n",
    "cell_types_98 = df_raw['cell_types_98']\n",
    "cell_types_50 = df_raw['cell_types_50']\n",
    "cell_types_25 = df_raw['cell_types_25']\n",
    "\n",
    "# Convert to numerical\n",
    "cell_types_98_num = pd.Categorical(cell_types_98).codes\n",
    "cell_types_50_num = pd.Categorical(cell_types_50).codes\n",
    "cell_types_25_num = pd.Categorical(cell_types_25).codes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "011f32bf-0e86-4829-b44f-3d7f463d27bb",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Arrange raw and CPM data\n",
    "raw = df_raw.values[:, :-3].astype(np.float32)\n",
    "cpm = df_cpm.values[:, :-3].astype(np.float32)\n",
    "gene_names = np.array(df_raw.columns)[:-3]\n",
    "\n",
    "# Generate logarithmized and binarized data\n",
    "binary = (raw > 0).astype(np.float32)\n",
    "log = np.log(1 + raw)\n",
    "logcpm = np.log(1 + cpm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "3d90a8a0-ca83-49c7-9724-500eaad64a14",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "13349 total examples, 10679 training examples, 1336 validation examples, 1334 test examples\n"
     ]
    }
   ],
   "source": [
    "# For data splitting\n",
    "n = len(raw)\n",
    "n_train = int(0.8 * n)\n",
    "n_test = int(0.1 * n)\n",
    "all_rows = np.arange(n)\n",
    "np.random.seed(0)\n",
    "np.random.shuffle(all_rows)\n",
    "train_inds = all_rows[:n_train]\n",
    "val_inds = all_rows[n_train:-n_test]\n",
    "test_inds = all_rows[-n_test:]\n",
    "print(f'{n} total examples, {len(train_inds)} training examples, {len(val_inds)} validation examples, {len(test_inds)} test examples')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f57f9665-29d6-476a-90d6-ace2b5711235",
   "metadata": {},
   "source": [
    "# Run PROPOSE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "3d766536-f9f6-4e82-aedc-2ee06aeaf3b5",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import pickle\n",
    "import numpy as np\n",
    "import torch.nn as nn\n",
    "from propose import PROPOSE, HurdleLoss, ExpressionDataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "f723348a-705a-4c06-81e4-cf79909bec39",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set up datasets\n",
    "train_dataset = ExpressionDataset(binary[train_inds], logcpm[train_inds])\n",
    "val_dataset = ExpressionDataset(binary[val_inds], logcpm[val_inds])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "e0e2a58f-b95a-46d2-a058-67023174c1d2",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set up GPU device\n",
    "device = torch.device('cuda', 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "46e3bbc5-fcbd-43f6-b836-42f57dc4e69d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Number of genes to select\n",
    "num_genes = (32, 64, 128)\n",
    "propose_results = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "3b9c327b-130d-41dc-a263-34a3bff80df5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "using HurdleLoss, starting with lam = 0.01\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "324c465c2ae249bc8b0515f252ad2339",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/500 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lam = 0.010000 yielded 10000 genes\n",
      "Warm starting model for next iteration\n",
      "next attempt is lam = 0.100000\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "af1055fdb86a44479588546902e12dd6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/500 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lam = 0.100000 yielded 6914 genes\n",
      "Warm starting model for next iteration\n",
      "next attempt is lam = 1.000000\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a7e2db422c7a4db1920a63ea87c5e11c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/500 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lam = 1.000000 yielded 495 genes\n",
      "done, lam = 1.000000 yielded 495 genes\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1f464c398dbd4c85be3b08870e124b27",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/500 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "done, selected 32 genes\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1354809243e54068857eb33be08a0632",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/500 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "done, selected 64 genes\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f1f97be981664ac794b131a9844a1ee3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/500 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "done, selected 128 genes\n"
     ]
    }
   ],
   "source": [
    "# Set up selector\n",
    "selector = PROPOSE(train_dataset,\n",
    "                   val_dataset,\n",
    "                   loss_fn=HurdleLoss(),\n",
    "                   device=device,\n",
    "                   hidden=[128, 128])\n",
    "\n",
    "# Eliminate many candidates\n",
    "candidates, model = selector.eliminate(target=500, mbsize=128, max_nepochs=500)\n",
    "\n",
    "for num in num_genes:\n",
    "    # Select specific number of genes\n",
    "    inds, model = selector.select(num_genes=num, mbsize=128, max_nepochs=500)\n",
    "    propose_results[num] = inds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "0236795f-9512-444f-afd8-8c5a52f0a527",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save results\n",
    "with open('propose_results.pkl', 'wb') as f:\n",
    "    pickle.dump(propose_results, f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "69ac9e9e-5305-4f9f-99bc-bfcbaafc19a1",
   "metadata": {},
   "source": [
    "# Run PROPOSE for cell type classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "9c5b4495-edba-4105-89c4-a431b7cb8dff",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import pickle\n",
    "import numpy as np\n",
    "import torch.nn as nn\n",
    "from propose import PROPOSE, ExpressionDataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "7dae6048-608b-4bcb-9f90-90e9e3b4fc1c",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Make cell types numerical\n",
    "cell_types = pd.Categorical(cell_types_98).codes\n",
    "\n",
    "# Set up datasets\n",
    "train_dataset = ExpressionDataset(binary[train_inds], cell_types[train_inds])\n",
    "val_dataset = ExpressionDataset(binary[val_inds], cell_types[val_inds])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "a29bae6e-b894-4fa2-99e1-ede29543bd04",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set up GPU device\n",
    "device = torch.device('cuda', 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "d7f225da-900b-4f0c-afcd-a7583f951fdd",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Number of genes to select\n",
    "num_genes = (32, 64, 128)\n",
    "propose_class_results = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "cd4b53b8-e274-486c-98e4-e0af7114ec3b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "using CrossEntropyLoss, starting with lam = 0.0001\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4190ebad58074c6c86a291db4bce2d7a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/600 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lam = 0.000100 yielded 606 genes\n",
      "Warm starting model for next iteration\n",
      "next attempt is lam = 0.000123\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dfe48af593e94d75a3893625e8e8c7e5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/600 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lam = 0.000123 yielded 341 genes\n",
      "Reinitializing model for next iteration\n",
      "next attempt is lam = 0.000090\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a6206b23951449ccadfffce96eb4ab1f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/600 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lam = 0.000090 yielded 677 genes\n",
      "Warm starting model for next iteration\n",
      "next attempt is lam = 0.000113\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "48efc9b1a75342f69e356de992596bb1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/600 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lam = 0.000113 yielded 369 genes\n",
      "Reinitializing model for next iteration\n",
      "next attempt is lam = 0.000091\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1f2ca8827139467090e9f094171b4a11",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/600 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lam = 0.000091 yielded 690 genes\n",
      "Warm starting model for next iteration\n",
      "next attempt is lam = 0.000111\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dc54b135d06c401ea253fc01e57b997d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/600 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lam = 0.000111 yielded 403 genes\n",
      "done, lam = 0.000111 yielded 403 genes\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "36a72111c7524b5ebca15e8c6be64cb1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/600 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "done, selected 32 genes\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6abc81f2d4484277b1827b184bdfd08e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/600 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "done, selected 64 genes\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9e4358db942d4f439bb625a2646a3835",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training epochs:   0%|          | 0/600 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "done, selected 128 genes\n"
     ]
    }
   ],
   "source": [
    "# Set up selector\n",
    "selector = PROPOSE(train_dataset,\n",
    "                   val_dataset,\n",
    "                   loss_fn=torch.nn.CrossEntropyLoss(),\n",
    "                   device=device,\n",
    "                   hidden=[128, 128])\n",
    "\n",
    "# Eliminate many candidates\n",
    "candidates, model = selector.eliminate(target=500, mbsize=128, max_nepochs=600)\n",
    "\n",
    "for num in num_genes:\n",
    "    # Select specific number of genes\n",
    "    inds, model = selector.select(num_genes=num, mbsize=128, max_nepochs=600)\n",
    "    propose_class_results[num] = inds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "178750a0-da81-4d2d-b745-18984b43ce08",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save results\n",
    "with open('propose_class_results.pkl', 'wb') as f:\n",
    "    pickle.dump(propose_class_results, f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "440ba77a-6320-451d-bb19-8ecaebf6f01a",
   "metadata": {},
   "source": [
    "# Run Seurat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "dd6fa316-b788-4e40-a4d5-32adcae78244",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: scanpy in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (1.7.2)\n",
      "Requirement already satisfied: anndata in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (0.7.8)\n",
      "Requirement already satisfied: pandas>=1.1.1 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from anndata) (1.1.5)\n",
      "Requirement already satisfied: xlrd<2.0 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from anndata) (1.2.0)\n",
      "Requirement already satisfied: scipy>1.4 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from anndata) (1.5.4)\n",
      "Requirement already satisfied: packaging>=20 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from anndata) (20.9)\n",
      "Requirement already satisfied: natsort in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from anndata) (7.1.1)\n",
      "Requirement already satisfied: h5py in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from anndata) (3.1.0)\n",
      "Requirement already satisfied: numpy>=1.16.5 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from anndata) (1.19.2)\n",
      "Requirement already satisfied: importlib_metadata>=0.7 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from anndata) (2.0.0)\n",
      "Requirement already satisfied: zipp>=0.5 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from importlib_metadata>=0.7->anndata) (3.4.0)\n",
      "Requirement already satisfied: pyparsing>=2.0.2 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from packaging>=20->anndata) (2.4.7)\n",
      "Requirement already satisfied: pytz>=2017.2 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from pandas>=1.1.1->anndata) (2021.1)\n",
      "Requirement already satisfied: python-dateutil>=2.7.3 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from pandas>=1.1.1->anndata) (2.8.1)\n",
      "Requirement already satisfied: six>=1.5 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from python-dateutil>=2.7.3->pandas>=1.1.1->anndata) (1.15.0)\n",
      "Requirement already satisfied: statsmodels>=0.10.0rc2 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (0.12.2)\n",
      "Requirement already satisfied: umap-learn>=0.3.10 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (0.5.1)\n",
      "Requirement already satisfied: joblib in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (1.0.1)\n",
      "Requirement already satisfied: matplotlib>=3.1.2 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (3.3.4)\n",
      "Requirement already satisfied: tables in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (3.6.1)\n",
      "Requirement already satisfied: tqdm in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (4.59.0)\n",
      "Requirement already satisfied: legacy-api-wrap in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (1.2)\n",
      "Requirement already satisfied: seaborn in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (0.11.2)\n",
      "Requirement already satisfied: sinfo in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (0.3.4)\n",
      "Requirement already satisfied: scikit-learn>=0.21.2 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (0.24.2)\n",
      "Requirement already satisfied: patsy in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (0.5.1)\n",
      "Requirement already satisfied: networkx>=2.3 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (2.5.1)\n",
      "Requirement already satisfied: numba>=0.41.0 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scanpy) (0.52.0)\n",
      "Requirement already satisfied: cached-property in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from h5py->anndata) (1.5.2)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from matplotlib>=3.1.2->scanpy) (1.3.1)\n",
      "Requirement already satisfied: cycler>=0.10 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from matplotlib>=3.1.2->scanpy) (0.10.0)\n",
      "Requirement already satisfied: pillow>=6.2.0 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from matplotlib>=3.1.2->scanpy) (8.1.2)\n",
      "Requirement already satisfied: decorator<5,>=4.3 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from networkx>=2.3->scanpy) (4.4.2)\n",
      "Requirement already satisfied: llvmlite<0.36,>=0.35.0 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from numba>=0.41.0->scanpy) (0.35.0)\n",
      "Requirement already satisfied: setuptools in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from numba>=0.41.0->scanpy) (52.0.0.post20210125)\n",
      "Requirement already satisfied: threadpoolctl>=2.0.0 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from scikit-learn>=0.21.2->scanpy) (2.1.0)\n",
      "Requirement already satisfied: pynndescent>=0.5 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from umap-learn>=0.3.10->scanpy) (0.5.4)\n",
      "Requirement already satisfied: get-version>=2.0.4 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from legacy-api-wrap->scanpy) (2.1)\n",
      "Requirement already satisfied: stdlib-list in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from sinfo->scanpy) (0.8.0)\n",
      "Requirement already satisfied: numexpr>=2.6.2 in /homes/gws/icovert/anaconda3/envs/numbaenv/lib/python3.6/site-packages (from tables->scanpy) (2.7.3)\n"
     ]
    }
   ],
   "source": [
    "# Install scanpy package\n",
    "!pip install --upgrade scanpy anndata"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "f601a25e-8b7d-4b7b-ad12-ef9262a801bb",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "import scanpy\n",
    "from anndata import AnnData"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "92864838-43bf-4cc3-b957-d189a332f5fc",
   "metadata": {},
   "outputs": [],
   "source": [
    "def seurat(x, num):\n",
    "    ann = AnnData(X=x)\n",
    "    annotations = scanpy.pp.highly_variable_genes(ann, n_top_genes=num, flavor='seurat_v3', inplace=False)\n",
    "    inds = np.where(~np.isnan(annotations['highly_variable_rank'].values))[0]\n",
    "    return np.sort(inds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "ec7ebae5-b306-42c2-88f4-33b79378cabe",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get panels of different sizes\n",
    "num_genes = (32, 64, 128)\n",
    "seurat_results = {num: seurat(raw[train_inds], num) for num in num_genes}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "038a0d90-a91b-48d7-b7df-eb5fb6ae6a27",
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('seurat_results.pkl', 'wb') as f:\n",
    "    pickle.dump(seurat_results, f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6ab8d275-cfb5-4553-9308-bc922c88a87e",
   "metadata": {},
   "source": [
    "# Expression profile reconstruction metric"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "945941d5-d20a-4b89-bbca-e4e914d74351",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import pickle\n",
    "import matplotlib.pyplot as plt\n",
    "from propose import models, MSELoss, ExpressionDataset\n",
    "from torch.utils.data import DataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "e173db54-3367-4c3c-8f6a-328b6e25fcab",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set up datasets\n",
    "train_dataset = ExpressionDataset(binary[train_inds], logcpm[train_inds])\n",
    "val_dataset = ExpressionDataset(binary[val_inds], logcpm[val_inds])\n",
    "test_dataset = ExpressionDataset(binary[test_inds], logcpm[test_inds])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "0f859ede-f1f3-46bb-906b-83b9b61db43d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set up GPU device\n",
    "device = torch.device('cuda', 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "34ae1a9c-0d27-417a-8f96-650916306c2b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare gene sets\n",
    "with open('propose_results.pkl', 'rb') as f:\n",
    "    propose_results = pickle.load(f)\n",
    "with open('propose_class_results.pkl', 'rb') as f:\n",
    "    propose_class_results = pickle.load(f)\n",
    "with open('seurat_results.pkl', 'rb') as f:\n",
    "    seurat_results = pickle.load(f)\n",
    "num_genes = list(propose_results.keys())\n",
    "\n",
    "# Dictionary of methods\n",
    "methods = {\n",
    "    'PROPOSE': propose_results,\n",
    "    'PROPOSE-Class': propose_class_results,\n",
    "    'Seurat': seurat_results\n",
    "}\n",
    "\n",
    "# Dictionary of results\n",
    "results = {name: {} for name in methods}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "e317e6e7-aede-4858-8e71-f1ccc459977f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fit models\n",
    "for name in methods:\n",
    "    for num in num_genes:\n",
    "        # Get inds\n",
    "        inds = methods[name][num]\n",
    "        \n",
    "        # Set up datasets\n",
    "        train_dataset.set_inds(inds)\n",
    "        val_dataset.set_inds(inds)\n",
    "        test_dataset.set_inds(inds)\n",
    "        \n",
    "        # Train model\n",
    "        model = models.MLP(\n",
    "            input_size=num,\n",
    "            output_size=train_dataset.output_size,\n",
    "            hidden=[128, 128]).to(device)\n",
    "        model.fit(\n",
    "            train_dataset,\n",
    "            val_dataset,\n",
    "            mbsize=512,\n",
    "            max_nepochs=500,\n",
    "            loss_fn=MSELoss(),\n",
    "            verbose=False)\n",
    "\n",
    "        # Validation performance\n",
    "        test_mse = model.validate(DataLoader(test_dataset, batch_size=1024), MSELoss()).item()\n",
    "        results[name][num] = test_mse"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "606e8bdd-7d22-4c73-847e-531d03b7c180",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save results\n",
    "with open('reconstruction_metric.pkl', 'wb') as f:\n",
    "    pickle.dump(results, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "b3c55c6b-e651-466a-ba21-474a366ff02f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot results\n",
    "with open('reconstruction_metric.pkl', 'rb') as f:\n",
    "    results = pickle.load(f)\n",
    "num_features = list(results['PROPOSE'].keys())\n",
    "\n",
    "# Get total variance\n",
    "total_variance = np.var(logcpm, axis=0).sum()\n",
    "\n",
    "# Make plot\n",
    "plt.figure(figsize=(9, 6))\n",
    "\n",
    "plt.plot(num_features, [1 - results['PROPOSE'][num] / total_variance for num in num_features],\n",
    "         marker='o', markersize=7, color='C0', label='PROPOSE')\n",
    "plt.plot(num_features, [1 - results['PROPOSE-Class'][num] / total_variance for num in num_features],\n",
    "         marker='o', markersize=7, color='C2', label='PROPOSE-Class')\n",
    "plt.plot(num_features, [1 - results['Seurat'][num] / total_variance for num in num_features],\n",
    "         marker='o', markersize=7, color='C1', label='Seurat')\n",
    "\n",
    "plt.legend(loc='lower right', fontsize=18, frameon=False)\n",
    "plt.tick_params(labelsize=14)\n",
    "plt.title('Genome-Wide Expression Profile Reconstruction', fontsize=18)\n",
    "plt.xlabel('# Genes', fontsize=16)\n",
    "plt.ylabel('Explained Variance', fontsize=16)\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2a26e882-fb79-46f5-906a-d3c7f117d08f",
   "metadata": {},
   "source": [
    "# Cell type classification metric"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "77787586-3ac2-422f-beda-0938036ac70b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import pickle\n",
    "import matplotlib.pyplot as plt\n",
    "from propose import models, Accuracy, ExpressionDataset\n",
    "from torch.utils.data import DataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "1aa68bbe-86ca-4e13-be24-efd6450e7b2f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set up datasets\n",
    "train_dataset = ExpressionDataset(binary[train_inds], cell_types_98_num[train_inds])\n",
    "val_dataset = ExpressionDataset(binary[val_inds], cell_types_98_num[val_inds])\n",
    "test_dataset = ExpressionDataset(binary[test_inds], cell_types_98_num[test_inds])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "8036cb3d-0bfa-46c7-9a78-73e0140c3d79",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare gene sets\n",
    "with open('propose_results.pkl', 'rb') as f:\n",
    "    propose_results = pickle.load(f)\n",
    "with open('propose_class_results.pkl', 'rb') as f:\n",
    "    propose_class_results = pickle.load(f)\n",
    "with open('seurat_results.pkl', 'rb') as f:\n",
    "    seurat_results = pickle.load(f)\n",
    "num_genes = list(propose_results.keys())\n",
    "\n",
    "# Dictionary of methods\n",
    "methods = {\n",
    "    'PROPOSE': propose_results,\n",
    "    'PROPOSE-Class': propose_class_results,\n",
    "    'Seurat': seurat_results\n",
    "}\n",
    "\n",
    "# Dictionary of results\n",
    "results = {name: {} for name in methods}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "ccd101c6-4692-4502-8fd6-772784955111",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fit models\n",
    "for name in methods:\n",
    "    for num in num_genes:\n",
    "        # Get inds\n",
    "        inds = methods[name][num]\n",
    "        \n",
    "        # Set up datasets\n",
    "        train_dataset.set_inds(inds)\n",
    "        val_dataset.set_inds(inds)\n",
    "        test_dataset.set_inds(inds)\n",
    "        \n",
    "        # Train model\n",
    "        model = models.MLP(\n",
    "            input_size=num,\n",
    "            output_size=train_dataset.output_size,\n",
    "            hidden=[128, 128]).to(device)\n",
    "        model.fit(\n",
    "            train_dataset,\n",
    "            val_dataset,\n",
    "            mbsize=512,\n",
    "            max_nepochs=500,\n",
    "            loss_fn=torch.nn.CrossEntropyLoss(),\n",
    "            verbose=False)\n",
    "\n",
    "        # Validation performance\n",
    "        test_acc = model.validate(DataLoader(test_dataset, batch_size=1024), Accuracy()).item()\n",
    "        results[name][num] = test_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "29c780c7-efac-4065-a6e0-f8d581dcf5e8",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save results\n",
    "with open('classification_metric.pkl', 'wb') as f:\n",
    "    pickle.dump(results, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "3db896b0-5fc8-42a8-b768-5cf4ed572273",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot results\n",
    "with open('classification_metric.pkl', 'rb') as f:\n",
    "    results = pickle.load(f)\n",
    "num_features = list(results['PROPOSE'].keys())\n",
    "\n",
    "# Make plot\n",
    "plt.figure(figsize=(9, 6))\n",
    "\n",
    "plt.plot(num_features, [results['PROPOSE'][num] for num in num_features],\n",
    "         marker='o', markersize=7, color='C0', label='PROPOSE')\n",
    "plt.plot(num_features, [results['PROPOSE-Class'][num] for num in num_features],\n",
    "         marker='o', markersize=7, color='C2', label='PROPOSE-Class')\n",
    "plt.plot(num_features, [results['Seurat'][num] for num in num_features],\n",
    "         marker='o', markersize=7, color='C1', label='Seurat')\n",
    "\n",
    "plt.legend(loc='lower right', fontsize=18, frameon=False)\n",
    "plt.tick_params(labelsize=14)\n",
    "plt.title('Cell Type Classification (98 Types)', fontsize=18)\n",
    "plt.xlabel('# Genes', fontsize=16)\n",
    "plt.ylabel('Accuracy', fontsize=16)\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "999fdb1d-e825-47e2-943d-339bb92386e8",
   "metadata": {},
   "source": [
    "# Cell type classification with less granularity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "350e5e41-8c43-45e3-b38c-26d060c9c065",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import pickle\n",
    "import matplotlib.pyplot as plt\n",
    "from propose import models, Accuracy, ExpressionDataset\n",
    "from torch.utils.data import DataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "ffb3b291-1665-428e-8aab-0c3591d870df",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare gene sets\n",
    "with open('propose_class_results.pkl', 'rb') as f:\n",
    "    propose_class_results = pickle.load(f)\n",
    "num_genes = list(propose_class_results.keys())\n",
    "\n",
    "# Dictionary of granularities\n",
    "granularities = {\n",
    "    98: cell_types_98_num,\n",
    "    50: cell_types_50_num,\n",
    "    25: cell_types_25_num\n",
    "}\n",
    "\n",
    "# Dictionary of results\n",
    "results = {num_types: {} for num_types in granularities}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "c1b99b4e-44eb-451d-8e6b-580ebac74912",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Fit models\n",
    "for num_types in granularities:\n",
    "    # Set up datasets\n",
    "    train_dataset = ExpressionDataset(binary[train_inds], granularities[num_types][train_inds])\n",
    "    val_dataset = ExpressionDataset(binary[val_inds], granularities[num_types][val_inds])\n",
    "    test_dataset = ExpressionDataset(binary[test_inds], granularities[num_types][test_inds])\n",
    "    \n",
    "    for num in num_genes:\n",
    "        # Get inds\n",
    "        inds = methods[name][num]\n",
    "        \n",
    "        # Set up datasets\n",
    "        train_dataset.set_inds(inds)\n",
    "        val_dataset.set_inds(inds)\n",
    "        test_dataset.set_inds(inds)\n",
    "        \n",
    "        # Train model\n",
    "        model = models.MLP(\n",
    "            input_size=num,\n",
    "            output_size=train_dataset.output_size,\n",
    "            hidden=[128, 128]).to(device)\n",
    "        model.fit(\n",
    "            train_dataset,\n",
    "            val_dataset,\n",
    "            mbsize=512,\n",
    "            max_nepochs=500,\n",
    "            loss_fn=torch.nn.CrossEntropyLoss(),\n",
    "            verbose=False)\n",
    "\n",
    "        # Validation performance\n",
    "        test_acc = model.validate(DataLoader(test_dataset, batch_size=1024), Accuracy()).item()\n",
    "        results[num_types][num] = test_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "ef116838-394c-4dd4-870e-dac2f7d13a3e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save results\n",
    "with open('classification_granularity_metric.pkl', 'wb') as f:\n",
    "    pickle.dump(results, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "0e0d3ad9-5461-4f82-9ce9-b25f8f6a3de0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot results\n",
    "with open('classification_granularity_metric.pkl', 'rb') as f:\n",
    "    results = pickle.load(f)\n",
    "num_features = list(results[98].keys())\n",
    "\n",
    "# Make plot\n",
    "plt.figure(figsize=(9, 6))\n",
    "\n",
    "plt.plot(num_features, [results[98][num] for num in num_features],\n",
    "         marker='o', markersize=7, color='C2', label='98 Types')\n",
    "plt.plot(num_features, [results[50][num] for num in num_features],\n",
    "         marker='o', markersize=7, color='C3', label='50 Types')\n",
    "plt.plot(num_features, [results[25][num] for num in num_features],\n",
    "         marker='o', markersize=7, color='C4', label='25 Types')\n",
    "\n",
    "plt.legend(loc='lower right', fontsize=18, frameon=False)\n",
    "plt.tick_params(labelsize=14)\n",
    "plt.title('PROPOSE-Class Cell Type Classification (Less Granularity)', fontsize=18)\n",
    "plt.xlabel('# Genes', fontsize=16)\n",
    "plt.ylabel('Accuracy', fontsize=16)\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1c02c829-f7d9-4cb5-81f0-6a686f822b8c",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}