imprv_AUC_gexp_prep_v1.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "e545eb1e-594a-4555-8005-e36a28d7ce64",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Experiment goals\n",
    "Determine features that differentiate across:  \n",
    "    1) Data types  \n",
    "    2) Cancer model types  \n",
    "    3) Tissues-of-origin (Cancer Types)  "
   ]
  },
  {
   "cell_type": "raw",
   "id": "850fb4cf-a663-4a2b-b310-df8ee9b711e2",
   "metadata": {},
   "source": [
    "# link to repo:\n",
    "https://github.com/PNNL-CompBio/panCancerModelSystems/tree/main\n",
    "\n",
    "# link to model template\n",
    "https://keras.io/examples/structured_data/tabtransformer/"
   ]
  },
  {
   "cell_type": "raw",
   "id": "727688f4-b6da-4115-9752-14304f8bccd6",
   "metadata": {},
   "source": [
    "conda activate transformer_env"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "575f4f81-bde4-445e-be7c-6ca610efa009",
   "metadata": {},
   "outputs": [],
   "source": [
    "# This notebook:\n",
    "    # write averaged and transposed expression file to disk\n",
    "        # for read-in by 3 feature selection methods\n",
    "    # write binary and ternary response objects to disk\n",
    "\n",
    "# Update - pairs patch code file removes duplicates for canSMILES that map\n",
    "# to multiple drug ids"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1daa72d3-31e8-4cda-b7d4-145b1b6769db",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "037a88a7-def5-49d3-a9cd-b4084ec5f602",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import math\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "# import tensorflow as tf\n",
    "# from tensorflow import keras\n",
    "# from tensorflow.keras import layers\n",
    "# import tensorflow_addons as tfa\n",
    "import matplotlib.pyplot as plt\n",
    "import collections"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "c6064e86-5f8e-476d-96f3-aee859c86e5b",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "copy_number.csv.gz  drugs.tsv.gz        experiments.tsv.gz  expression.csv.gz\n"
     ]
    }
   ],
   "source": [
    "ls *.gz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "f22bf837-7b5e-4bc6-9478-4a68391ff585",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "genes.csv       mutations.csv   samples.csv     test_data.csv   train_data.csv\n"
     ]
    }
   ],
   "source": [
    "ls *.csv"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f7970907-f7a1-434e-a004-9af3b4b8b495",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Average counts for duplicate genes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8da48031-ca7d-4528-a109-5974686ac8e3",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Send to diff, UMAP, and VAE feature methods after averaging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "0b29413d-a6ce-4f56-8f6f-a9dc4ba9ab29",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "exp_df = pd.read_csv('expression.csv.gz')\n",
    "exp_df_lite = exp_df.iloc[:, :3] # remove study and source cols\n",
    "del(exp_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4de9af1c-8016-40b7-aede-83bad0685377",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Average expression values of duplicate genes\n",
    "# within each improve sample\n",
    "# Run time is about 6 - 7 minutes\n",
    "# Result is removal of ~1.3M genes, ~5% of file\n",
    "\n",
    "# List for each new column\n",
    "e_lst = []\n",
    "i_lst = []\n",
    "c_lst = []\n",
    "\n",
    "# Improve sample loop, 1007 total\n",
    "for smp_id in exp_df_lite.improve_sample_id.unique():\n",
    "    sub = exp_df_lite[exp_df_lite.improve_sample_id == smp_id]\n",
    "    single_list = [] # only populate with gene id if unique occurance\n",
    "    for index, value in sub.entrez_id.value_counts().items():\n",
    "        \n",
    "        if value > 1:\n",
    "\n",
    "            sub_sub = sub[sub.entrez_id == index]\n",
    "            e_lst.append(index)\n",
    "            i_lst.append(smp_id)\n",
    "            c_lst.append(sub_sub.counts.mean())\n",
    "        else:\n",
    "            single_list.append(index) # only add single expression count below\n",
    "                                      # if averaged counts list not modded above\n",
    "    c_lst = c_lst + list(sub[sub.entrez_id.isin(single_list)].counts)\n",
    "    i_lst = i_lst + list(sub[sub.entrez_id.isin(single_list)].improve_sample_id)\n",
    "    e_lst = e_lst + list(sub[sub.entrez_id.isin(single_list)].entrez_id)\n",
    "exp_avrgd = pd.DataFrame({'entrez_id': e_lst, 'improve_sample_id': i_lst, 'counts': c_lst})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "id": "f7fe88e8-972e-487f-b659-3ef05bb44b2c",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "exp_avrgd.to_csv('disk_chkpnts_tssue/exp_avrgd.tsv', sep = '\\t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "41fa68e0-c9b8-4271-bee3-b4ad968bb7e9",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Checkpoint\n",
    "exp_avrgd = pd.read_csv('disk_chkpnts_tssue/exp_avrgd.tsv', sep = '\\t', index_col = 0)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "95eec0bf-5ef1-42f5-b938-430f55f58af4",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Intersection of genes across IMPROVE samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "591dbd79-1c71-4445-924d-7e3cfc4669f9",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "gene_vectors = []\n",
    "for smp_id in exp_avrgd.improve_sample_id.unique():\n",
    "    sub = exp_avrgd[exp_avrgd.improve_sample_id == smp_id]\n",
    "    sub_gene_list = list(sub.entrez_id)\n",
    "    gene_vectors.append(sub_gene_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "e2013617-261c-4ce1-ae43-dedce1005ace",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "ntrsct = set.intersection(*map(set,gene_vectors))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "49db095a-9376-4e12-abb1-f494a9c2f1b9",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "30888"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(ntrsct)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d5820fa0-7213-4cf5-88db-98ca22a59c0f",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Transpose into IMPROVE samples x genes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "adc84e37-7e98-4b92-a183-52deeb9bb331",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "exprsn_dot_T = pd.DataFrame()\n",
    "for smp_id in exp_avrgd.improve_sample_id.unique():\n",
    "    sub = exp_avrgd[exp_avrgd.improve_sample_id == smp_id]\n",
    "    transposed_df = sub.pivot(\n",
    "        index='improve_sample_id',\n",
    "        columns='entrez_id',\n",
    "        values='counts')\n",
    "    exprsn_dot_T = pd.concat([exprsn_dot_T, transposed_df])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "e6e2b7d0-9d73-4376-9632-79e975e02446",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "exprsn_dot_T.to_csv('AUC_v1/exprsn_dot_T.tsv',\n",
    "                  sep = '\\t')\n",
    "# To three feature selection methods"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f74cd005-5191-4ff6-8d2f-7524508d49c2",
   "metadata": {
    "tags": []
   },
   "source": [
    "### Build summary AUC inner join object"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "e7547be7-7fac-4862-a10c-1104c41f804d",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "xpr_df = pd.read_csv('experiments.tsv.gz', sep = '\\t')\n",
    "xpr_auc = xpr_df[['improve_sample_id', 'improve_drug_id', 'AUC']].copy()\n",
    "xpr_check = xpr_auc[['improve_sample_id', 'improve_drug_id']].copy()\n",
    "xpr_check.drop_duplicates(inplace = True)\n",
    "unq_pairs = xpr_auc[xpr_auc.index.isin(xpr_check.index)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "029df348-85e0-484b-94f8-5fc3a5f696d0",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "drg_df = pd.read_csv('drugs.tsv.gz', sep = '\\t')\n",
    "drg_canSMI = drg_df[['canSMILES', 'improve_drug_id']].copy()\n",
    "drg_canSMI.dropna(inplace=True)\n",
    "drg_canSMI.drop_duplicates(inplace = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "12edc17a-1253-43b5-a8bc-a4c4f0c04fcb",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC = unq_pairs.merge(drg_canSMI, how=\"inner\", on=\"improve_drug_id\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "e91d6b55-c981-448b-bc64-7dc52e862e42",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC.to_csv('AUC_v1/inner_join_AUC.tsv', sep = '\\t')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "00c17e1f-764e-40ed-a41a-9b7880722f05",
   "metadata": {},
   "source": [
    "### Binarize and trinarize the response scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "b561a523-6473-4575-970a-c8d2cb77dabb",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC = pd.read_csv('AUC_v1/inner_join_AUC.tsv',\n",
    "                            sep = '\\t',\n",
    "                            index_col = 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "93cf94ed-86fe-4362-9a8e-e572ba3bc270",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC.rename(\n",
    "    columns={col: str(col) for col in inner_join_AUC.columns}, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "c7a839f2-6f25-453d-9c2e-c16e0ff20ec0",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "del inner_join_AUC['improve_drug_id']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "28e0fdec-7704-41b0-a28b-153bfd7e3174",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'AUC, merged PhrmGx drg rspns')"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(inner_join_AUC.AUC, bins = 800)\n",
    "plt.title('AUC, merged PhrmGx drg rspns')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "824a2ce5-6918-43e7-918c-9da9454b4fc6",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC = inner_join_AUC[inner_join_AUC['AUC'].between(0, 1)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "773a9bd6-38c8-4d30-bdd5-8083f146cc85",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'AUC, merged PhrmGx drg rspns')"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(inner_join_AUC.AUC, bins = 800)\n",
    "plt.title('AUC, merged PhrmGx drg rspns')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "c0a3bce5-204f-46a0-827e-884bafe8882d",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inner_join_AUC.AUC.min()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "953ffa17-f73a-4fd7-b97a-62820b7a55ae",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inner_join_AUC.AUC.max()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e83bd2ca-da0b-4a93-89d6-a959bfd4ecfd",
   "metadata": {
    "tags": []
   },
   "source": [
    "#### binarize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "d0017bc6-d788-4547-b3f1-16c868a0271e",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "binary_list = []\n",
    "for auc in inner_join_AUC.AUC:\n",
    "    if auc > .95:\n",
    "        binary_list.append('high')\n",
    "    elif auc < .8:\n",
    "        binary_list.append('low')\n",
    "    else:\n",
    "        binary_list.append('delete')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "aa347fc6-8459-4af9-b28a-c93bd0ba60a0",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC.AUC = binary_list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "93745cd4-0ba4-4251-b41c-5dad465d4fd0",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC = inner_join_AUC[~inner_join_AUC.AUC.str.contains('delete')]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "9129002c-d65c-4f86-ac2c-ae881fd9068d",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC.to_csv('AUC_v1/inner_join_AUC_binary.tsv', sep = '\\t')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5529cdb0-7f2b-4d54-bb4e-13aa771fd634",
   "metadata": {},
   "source": [
    "#### trinarize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "721e4c62-668e-43af-af05-f73d31ea74b9",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "trinary_list = []\n",
    "for auc in inner_join_AUC.AUC:\n",
    "    if auc > .95:\n",
    "        trinary_list.append('high')\n",
    "    elif auc < .5:\n",
    "        trinary_list.append('low')\n",
    "    else:\n",
    "        trinary_list.append('med')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "37ac2694-79d4-42a2-b36e-74aee41dff95",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC.AUC = trinary_list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "d103db97-c6ab-493f-ab4f-08e7295f6bdf",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "inner_join_AUC.to_csv('AUC_v1/inner_join_AUC_trinary.tsv', sep = '\\t')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2a11bb24-794d-4b66-a6ea-30d002380524",
   "metadata": {},
   "outputs": [],
   "source": [
    "# To feature set merges on front-end of transformer"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}