GRPM_02_mesh_selection.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Ref MeSH Builder"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " This notebook allows you to create MeSH lists to inquire the GRPM Dataset starting with a biomedical topic of choice.\n",
    " The system is based on the complete MESH datasheet.\n",
    " The system uses ChatGPT API to supply the initial biomedical terms for constructing the Ref MeSH set."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Import Packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T13:55:44.981701900Z",
     "start_time": "2024-11-15T13:55:44.848671100Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Current working directory: G:\\Altri computer\\Horizon\\horizon_workspace\\projects\\work\\semantics\\GRPM\\GRPM_github\n"
     ]
    }
   ],
   "source": [
    "import importlib\n",
    "import subprocess\n",
    "\n",
    "try:\n",
    "    importlib.import_module('pygrpm')\n",
    "except ImportError:\n",
    "    subprocess.check_call([\"pip\", \"install\", \"git+https://github.com/johndef64/GRPM_system.git\"])\n",
    "\n",
    "from pygrpm import *\n",
    "\n",
    "import os\n",
    "import glob\n",
    "import requests\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from datetime import datetime\n",
    "\n",
    "current_directory = os.getcwd()\n",
    "print(\"Current working directory:\", current_directory)\n",
    "\n",
    "def simple_bool(message):\n",
    "    choose = input(message+\" (y/n): \").lower()\n",
    "    your_bool = choose in [\"y\", \"yes\",\"yea\",\"sure\"]\n",
    "    return your_bool\n",
    "\n",
    "def get_file(url, file_name, dir = os.getcwd()):\n",
    "    url = url\n",
    "    file_name = file_name\n",
    "    response = requests.get(url)\n",
    "    if response.status_code == 200:\n",
    "        content = response.content\n",
    "        file_path = os.path.join(dir, file_name)\n",
    "        with open(file_path, 'wb') as file:\n",
    "            file.write(content)\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Get requirements"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Get MESH.csv from 'bioportal.bioontology.org'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import gzip\n",
    "import shutil\n",
    "# Get MESH.csv from https://bioportal.bioontology.org/ontologies/MESH?p=summary\n",
    "\n",
    "if not os.path.exists('ref-mesh'):\n",
    "    os.makedirs('ref-mesh')\n",
    "\n",
    "if not os.path.exists('ref-mesh/MESH.csv'):\n",
    "    get_file( url='https://data.bioontology.org/ontologies/MESH/download?apikey=8b5b7825-538d-40e0-9e9e-5ab9274a9aeb&download_format=csv', file_name='MESH.gz', dir = 'ref-mesh')\n",
    "\n",
    "    # Open the gzipped input file and the output file\n",
    "    file_path = 'ref-mesh/'\n",
    "    with gzip.open(file_path+'MESH.gz', 'rb') as input_file, open(file_path+'MESH.csv', 'wb') as output_file:\n",
    "        # Copy the content from the input gzipped file to the output file\n",
    "        shutil.copyfileobj(input_file, output_file)\n",
    "    \n",
    "    os.remove(file_path+'MESH.gz')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Get pre-made datasets from Zenodo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-08T12:30:02.096036100Z",
     "start_time": "2024-11-08T12:29:55.384037500Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14052302\n",
      "Downloading ref-mesh.zip from https://zenodo.org/record/14052302/files/ref-mesh.zip?download=1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Downloading...\n",
      "From: https://zenodo.org/record/14052302/files/ref-mesh.zip?download=1\n",
      "To: G:\\Altri computer\\Horizon\\horizon_workspace\\projects\\work\\semantics\\GRPM\\GRPM_github\\ref-mesh.zip\n",
      "100%|██████████| 4.02M/4.02M [00:01<00:00, 3.23MB/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting...\n",
      "ZIP file 'ref-mesh.zip' extracted in 'G:\\Altri computer\\Horizon\\horizon_workspace\\projects\\work\\semantics\\GRPM\\GRPM_github' successfully.\n"
     ]
    }
   ],
   "source": [
    "# Get pre-made datasets from Zenodo Repository\n",
    "#https://zenodo.org/record/8205724  DOI: 10.5281/zenodo.8205724\n",
    "dataset_releases = {'0.1':'8205724',\n",
    "                    '0.2':'14052302'}\n",
    "    \n",
    "if not os.path.exists('ref-mesh/MESH_STY_LITVAR1.csv'):\n",
    "    get_and_extract('ref-mesh', record_id=dataset_releases['0.2'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Load Mesh full dataset\n",
    "(skip this if MESH.csv is not provided)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T13:56:47.208472100Z",
     "start_time": "2024-11-15T13:56:37.341470Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "                            0\n0                  blood (BL)\n1    cerebrospinal fluid (CF)\n2     chemically induced (CI)\n3         classification (CL)\n4          complications (CO)\n5             congenital (CN)\n6              diagnosis (DI)\n7     diagnostic imaging (DG)\n8           diet therapy (DH)\n9           drug therapy (DT)\n10             economics (EC)\n11            embryology (EM)\n12            enzymology (EN)\n13          epidemiology (EP)\n14             ethnology (EH)\n15              etiology (ET)\n16              genetics (GE)\n17               history (HI)\n18            immunology (IM)\n19            metabolism (ME)\n20          microbiology (MI)\n21             mortality (MO)\n22               nursing (NU)\n23          parasitology (PS)\n24             pathology (PA)\n25       physiopathology (PP)\n26  prevention & control (PC)\n27            psychology (PX)\n28          radiotherapy (RT)\n29        rehabilitation (RH)\n30               surgery (SU)\n31               therapy (TH)\n32                 urine (UR)\n33            veterinary (VE)\n34              virology (VI)",
      "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>0</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>blood (BL)</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>cerebrospinal fluid (CF)</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>chemically induced (CI)</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>classification (CL)</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>complications (CO)</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>congenital (CN)</td>\n    </tr>\n    <tr>\n      <th>6</th>\n      <td>diagnosis (DI)</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>diagnostic imaging (DG)</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>diet therapy (DH)</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>drug therapy (DT)</td>\n    </tr>\n    <tr>\n      <th>10</th>\n      <td>economics (EC)</td>\n    </tr>\n    <tr>\n      <th>11</th>\n      <td>embryology (EM)</td>\n    </tr>\n    <tr>\n      <th>12</th>\n      <td>enzymology (EN)</td>\n    </tr>\n    <tr>\n      <th>13</th>\n      <td>epidemiology (EP)</td>\n    </tr>\n    <tr>\n      <th>14</th>\n      <td>ethnology (EH)</td>\n    </tr>\n    <tr>\n      <th>15</th>\n      <td>etiology (ET)</td>\n    </tr>\n    <tr>\n      <th>16</th>\n      <td>genetics (GE)</td>\n    </tr>\n    <tr>\n      <th>17</th>\n      <td>history (HI)</td>\n    </tr>\n    <tr>\n      <th>18</th>\n      <td>immunology (IM)</td>\n    </tr>\n    <tr>\n      <th>19</th>\n      <td>metabolism (ME)</td>\n    </tr>\n    <tr>\n      <th>20</th>\n      <td>microbiology (MI)</td>\n    </tr>\n    <tr>\n      <th>21</th>\n      <td>mortality (MO)</td>\n    </tr>\n    <tr>\n      <th>22</th>\n      <td>nursing (NU)</td>\n    </tr>\n    <tr>\n      <th>23</th>\n      <td>parasitology (PS)</td>\n    </tr>\n    <tr>\n      <th>24</th>\n      <td>pathology (PA)</td>\n    </tr>\n    <tr>\n      <th>25</th>\n      <td>physiopathology (PP)</td>\n    </tr>\n    <tr>\n      <th>26</th>\n      <td>prevention &amp; control (PC)</td>\n    </tr>\n    <tr>\n      <th>27</th>\n      <td>psychology (PX)</td>\n    </tr>\n    <tr>\n      <th>28</th>\n      <td>radiotherapy (RT)</td>\n    </tr>\n    <tr>\n      <th>29</th>\n      <td>rehabilitation (RH)</td>\n    </tr>\n    <tr>\n      <th>30</th>\n      <td>surgery (SU)</td>\n    </tr>\n    <tr>\n      <th>31</th>\n      <td>therapy (TH)</td>\n    </tr>\n    <tr>\n      <th>32</th>\n      <td>urine (UR)</td>\n    </tr>\n    <tr>\n      <th>33</th>\n      <td>veterinary (VE)</td>\n    </tr>\n    <tr>\n      <th>34</th>\n      <td>virology (VI)</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#load MESH dataframe:\n",
    "if os.path.exists('ref-mesh/MESH.csv'):\n",
    "    df = pd.read_csv('ref-mesh/MESH.csv', low_memory=False)\n",
    "else:\n",
    "    print(\"no MESH.csv available.\\n => skip to Section 2\")\n",
    "\n",
    "#info on Proprieties: https://bioportal.bioontology.org/ontologies/MESH?p=properties\n",
    "#Mesh Browser: https://meshb.nlm.nih.gov/\n",
    "\n",
    "# AQL: Allowable Qualifiers\n",
    "AQL = [\"blood (BL)\",\"cerebrospinal fluid (CF)\",\"chemically induced (CI)\",\"classification (CL)\",\"complications (CO)\",\"congenital (CN)\",\"diagnosis (DI)\",\"diagnostic imaging (DG)\",\"diet therapy (DH)\",\"drug therapy (DT)\",\"economics (EC)\",\"embryology (EM)\",\"enzymology (EN)\",\"epidemiology (EP)\",\"ethnology (EH)\",\"etiology (ET)\",\"genetics (GE)\",\"history (HI)\",\"immunology (IM)\",\"metabolism (ME)\",\"microbiology (MI)\",\"mortality (MO)\",\"nursing (NU)\",\"parasitology (PS)\",\"pathology (PA)\",\"physiopathology (PP)\",\"prevention & control (PC)\",\"psychology (PX)\",\"radiotherapy (RT)\",\"rehabilitation (RH)\",\"surgery (SU)\",\"therapy (TH)\",\"urine (UR)\",\"veterinary (VE)\",\"virology (VI)\"]\n",
    "aql=pd.DataFrame(AQL)\n",
    "aql"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T13:56:52.637469300Z",
     "start_time": "2024-11-15T13:56:52.439470200Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "display MeSH data:\n"
     ]
    },
    {
     "data": {
      "text/plain": "                                            Class ID  \\\n0  http://purl.bioontology.org/ontology/MESH/C039049   \n1  http://purl.bioontology.org/ontology/MESH/C039048   \n2  http://purl.bioontology.org/ontology/MESH/C039047   \n3  http://purl.bioontology.org/ontology/MESH/C039046   \n4  http://purl.bioontology.org/ontology/MESH/C039040   \n\n                       Preferred Label  \\\n0                             UP 5-207   \n1               manganese-zinc ferrite   \n2               2,3,6-trichlorotoluene   \n3  2,3,6-trichloro-4-tert-butyltoluene   \n4           ethiodized oil emulsion 13   \n\n                                            Synonyms Definitions  Obsolete  \\\n0  Phenol, 4,4'-(1-methylethylidene)bis-, polymer...         NaN     False   \n1                                                NaN         NaN     False   \n2                                          2,3,6-TCT         NaN     False   \n3  2,3,6-TTBT|Benzene, 1,3,4-trichloro-5-(1,1-dim...         NaN     False   \n4                            ethiodiol|EOE 13|EOE-13         NaN     False   \n\n                 CUI                                     Semantic Types  \\\n0           C0619384      http://purl.bioontology.org/ontology/STY/T109   \n1           C0065653      http://purl.bioontology.org/ontology/STY/T197   \n2           C0045332      http://purl.bioontology.org/ontology/STY/T109   \n3           C0619382      http://purl.bioontology.org/ontology/STY/T109   \n4  C0059720|C0116736  http://purl.bioontology.org/ontology/STY/T109|...   \n\n  Parents   AN  AQL  ...   OL   PA                                PI  \\\n0     NaN  NaN  NaN  ...  NaN  NaN                               NaN   \n1     NaN  NaN  NaN  ...  NaN  NaN  *MANGANESE (83-93)|*ZINC (83-93)   \n2     NaN  NaN  NaN  ...  NaN  NaN                               NaN   \n3     NaN  NaN  NaN  ...  NaN  NaN                               NaN   \n4     NaN  NaN  NaN  ...  NaN  NaN                               NaN   \n\n                                   RR   SC  \\\n0                                 NaN  1.0   \n1                                 NaN  1.0   \n2  2077-46-5 (2,3,6-trichlorotoluene)  1.0   \n3                                 NaN  1.0   \n4                                 NaN  1.0   \n\n                                  Scope Statement  \\\n0                                      epoxy glue   \n1                                             NaN   \n2                                             NaN   \n3                                             NaN   \n4  used for detection of tumors in liver & spleen   \n\n                         Semantic type UMLS property  \\\n0      http://purl.bioontology.org/ontology/STY/T109   \n1      http://purl.bioontology.org/ontology/STY/T197   \n2      http://purl.bioontology.org/ontology/STY/T109   \n3      http://purl.bioontology.org/ontology/STY/T109   \n4  http://purl.bioontology.org/ontology/STY/T109|...   \n\n                                    SRC                           TERMUI  \\\n0         Gig Tr Prof Zabol 1983;(7):54                  T147161|T147162   \n1                 Gig Sanit 1983;(6):13                          T147160   \n2                 Gig Sanit 1983;(6):61                  T147158|T147159   \n3                 Gig Sanit 1983;(6):61                  T147156|T147157   \n4  J Comput Assist Tomogr 1983;7(5):905  T147143|T147142|T147140|T147141   \n\n                                     TH  \n0                            NLM (1983)  \n1                            NLM (1983)  \n2  INN (19XX)|FDA SRS (2015)|NLM (1983)  \n3                            NLM (1983)  \n4                            NLM (1983)  \n\n[5 rows x 45 columns]",
      "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>Class ID</th>\n      <th>Preferred Label</th>\n      <th>Synonyms</th>\n      <th>Definitions</th>\n      <th>Obsolete</th>\n      <th>CUI</th>\n      <th>Semantic Types</th>\n      <th>Parents</th>\n      <th>AN</th>\n      <th>AQL</th>\n      <th>...</th>\n      <th>OL</th>\n      <th>PA</th>\n      <th>PI</th>\n      <th>RR</th>\n      <th>SC</th>\n      <th>Scope Statement</th>\n      <th>Semantic type UMLS property</th>\n      <th>SRC</th>\n      <th>TERMUI</th>\n      <th>TH</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C039049</td>\n      <td>UP 5-207</td>\n      <td>Phenol, 4,4'-(1-methylethylidene)bis-, polymer...</td>\n      <td>NaN</td>\n      <td>False</td>\n      <td>C0619384</td>\n      <td>http://purl.bioontology.org/ontology/STY/T109</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>1.0</td>\n      <td>epoxy glue</td>\n      <td>http://purl.bioontology.org/ontology/STY/T109</td>\n      <td>Gig Tr Prof Zabol 1983;(7):54</td>\n      <td>T147161|T147162</td>\n      <td>NLM (1983)</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C039048</td>\n      <td>manganese-zinc ferrite</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>False</td>\n      <td>C0065653</td>\n      <td>http://purl.bioontology.org/ontology/STY/T197</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>*MANGANESE (83-93)|*ZINC (83-93)</td>\n      <td>NaN</td>\n      <td>1.0</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T197</td>\n      <td>Gig Sanit 1983;(6):13</td>\n      <td>T147160</td>\n      <td>NLM (1983)</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C039047</td>\n      <td>2,3,6-trichlorotoluene</td>\n      <td>2,3,6-TCT</td>\n      <td>NaN</td>\n      <td>False</td>\n      <td>C0045332</td>\n      <td>http://purl.bioontology.org/ontology/STY/T109</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>2077-46-5 (2,3,6-trichlorotoluene)</td>\n      <td>1.0</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T109</td>\n      <td>Gig Sanit 1983;(6):61</td>\n      <td>T147158|T147159</td>\n      <td>INN (19XX)|FDA SRS (2015)|NLM (1983)</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C039046</td>\n      <td>2,3,6-trichloro-4-tert-butyltoluene</td>\n      <td>2,3,6-TTBT|Benzene, 1,3,4-trichloro-5-(1,1-dim...</td>\n      <td>NaN</td>\n      <td>False</td>\n      <td>C0619382</td>\n      <td>http://purl.bioontology.org/ontology/STY/T109</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>1.0</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T109</td>\n      <td>Gig Sanit 1983;(6):61</td>\n      <td>T147156|T147157</td>\n      <td>NLM (1983)</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C039040</td>\n      <td>ethiodized oil emulsion 13</td>\n      <td>ethiodiol|EOE 13|EOE-13</td>\n      <td>NaN</td>\n      <td>False</td>\n      <td>C0059720|C0116736</td>\n      <td>http://purl.bioontology.org/ontology/STY/T109|...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>1.0</td>\n      <td>used for detection of tumors in liver &amp; spleen</td>\n      <td>http://purl.bioontology.org/ontology/STY/T109|...</td>\n      <td>J Comput Assist Tomogr 1983;7(5):905</td>\n      <td>T147143|T147142|T147140|T147141</td>\n      <td>NLM (1983)</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows × 45 columns</p>\n</div>"
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('display MeSH data:\\n')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#How many MeSH has mapping qualifier (subheadings)?\n",
    "all_mesh = df['Preferred Label'].drop_duplicates()\n",
    "mapping_qf = df[['Mapping qualifier of','Preferred Label']].drop_duplicates().dropna()\n",
    "has = df[['Has mapping qualifier','Preferred Label']].drop_duplicates().dropna()\n",
    "has_len = has['Preferred Label'].nunique()\n",
    "\n",
    "print('Total number of MeSH terms:\\n', df['Preferred Label'].nunique())\n",
    "print('\\nHow many Mesh has mapping qualifier (subheadings)?\\n',has_len)\n",
    "print(' %', round((has_len/len(all_mesh)*100),3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# search for single exact value\n",
    "mesh = 'Disease or Syndrome'\n",
    "mesh = 'Heart Failure'\n",
    "print('search for single exact value:')\n",
    "df[df['Preferred Label']==mesh].T.dropna()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Extract subset of Mesh\n",
    "str_1 = 'Fabry'\n",
    "str_2 = 'Lysosomal Sto'\n",
    "str_3 = ''\n",
    "str_full = str_2 or str_1 or str_3\n",
    "\n",
    "ref_search = df[df['Preferred Label'].str.contains(str_full).fillna(False)]\n",
    "print('look for mesh containing \"',str_1,',',str_2,',',str_3,  '\":')\n",
    "ref_search"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Semantic Types Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Extract subset of all semantic types\n",
    "df_sem = df[df['Class ID'].str.contains('STY').fillna(False)]\n",
    "df_sem"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Characterize a Semantic Type"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T13:57:22.028366200Z",
     "start_time": "2024-11-15T13:57:21.671569500Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "semantic type: Receptor\n",
      "number of mesh: 4617\n",
      "\n",
      "Mesh for \" T192 Receptor \" semantic type:\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Utente\\AppData\\Local\\Temp\\ipykernel_13968\\985110198.py:7: FutureWarning: Downcasting object dtype arrays on .fillna, .ffill, .bfill is deprecated and will change in a future version. Call result.infer_objects(copy=False) instead. To opt-in to the future behavior, set `pd.set_option('future.no_silent_downcasting', True)`\n",
      "  mask2 = df['Semantic Types'].str.contains(semantictype).fillna(False)\n"
     ]
    },
    {
     "data": {
      "text/plain": "                                                 Class ID  \\\n157     http://purl.bioontology.org/ontology/MESH/C033482   \n488     http://purl.bioontology.org/ontology/MESH/C491405   \n557     http://purl.bioontology.org/ontology/MESH/C116483   \n560     http://purl.bioontology.org/ontology/MESH/C116484   \n626     http://purl.bioontology.org/ontology/MESH/C087430   \n...                                                   ...   \n353374  http://purl.bioontology.org/ontology/MESH/C515077   \n353739  http://purl.bioontology.org/ontology/MESH/C030708   \n353784  http://purl.bioontology.org/ontology/MESH/C112493   \n353822  http://purl.bioontology.org/ontology/MESH/D047629   \n353823  http://purl.bioontology.org/ontology/MESH/D047628   \n\n                                Preferred Label  \\\n157                serum thymic factor receptor   \n488                     TNFRSF18 protein, human   \n557                   plexA protein, Drosophila   \n560                   plexB protein, Drosophila   \n626     22.6-kDa antigen, Schistosoma japonicum   \n...                                         ...   \n353374                   KIR2DL2 protein, human   \n353739              triazine herbicide receptor   \n353784                EGL-15 protein, C elegans   \n353822                   Estrogen Receptor beta   \n353823                  Estrogen Receptor alpha   \n\n                                              Definitions  \\\n157                                                   NaN   \n488                                                   NaN   \n557                                                   NaN   \n560                                                   NaN   \n626                                                   NaN   \n...                                                   ...   \n353374                                                NaN   \n353739                                                NaN   \n353784                                                NaN   \n353822  One of the ESTROGEN RECEPTORS that has greater...   \n353823  One of the ESTROGEN RECEPTORS that has marked ...   \n\n                                           Semantic Types  \n157     http://purl.bioontology.org/ontology/STY/T116|...  \n488     http://purl.bioontology.org/ontology/STY/T116|...  \n557     http://purl.bioontology.org/ontology/STY/T116|...  \n560     http://purl.bioontology.org/ontology/STY/T116|...  \n626     http://purl.bioontology.org/ontology/STY/T116|...  \n...                                                   ...  \n353374  http://purl.bioontology.org/ontology/STY/T116|...  \n353739  http://purl.bioontology.org/ontology/STY/T116|...  \n353784  http://purl.bioontology.org/ontology/STY/T116|...  \n353822  http://purl.bioontology.org/ontology/STY/T116|...  \n353823  http://purl.bioontology.org/ontology/STY/T116|...  \n\n[4617 rows x 4 columns]",
      "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>Class ID</th>\n      <th>Preferred Label</th>\n      <th>Definitions</th>\n      <th>Semantic Types</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>157</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C033482</td>\n      <td>serum thymic factor receptor</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n    <tr>\n      <th>488</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C491405</td>\n      <td>TNFRSF18 protein, human</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n    <tr>\n      <th>557</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C116483</td>\n      <td>plexA protein, Drosophila</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n    <tr>\n      <th>560</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C116484</td>\n      <td>plexB protein, Drosophila</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n    <tr>\n      <th>626</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C087430</td>\n      <td>22.6-kDa antigen, Schistosoma japonicum</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>353374</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C515077</td>\n      <td>KIR2DL2 protein, human</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n    <tr>\n      <th>353739</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C030708</td>\n      <td>triazine herbicide receptor</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n    <tr>\n      <th>353784</th>\n      <td>http://purl.bioontology.org/ontology/MESH/C112493</td>\n      <td>EGL-15 protein, C elegans</td>\n      <td>NaN</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n    <tr>\n      <th>353822</th>\n      <td>http://purl.bioontology.org/ontology/MESH/D047629</td>\n      <td>Estrogen Receptor beta</td>\n      <td>One of the ESTROGEN RECEPTORS that has greater...</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n    <tr>\n      <th>353823</th>\n      <td>http://purl.bioontology.org/ontology/MESH/D047628</td>\n      <td>Estrogen Receptor alpha</td>\n      <td>One of the ESTROGEN RECEPTORS that has marked ...</td>\n      <td>http://purl.bioontology.org/ontology/STY/T116|...</td>\n    </tr>\n  </tbody>\n</table>\n<p>4617 rows × 4 columns</p>\n</div>"
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#how many mesh for semantic type?\n",
    "semantictype = 'T192'\n",
    "sam_label = df[df['Class ID'].str.contains(semantictype).fillna(False)].reset_index().at[0,'Preferred Label']\n",
    "# Get Semantic Type name for STY ID\n",
    "print('semantic type:', sam_label)\n",
    "\n",
    "mask2 = df['Semantic Types'].str.contains(semantictype).fillna(False)\n",
    "mesh_sem = df[mask2][['Class ID','Preferred Label','Definitions','Semantic Types']]\n",
    "print('number of mesh:', len(mesh_sem))\n",
    "print('\\nMesh for \"',semantictype,sam_label,'\" semantic type:')\n",
    "\n",
    "mesh_sem"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Characterize a single mesh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T13:57:55.999128700Z",
     "start_time": "2024-11-15T13:57:55.776453Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mesh: Diethylamines \n",
      "semantic type: Organic Chemical T109 \n",
      "descrition: Diethylamine with the formula CH3CH2NHCH2CH3 and its derivatives.\n"
     ]
    }
   ],
   "source": [
    "# search for single Mesh\n",
    "mesh_classid ='D004048'\n",
    "maskmesh = df['Class ID'].str.contains(mesh_classid).fillna(False)\n",
    "\n",
    "sty_classid = df[maskmesh]['Semantic Types'].reset_index().iat[0, 1][-4:]\n",
    "\n",
    "masksty = df['Class ID'].str.contains(sty_classid).fillna(False)\n",
    "\n",
    "masked = df[maskmesh][['Semantic Types','Preferred Label','Definitions']].reset_index()\n",
    "\n",
    "#get mesh definition:\n",
    "print('mesh:', masked.at[0,'Preferred Label'],\n",
    "      '\\nsemantic type:',df[masksty].reset_index().at[0,'Preferred Label'],sty_classid,\n",
    "      '\\ndescrition:',masked.at[0,'Definitions'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Row aggregation\n",
    "data_mesh = []\n",
    "data_sty = ['T021','T116']\n",
    "dfd = pd.DataFrame()\n",
    "for i in data_sty:\n",
    "    mask = df['Semantic Types'].str.contains(i).fillna(False) # aggregate mesh for sty\n",
    "    dfd = pd.concat([dfd, df[mask]])\n",
    "dfd"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Load/create MESH-LITVAR dataset (required)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T13:59:11.013213Z",
     "start_time": "2024-11-15T13:59:04.350218200Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MESH_STY sty imported from csv\n",
      "MESH_STY_LITVAR1 imported from csv\n",
      "\n",
      " => For Reference MeSH list definition skip to Section 3\n"
     ]
    }
   ],
   "source": [
    "# Subset MeSH dataset\n",
    "\n",
    "#set options:\n",
    "import_mesh_sty = True\n",
    "\n",
    "# concat the exploded mesh table or import prearranged csv\n",
    "if os.path.isfile('ref-mesh/MESH_STY.csv') and import_mesh_sty:\n",
    "    mesh_large_df_sty = pd.read_csv('ref-mesh/MESH_STY.csv', index_col=0)\n",
    "    print('MESH_STY sty imported from csv')\n",
    "elif import_mesh_sty:\n",
    "        if os.path.exists('ref-mesh/MESH.csv'):\n",
    "            # correct list format\n",
    "            df['STY_ID'] = df['Semantic Types'].str.replace(r'http://purl.bioontology.org/ontology/STY/','', regex = False)\n",
    "            start_word = '[\\\"'\n",
    "            end_word = '\\\"]'\n",
    "            df['STY_ID'] = f'{start_word} ' + df['STY_ID'] + f' {end_word}'\n",
    "            df['STY_ID'] = df['STY_ID'].str.replace(' ','', regex = False)\n",
    "            df['STY_ID'] = df['STY_ID'].str.replace('|','\\\",\\\"', regex = False)\n",
    "            #print(df['STY_ID'].isna().sum())\n",
    "            df.dropna(subset=['STY_ID'], inplace=True)\n",
    "            df.reset_index(drop=True,inplace=True)\n",
    "\n",
    "            from ast import literal_eval\n",
    "            df['STY_ID'] = df['STY_ID'].apply(literal_eval)\n",
    "\n",
    "            # Import exhcange table sty-code\n",
    "            sty = pd.read_csv('ref-mesh/MeshSTY-code.csv',sep=';')\n",
    "            sty = sty.rename(columns={'ID':'Semantic Types'})\n",
    "            sty = sty.rename(columns={'ID':'Semantic Types'})\n",
    "            #print(sty['Semantic Types'].nunique())\n",
    "\n",
    "            #mesh_large = pd.DataFrame()\n",
    "            mesh_large = []\n",
    "            for i in range(len(df)):\n",
    "                for sem in df['STY_ID'][i]: #dfrspost = mother table\n",
    "                    out = df['Preferred Label'][i],df['Class ID'][i],sem,df['Synonyms'][i],df['Parents'][i],df['CUI'][i],df['AQL'][i],df['TERMUI'][i]\n",
    "                    mesh_large.append(out)\n",
    "\n",
    "            mesh_large_df = pd.DataFrame(mesh_large)\n",
    "            new_col_names = ['Preferred Label','Class ID','Semantic Types','Synonyms','Parents','CUI','AQL','TERMUI']\n",
    "            mesh_large_df.columns = new_col_names\n",
    "\n",
    "            ## Add STY Labels\n",
    "            mesh_large_df_sty = pd.merge(mesh_large_df, sty, on='Semantic Types', how='inner').reset_index(drop=True)\n",
    "            #Add rsid coulmn con merge\n",
    "            mesh_large_df_sty = mesh_large_df_sty.rename(columns={'Preferred Label_y':'Semantic Types Label','Preferred Label_x':'Preferred Label'})\n",
    "\n",
    "            mesh_large_df_sty = mesh_large_df_sty[['Preferred Label', 'Semantic Types Label', 'Class ID', 'Semantic Types', 'Synonyms', 'Parents', 'CUI', 'AQL', 'TERMUI']]\n",
    "            mesh_large_df_sty.to_csv('ref-mesh/MESH_STY.csv')\n",
    "            print('MESH_STY created')\n",
    "        else:\n",
    "            print('MESH.csv not available')\n",
    "\n",
    "#---------------------------------------------------\n",
    "#### Create MESH-STY-LITVAR subset from MESH-STY.csv\n",
    "if os.path.exists('ref-mesh/MESH_STY_LITVAR1.csv'):\n",
    "    mesh_litvar_sty = pd.read_csv('ref-mesh/MESH_STY_LITVAR1.csv',index_col=0)\n",
    "    print('MESH_STY_LITVAR1 imported from csv')\n",
    "else:\n",
    "    grpm_mesh = pd.read_csv('ref-mesh/grpm_db_mesh.csv', index_col=0)\n",
    "    mask = mesh_large_df_sty['Preferred Label'].isin(grpm_mesh.mesh)\n",
    "    mesh_litvar_sty = mesh_large_df_sty[mask]\n",
    "    mesh_litvar_sty.to_csv('ref-mesh/MESH_STY_LITVAR1.csv')\n",
    "    print('MESH_STY_LITVAR1 created')\n",
    "\n",
    "print('\\n => For Reference MeSH list definition skip to Section 3')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# MeSH stats\n",
    "if 'mesh_large_df_sty' in globals():\n",
    "    print('MESH_STY', mesh_large_df_sty['Preferred Label'].nunique(), 'mesh')\n",
    "    print('MESH_STY', mesh_large_df_sty['Semantic Types Label'].nunique(), 'semantic types')\n",
    "\n",
    "print('MESH_STY_LITVAR1', mesh_litvar_sty['Preferred Label'].nunique(), 'mesh')\n",
    "print('MESH_STY_LITVAR1', mesh_litvar_sty['Semantic Types Label'].nunique(), 'semantic types\\n')\n",
    "\n",
    "mesh_litvar_sty[['Preferred Label', 'Semantic Types Label', 'Class ID', 'mesh_id','Semantic Types']]#.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## ADD mesh_id col\n",
    "if 'mesh_id' not in mesh_litvar_sty.columns:\n",
    "    mesh_litvar_sty['mesh_id'] = mesh_litvar_sty['Class ID'].str.replace('http://purl.bioontology.org/ontology/MESH/', '')\n",
    "    mesh_litvar_sty['mesh_id']\n",
    "    mesh_litvar_sty.columns\n",
    "    new_order = ['Preferred Label', 'Semantic Types Label', 'Class ID', 'mesh_id', 'Semantic Types',\n",
    "                 'Synonyms', 'Parents', 'CUI', 'AQL', 'TERMUI']\n",
    "    mesh_litvar_sty = mesh_litvar_sty[new_order]\n",
    "    mesh_litvar_sty.to_csv('ref-mesh/MESH_STY_LITVAR1.csv')\n",
    "else:\n",
    "    print('mesh_id aready added to csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Add mesh_id to GRPM dataset (not required)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T14:01:09.591297500Z",
     "start_time": "2024-11-15T14:00:57.996011900Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "note: in GRPM Dataset screening using 'Preferred Label' results faster then numerical 'mesh_id'\n"
     ]
    }
   ],
   "source": [
    "print(\"note: in GRPM Dataset screening using 'Preferred Label' results faster then numerical 'mesh_id'\")\n",
    "\n",
    "if simple_bool(\"note: in GRPM Dataset screening using 'Preferred Label' results faster then numerical 'mesh_id'\\n add mesh_id to grpm_ds and replace 'Preferred Label' anyway?\") and os.path.exists('grpm_dataset/grpm_db_pcg/grpm_table_output.csv'):\n",
    "    grpm_b_df = pd.read_parquet('grpm_dataset/grpm_dataset.parquet')\n",
    "\n",
    "    if not 'mesh_id' in grpm_b_df.columns:\n",
    "        # add mesh_id to grpm db\n",
    "        mesh_id_ref_df = mesh_litvar_sty[['Preferred Label','mesh_id']]\n",
    "        grpm_db_merge_id = pd.merge(grpm_b_df, mesh_id_ref_df, left_on='mesh', right_on='Preferred Label')\n",
    "\n",
    "        # replace 'Preferred Label' with 'mesh_id'\n",
    "        grpm_db_merge_id = grpm_db_merge_id.drop('Preferred Label', axis=1)\n",
    "        grpm_db_merge_id.columns\n",
    "        new_order = ['gene', 'rsid', 'pmid', 'mesh_id']\n",
    "        grpm_db_merge_id = grpm_db_merge_id[new_order].drop_duplicates()\n",
    "        grpm_db_merge_id"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "if simple_bool('save it?'):\n",
    "    grpm_db_merge_id.to_parquet('grpm_dataset/grpm_datset_id.parquet')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "if simple_bool('check?'):\n",
    "    check = pd.read_parquet('grpm_dataset/grpm_datset_id.parquet')\n",
    "    check"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Sorting Mesh-STY"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T14:02:06.935412700Z",
     "start_time": "2024-11-15T14:02:05.542409900Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MESH_STY\n",
      "Semantic Type count: 126 \n",
      "\n",
      "MeSH count for Sty:\n"
     ]
    },
    {
     "data": {
      "text/plain": "    mesh-count             Semantic Types Label\n0       136114                 Organic Chemical\n1       104554  Amino Acid, Peptide, or Protein\n2        62414    Biologically Active Substance\n3        60593          Pharmacologic Substance\n4        45706                           Fungus\n..         ...                              ...\n121          4                       Vertebrate\n122          3             Drug Delivery Device\n123          3                            Human\n124          2                           Entity\n125          1            Carbohydrate Sequence\n\n[126 rows x 2 columns]",
      "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>mesh-count</th>\n      <th>Semantic Types Label</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>136114</td>\n      <td>Organic Chemical</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>104554</td>\n      <td>Amino Acid, Peptide, or Protein</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>62414</td>\n      <td>Biologically Active Substance</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>60593</td>\n      <td>Pharmacologic Substance</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>45706</td>\n      <td>Fungus</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>121</th>\n      <td>4</td>\n      <td>Vertebrate</td>\n    </tr>\n    <tr>\n      <th>122</th>\n      <td>3</td>\n      <td>Drug Delivery Device</td>\n    </tr>\n    <tr>\n      <th>123</th>\n      <td>3</td>\n      <td>Human</td>\n    </tr>\n    <tr>\n      <th>124</th>\n      <td>2</td>\n      <td>Entity</td>\n    </tr>\n    <tr>\n      <th>125</th>\n      <td>1</td>\n      <td>Carbohydrate Sequence</td>\n    </tr>\n  </tbody>\n</table>\n<p>126 rows × 2 columns</p>\n</div>"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if 'mesh_large_df_sty' in globals():\n",
    "    # groupby and bar module\n",
    "    mesh_large_df_sty_less = mesh_large_df_sty[['Preferred Label','Semantic Types Label']]\n",
    "\n",
    "    ### groupby.describe analysis by rsid--------------------\n",
    "    mesh_large_df_sty_less_count = mesh_large_df_sty_less.groupby('Semantic Types Label').describe().reset_index()\n",
    "    mesh_large_df_sty_less_count.columns = mesh_large_df_sty_less_count.columns.to_flat_index()\n",
    "    new_column_names = ['Semantic Types Label', 'mesh-count', 'mesh-unique','mesh-top','mesh-freq']\n",
    "    mesh_large_df_sty_less_count.columns = new_column_names\n",
    "\n",
    "    mesh_large_df_sty_less_count_sort = mesh_large_df_sty_less_count.sort_values(by='mesh-count',ascending=False).reset_index(drop=True)\n",
    "    print('MESH_STY')\n",
    "    print('Semantic Type count:',len(mesh_large_df_sty_less_count_sort),'\\n\\nMeSH count for Sty:')\n",
    "    display(mesh_large_df_sty_less_count_sort[['mesh-count','Semantic Types Label']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T14:02:16.918568Z",
     "start_time": "2024-11-15T14:02:16.220531Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 400x500 with 1 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Graph Barh\n",
    "num = 20\n",
    "x = mesh_large_df_sty_less_count_sort['Semantic Types Label'].iloc[:num]\n",
    "y = mesh_large_df_sty_less_count_sort['mesh-count'].iloc[:num]\n",
    "plt.figure(figsize=(4, len(x) *0.25))\n",
    "plt.title('Global Mesh- Semantic Type enrichment', loc='center',pad=10)\n",
    "\n",
    "plt.barh(x,y)\n",
    "plt.gca().invert_yaxis()\n",
    "plt.tick_params(axis='x', which='both', top=True, bottom=False, labeltop=True, labelbottom=False)\n",
    "#plt.xlabel('pmid count', position=(0.5, 1.08))\n",
    "plt.ylabel('Semantic Types')\n",
    "plt.xlabel('mesh', position=(0.5, 1.08))\n",
    "ax = plt.gca()\n",
    "ax.xaxis.set_label_position('top')\n",
    "# use log scale\n",
    "plt.gca().set_xscale('log')\n",
    "#plt.savefig('Reference Mesh- Semantic Type enrichment.png',dpi=300, bbox_inches = \"tight\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#'Global \"Mesh-Semantic Type\" Zipf s law'\n",
    "x = mesh_large_df_sty_less_count_sort['Semantic Types Label'].iloc[:]\n",
    "y = mesh_large_df_sty_less_count_sort['mesh-count'].sort_values().iloc[:]\n",
    "plt.figure(figsize=(6, 5))\n",
    "plt.title('Global \"Mesh-Semantic Type\" Zipf s law', loc='center',pad=10)\n",
    "\n",
    "plt.plot(x,y)\n",
    "plt.gca().invert_yaxis()\n",
    "\n",
    "plt.ylabel('mesh number')\n",
    "plt.xlabel('semantic type rank', position=(0.5, 1.08))\n",
    "\n",
    "plt.gca().set_xscale('log')\n",
    "plt.gca().set_yscale('log')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## Density Rugged Plot\n",
    "data = mesh_large_df_sty_less_count_sort['mesh-count'].iloc[:]\n",
    "\n",
    "sns.set_style('whitegrid')\n",
    "sns.kdeplot(np.array(data), bw_method=0.5)\n",
    "sns.rugplot(np.array(data), color='r')\n",
    "\n",
    "#plt.yscale('log')\n",
    "plt.title('Mesh abundance for Semantic Type')\n",
    "#plt.yscale('log')\n",
    "print('Mesh abundance for Semantic Type:')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Analyze MESH-STY-LITVAR (subset of MESH-STY.csv)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "print('mesh_litvar_sty mesh:',mesh_litvar_sty['Preferred Label'].nunique())\n",
    "\n",
    "memory = mesh_litvar_sty.memory_usage().sum()\n",
    "print(f'The memory_usage of mesh_litvar_sty is {memory/ (1024 * 1024):.2f} MB.')\n",
    "\n",
    "file_size = os.path.getsize('ref-mesh/MESH_STY_LITVAR1.csv')\n",
    "print(f'The size of MESH_STY_LITVAR1.csv is {file_size/ (1024 * 1024):.2f} MB.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Sort MeSH-Sty-Litvar"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# groupby and bar module\n",
    "mesh_litvar_sty_less = mesh_litvar_sty[['Preferred Label','Semantic Types Label']]\n",
    "\n",
    "### groupby.describe analysis by rsid--------------------\n",
    "mesh_litvar_sty_less_count = mesh_litvar_sty_less.groupby('Semantic Types Label').describe().reset_index()\n",
    "mesh_litvar_sty_less_count.columns = mesh_litvar_sty_less_count.columns.to_flat_index()\n",
    "new_column_names = ['Semantic Types Label', 'mesh-count', 'mesh-unique','mesh-top','mesh-freq']\n",
    "mesh_litvar_sty_less_count.columns = new_column_names\n",
    "\n",
    "mesh_litvar_sty_less_count_sort = mesh_litvar_sty_less_count.sort_values(by='mesh-count',ascending=False).reset_index(drop=True)\n",
    "print('MESH_STY_LITVAR1')\n",
    "mesh_litvar_sty_less_count_sort[['mesh-count','Semantic Types Label']]#.to_clipboard()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Graph Barh\n",
    "num = 20\n",
    "x = mesh_litvar_sty_less_count_sort['Semantic Types Label'].iloc[:num]\n",
    "y = mesh_litvar_sty_less_count_sort['mesh-count'].iloc[:num]\n",
    "plt.figure(figsize=(4, len(x)*0.25))\n",
    "plt.title('Litvar1 Mesh- Semantic Type enrichment', loc='center',pad=10)\n",
    "\n",
    "plt.barh(x,y)\n",
    "plt.gca().invert_yaxis()\n",
    "plt.tick_params(axis='x', which='both', top=True, bottom=False, labeltop=True, labelbottom=False)\n",
    "plt.ylabel('Semantic Types')\n",
    "plt.xlabel('mesh', position=(0.5, 1.08))\n",
    "ax = plt.gca()\n",
    "ax.xaxis.set_label_position('top')\n",
    "# use log scale\n",
    "plt.gca().set_xscale('log')\n",
    "#plt.savefig('Reference Mesh- Semantic Type enrichment.png',dpi=300, bbox_inches = \"tight\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#'Global \"Mesh-Semantic Type\" Zipf s law'\n",
    "x = mesh_litvar_sty_less_count_sort['Semantic Types Label'].iloc[:]\n",
    "y = mesh_litvar_sty_less_count_sort['mesh-count'].sort_values().iloc[:]\n",
    "plt.figure(figsize=(6, 5))\n",
    "plt.title('LitVar \"Mesh-Semantic Type\" Zipf s law', loc='center',pad=10)\n",
    "\n",
    "plt.plot(x,y)\n",
    "plt.gca().invert_yaxis()\n",
    "\n",
    "plt.ylabel('mesh number')\n",
    "plt.xlabel('semantic type rank', position=(0.5, 1.08))\n",
    "\n",
    "plt.gca().set_xscale('log')\n",
    "plt.gca().set_yscale('log')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# VISUALIZE DIFFERENCES\n",
    "\n",
    "#reload sorted df\n",
    "if 'mesh_large_df_sty_less_count_sort' in locals() and isinstance(mesh_large_df_sty_less_count_sort, pd.DataFrame):\n",
    "    pass\n",
    "else:\n",
    "    mesh_large_df_sty_less = mesh_large_df_sty[['Preferred Label','Semantic Types Label']]\n",
    "    mesh_large_df_sty_less_count = mesh_large_df_sty_less.groupby('Semantic Types Label').describe().reset_index()\n",
    "    mesh_large_df_sty_less_count.columns = mesh_large_df_sty_less_count.columns.to_flat_index()\n",
    "    new_column_names = ['Semantic Types Label', 'mesh-count', 'mesh-unique','mesh-top','mesh-freq']\n",
    "    mesh_large_df_sty_less_count.columns = new_column_names\n",
    "    mesh_large_df_sty_less_count_sort = mesh_large_df_sty_less_count.sort_values(by='mesh-count',ascending=False).reset_index(drop=True)\n",
    "if 'mesh_litvar_sty_less_count_sort' in locals() and isinstance(mesh_litvar_sty_less_count_sort, pd.DataFrame):\n",
    "    pass\n",
    "else:\n",
    "    mesh_litvar_sty_less = mesh_litvar_sty[['Preferred Label','Semantic Types Label']]\n",
    "    mesh_litvar_sty_less_count = mesh_litvar_sty_less.groupby('Semantic Types Label').describe().reset_index()\n",
    "    mesh_litvar_sty_less_count.columns = mesh_litvar_sty_less_count.columns.to_flat_index()\n",
    "    new_column_names = ['Semantic Types Label', 'mesh-count', 'mesh-unique','mesh-top','mesh-freq']\n",
    "    mesh_litvar_sty_less_count.columns = new_column_names\n",
    "    mesh_litvar_sty_less_count_sort = mesh_litvar_sty_less_count.sort_values(by='mesh-count',ascending=False).reset_index(drop=True)\n",
    "\n",
    "# create two sample dataframes\n",
    "df1 = mesh_large_df_sty_less_count_sort[['mesh-count', 'Semantic Types Label']]\n",
    "df2 = mesh_litvar_sty_less_count_sort[['mesh-count', 'Semantic Types Label']]\n",
    "\n",
    "# add a 'source' column to each dataframe\n",
    "df1['source'] = 'global'\n",
    "df2['source'] = 'litvar'\n",
    "\n",
    "# combine the dataframes\n",
    "combined_df = pd.merge(df1, df2, on=['Semantic Types Label'], how='outer', suffixes=('_df1', '_df2'))\n",
    "\n",
    "# sort the dataframe by column 'A'\n",
    "#combined_df = combined_df.sort_values('mesh-count')\n",
    "\n",
    "# reset the index\n",
    "combined_df = combined_df.reset_index(drop=True)\n",
    "\n",
    "# display the combined dataframe\n",
    "combined_df = combined_df[-(combined_df['Semantic Types Label'] == 'Drug Delivery Device')]\n",
    "combined_df\n",
    "mesh_large_df_sty_less['Preferred Label'].nunique()\n",
    "combined_df = combined_df.sort_values(by='mesh-count_df2', ascending=False).reset_index(drop=True)\n",
    "\n",
    "\n",
    "# define a formatting function that generates a proportional bar\n",
    "def format_bar(value):\n",
    "    max_value = combined_df['mesh-count_df1'].max().max()  # get the maximum value in the dataframe\n",
    "    bar_width = int(value / max_value * 100)  # calculate the width of the bar as a percentage\n",
    "    return f'<div style=\"background-color: blue; width: {bar_width}%\">{value}</div>'\n",
    "\n",
    "def format_bar_2(value):\n",
    "    max_value = combined_df['mesh-count_df2'].max().max()  # get the maximum value in the dataframe\n",
    "    bar_width = int(value / max_value * 100)  # calculate the width of the bar as a percentage\n",
    "    return f'<div style=\"background-color: blue; width: {bar_width}%\">{value}</div>'\n",
    "\n",
    "# apply the formatting function to the dataframe\n",
    "df_formatted = combined_df.style.format({'mesh-count_df1': format_bar, 'mesh-count_df2': format_bar_2})\n",
    "\n",
    "# save the formatted dataframe to an HTML file\n",
    "with open('formatted_dataframe.html', 'w') as f:\n",
    "    f.write(df_formatted.render())\n",
    "\n",
    "# display the formatted dataframe\n",
    "df_formatted"
   ]
  },
  {
   "cell_type": "markdown",
   "source": [],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. Reference MeSH build"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Build coherent and comprehensive lists of mesh term related to different topics\n",
    "- nutrigentics\n",
    "- neurodegenerative diseases\n",
    "- skin diseases\n",
    "- infective diseases\n",
    "- reproductive physiology"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Check available ref-MeSH lists:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T14:02:50.151953800Z",
     "start_time": "2024-11-15T14:02:49.958924900Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Available reference mesh lists:\n"
     ]
    },
    {
     "data": {
      "text/plain": "['neuro',\n 'neuro_grpm',\n 'neuro_sty',\n 'ng_aller',\n 'ng_cvd',\n 'ng_dmt2_ms',\n 'ng_eat_taste',\n 'ng_intol',\n 'infect',\n 'ng_nutri',\n 'ng_ob_bmi',\n 'ng_oxi_stress',\n 'ng_vitam',\n 'ng_xeno',\n 'repro',\n 'repro_clean',\n 'heart_failure',\n 'nutri_old']"
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Check available refs:\n",
    "folder_path = \"ref-mesh\"  # Replace with the actual folder path\n",
    "#---------------------------\n",
    "\n",
    "# Create a file path pattern to match CSV files\n",
    "file_pattern = os.path.join(folder_path, \"*.csv\")\n",
    "\n",
    "# Use glob to get a list of file paths matching the pattern\n",
    "csv_files = glob.glob(file_pattern)\n",
    "csv_files_name = []\n",
    "# Print the list of CSV files\n",
    "for file in csv_files:\n",
    "    file_name = os.path.basename(file)\n",
    "    csv_files_name.append(file_name)\n",
    "\n",
    "print('Available reference mesh lists:')\n",
    "csv_files_df = pd.Series(csv_files_name)\n",
    "ref_files_df = csv_files_df[csv_files_df.str.contains('ref_mesh_')].reset_index(drop=True)\n",
    "ref_files_df_small = csv_files_df[csv_files_df.str.contains('small_ref_mesh_ng_')].reset_index(drop=True)\n",
    "\n",
    "file_names = [x.split('mesh_')[1].split('.csv')[0] for x in ref_files_df]\n",
    "file_names"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build ref-MeSH list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-04-11T17:03:14.424897700Z",
     "start_time": "2024-04-11T17:03:14.158146800Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "             Preferred Label  Semantic Types Label  \\\n1  Electronic Health Records  Intellectual Product   \n2              Consent Forms  Intellectual Product   \n3     Genealogy and Heraldry  Intellectual Product   \n4               Publications  Intellectual Product   \n5    Pharmaceutical Services  Intellectual Product   \n\n                                            Class ID  mesh_id Semantic Types  \\\n1  http://purl.bioontology.org/ontology/MESH/D057286  D057286           T170   \n2  http://purl.bioontology.org/ontology/MESH/D032962  D032962           T170   \n3  http://purl.bioontology.org/ontology/MESH/D005789  D005789           T170   \n4  http://purl.bioontology.org/ontology/MESH/D011642  D011642           T170   \n5  http://purl.bioontology.org/ontology/MESH/D010593  D010593           T170   \n\n                                            Synonyms  \\\n1  Electronic Medical Record|Medical Record, Elec...   \n2  Informed Consent Documents|Informed Consent Fo...   \n3  Geneology and Heraldry|Heraldry and Genealogy|...   \n4                                        Publication   \n5  Pharmaceutical Service|Services, Pharmaceutic|...   \n\n                                             Parents                CUI  \\\n1  http://purl.bioontology.org/ontology/MESH/D016347  C0079150|C2362543   \n2  http://purl.bioontology.org/ontology/MESH/D007...           C0009797   \n3  http://purl.bioontology.org/ontology/MESH/D006664  C0019232|C0017299   \n4  http://purl.bioontology.org/ontology/MESH/D003146           C0034036   \n5  http://purl.bioontology.org/ontology/MESH/D006296  C0031321|C1449618   \n\n                                AQL  \\\n1  CL EC ES HI IS LJ OG SD SN ST TD   \n2        CL EC ES HI LJ OG SN ST TD   \n3                               NaN   \n4        CL EC ES HI LJ SD SN ST TD   \n5     CL EC ES HI LJ OG SD SN ST TD   \n\n                                              TERMUI  \n1  T751083|T048813|T048814|T048815|T749887|T75108...  \n2                    T466073|T021753|T466072|T450735  \n3                    T017448|T017447|T017449|T017451  \n4                                            T034372  \n5  T552624|T031274|T031273|T031272|T031271|T03126...  ",
      "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>Preferred Label</th>\n      <th>Semantic Types Label</th>\n      <th>Class ID</th>\n      <th>mesh_id</th>\n      <th>Semantic Types</th>\n      <th>Synonyms</th>\n      <th>Parents</th>\n      <th>CUI</th>\n      <th>AQL</th>\n      <th>TERMUI</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>1</th>\n      <td>Electronic Health Records</td>\n      <td>Intellectual Product</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D057286</td>\n      <td>D057286</td>\n      <td>T170</td>\n      <td>Electronic Medical Record|Medical Record, Elec...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D016347</td>\n      <td>C0079150|C2362543</td>\n      <td>CL EC ES HI IS LJ OG SD SN ST TD</td>\n      <td>T751083|T048813|T048814|T048815|T749887|T75108...</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>Consent Forms</td>\n      <td>Intellectual Product</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D032962</td>\n      <td>D032962</td>\n      <td>T170</td>\n      <td>Informed Consent Documents|Informed Consent Fo...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D007...</td>\n      <td>C0009797</td>\n      <td>CL EC ES HI LJ OG SN ST TD</td>\n      <td>T466073|T021753|T466072|T450735</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>Genealogy and Heraldry</td>\n      <td>Intellectual Product</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D005789</td>\n      <td>D005789</td>\n      <td>T170</td>\n      <td>Geneology and Heraldry|Heraldry and Genealogy|...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D006664</td>\n      <td>C0019232|C0017299</td>\n      <td>NaN</td>\n      <td>T017448|T017447|T017449|T017451</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>Publications</td>\n      <td>Intellectual Product</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D011642</td>\n      <td>D011642</td>\n      <td>T170</td>\n      <td>Publication</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D003146</td>\n      <td>C0034036</td>\n      <td>CL EC ES HI LJ SD SN ST TD</td>\n      <td>T034372</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>Pharmaceutical Services</td>\n      <td>Intellectual Product</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D010593</td>\n      <td>D010593</td>\n      <td>T170</td>\n      <td>Pharmaceutical Service|Services, Pharmaceutic|...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D006296</td>\n      <td>C0031321|C1449618</td>\n      <td>CL EC ES HI LJ OG SD SN ST TD</td>\n      <td>T552624|T031274|T031273|T031272|T031271|T03126...</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "if 'mesh_litvar_sty' not in locals():\n",
    "    mesh_litvar_sty = pd.read_csv('ref-mesh/MESH_STY_LITVAR1.csv',index_col=0)\n",
    "    print('MESH_STY_LITVAR1 imported from csv')\n",
    "mesh_litvar_sty.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Generate Ref-MeSH from Topic Term Lists\n",
    "For list of terms defined by the user: search in Preferred Labels and Synonyms corresponding mesh entry"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T14:03:21.394170600Z",
     "start_time": "2024-11-15T14:03:21.270175900Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": "0    [Obesity, overweight and body weight control, ...\n1    [cardiovascular diseases, physiological proces...\n2    [Diabetes Melitus Type II and metabolic syndrome]\n3    [Vitamin metabolism and Vitamins recommended i...\n4               [eating behaviour and taste sensation]\n5                                  [food intolerances]\n6                                     [food allergies]\n7                      [diet-induced oxidative stress]\n8                          [metabolism of xenobiotics]\ndtype: object"
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Define Biomedical topics\n",
    "nutritional_topic = [['diseases and disorders realted to nutrition and diet ', 'diet, food consuption, eating behaviour and nutrition']]\n",
    "infective_topic = [['infective agents, bacteria, virus and protozoan','infective diseases']]\n",
    "reproductive_topic = [['reproductive system physiology','reproductive system pathology', 'Assisted reproductive technology']]\n",
    "female_infertility_topic = [['female infertility, genetic imprinting and maternal effect']]\n",
    "\n",
    "nutritional_topics = [\n",
    "    ['Obesity, overweight and body weight control', 'compulsive eating behavior'],\n",
    "    ['cardiovascular diseases','physiological processes realted to cardiovascular diseases','lipid metabolism in the context of cardiovascular diseases'],\n",
    "    ['Diabetes Melitus Type II and metabolic syndrome'],\n",
    "    ['Vitamin metabolism and Vitamins recommended intake levels','Micronutrients metabolism and Micronutrient recommended intake levels', 'disease related to vitamins and micronutrients deficiency'],\n",
    "    ['eating behaviour and taste sensation'],\n",
    "    ['food intolerances'],\n",
    "    ['food allergies'],\n",
    "    ['diet-induced oxidative stress'],\n",
    "    ['metabolism of xenobiotics'],\n",
    "]\n",
    "chosen_topic = nutritional_topics\n",
    "pd.Series(chosen_topic)"
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "### NLTK Processor"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package punkt_tab to\n",
      "[nltk_data]     C:\\Users\\Utente\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Unzipping tokenizers\\punkt_tab.zip.\n",
      "[nltk_data] Downloading package stopwords to\n",
      "[nltk_data]     C:\\Users\\Utente\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package stopwords is already up-to-date!\n"
     ]
    },
    {
     "data": {
      "text/plain": "True"
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# download nltk requirements\n",
    "import nltk\n",
    "nltk.download('punkt_tab')\n",
    "nltk.download('stopwords')"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-14T12:25:25.918583800Z",
     "start_time": "2024-11-14T12:25:25.280583100Z"
    }
   },
   "execution_count": 8
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "###python\n",
    "# Updating libraries from NLTK and adding string for punctuation removal\n",
    "import nltk\n",
    "import string\n",
    "from nltk.stem import PorterStemmer\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.tokenize import word_tokenize\n",
    "\n",
    "# Function for text preprocessing: tokenization, stopping, stemming, lowering, and punctuation removal\n",
    "def process_text(text, replace=False):\n",
    "    # Removing punctuation\n",
    "    if replace:\n",
    "        text = text.replace('-',' ')\n",
    "    text = text.translate(str.maketrans('', '', string.punctuation))\n",
    "\n",
    "    # Tokenizing the text\n",
    "    tokens = word_tokenize(text)\n",
    "    # Lowercasing all tokens\n",
    "    tokens = [token.lower() for token in tokens]\n",
    "    # Removing stop words\n",
    "    stop_words = set(stopwords.words('english'))\n",
    "    filtered_tokens = [token for token in tokens if token not in stop_words]\n",
    "    # Stemming words\n",
    "    stemmer = PorterStemmer()\n",
    "    stemmed_tokens = [stemmer.stem(token) for token in filtered_tokens]\n",
    "\n",
    "    return stemmed_tokens\n",
    "\n",
    "## Clean ref-mesh from biases term >>>>>>>>>>>>>>>\n",
    "def clean_mesh(get_rid_list, df, save_clean=False, path = 'ref-mesh/candidate_ref_mesh/'):\n",
    "    if not os.path.exists(path):\n",
    "        os.makedirs(path)\n",
    "    for get_rid in get_rid_list:\n",
    "        mask = df['Preferred Label'].str.contains(get_rid)\n",
    "        df = df[-mask]\n",
    "\n",
    "    if save_clean:\n",
    "        df.to_csv(path+'ref_mesh_'+topic_label+'.csv')\n",
    "    return df\n",
    "\n",
    "## Rip semantic-types from ref-mesh >>>>>>>>>>>>>>>\n",
    "def semantic_ripper(rip_list, topic_mesh_all, save=False, path = 'ref-mesh/candidate_ref_mesh/', tag=''):\n",
    "    if not os.path.exists(path):\n",
    "        os.makedirs(path)\n",
    "    mask = topic_mesh_all['Semantic Types Label'].isin(rip_list)\n",
    "    topic_mesh_all_redux = topic_mesh_all[-mask]\n",
    "\n",
    "    if save:\n",
    "        topic_mesh_all_redux[['Preferred Label', 'Semantic Types Label', 'mesh_id']].to_csv(path+'ref_mesh_'+topic_label+tag+'.csv')\n",
    "    return topic_mesh_all_redux\n",
    "\n",
    "# Sample text\n",
    "text = \"This is an example sentence demonstrating the stemming, stop-word removal, lowering operations, and punctuation removal!\"\n",
    "preprocessed_text = process_text(text)\n",
    "print(preprocessed_text)"
   ],
   "metadata": {
    "collapsed": false
   },
   "execution_count": null
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 31741/31741 [00:15<00:00, 2070.25it/s]\n"
     ]
    },
    {
     "data": {
      "text/plain": "                     Preferred Label                     label_tokens  \\\n1          Electronic Health Records       [electron, health, record]   \n2                      Consent Forms                  [consent, form]   \n3             Genealogy and Heraldry             [genealog, heraldri]   \n4                       Publications                         [public]   \n5            Pharmaceutical Services             [pharmaceut, servic]   \n7              Programming Languages               [program, languag]   \n8   Mental Status and Dementia Tests  [mental, statu, dementia, test]   \n9                          Mythology                       [mytholog]   \n10               Models, Theoretical                 [model, theoret]   \n11             Models, Psychological               [model, psycholog]   \n\n       mesh_id  Semantic Types Label  \n1      D057286  Intellectual Product  \n2      D032962  Intellectual Product  \n3      D005789  Intellectual Product  \n4      D011642  Intellectual Product  \n5      D010593  Intellectual Product  \n7      D011381  Intellectual Product  \n8   D000073216  Intellectual Product  \n9      D009229  Intellectual Product  \n10     D008962  Intellectual Product  \n11     D008960  Intellectual Product  ",
      "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>Preferred Label</th>\n      <th>label_tokens</th>\n      <th>mesh_id</th>\n      <th>Semantic Types Label</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>1</th>\n      <td>Electronic Health Records</td>\n      <td>[electron, health, record]</td>\n      <td>D057286</td>\n      <td>Intellectual Product</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>Consent Forms</td>\n      <td>[consent, form]</td>\n      <td>D032962</td>\n      <td>Intellectual Product</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>Genealogy and Heraldry</td>\n      <td>[genealog, heraldri]</td>\n      <td>D005789</td>\n      <td>Intellectual Product</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>Publications</td>\n      <td>[public]</td>\n      <td>D011642</td>\n      <td>Intellectual Product</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>Pharmaceutical Services</td>\n      <td>[pharmaceut, servic]</td>\n      <td>D010593</td>\n      <td>Intellectual Product</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>Programming Languages</td>\n      <td>[program, languag]</td>\n      <td>D011381</td>\n      <td>Intellectual Product</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>Mental Status and Dementia Tests</td>\n      <td>[mental, statu, dementia, test]</td>\n      <td>D000073216</td>\n      <td>Intellectual Product</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>Mythology</td>\n      <td>[mytholog]</td>\n      <td>D009229</td>\n      <td>Intellectual Product</td>\n    </tr>\n    <tr>\n      <th>10</th>\n      <td>Models, Theoretical</td>\n      <td>[model, theoret]</td>\n      <td>D008962</td>\n      <td>Intellectual Product</td>\n    </tr>\n    <tr>\n      <th>11</th>\n      <td>Models, Psychological</td>\n      <td>[model, psycholog]</td>\n      <td>D008960</td>\n      <td>Intellectual Product</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#====================================================================================\n",
    "# Tokenize and process MeSH 'Preferred Labels' =======================================\n",
    "mother_df = mesh_litvar_sty\n",
    "replace = True\n",
    "# Use tqdm with apply to show a progress bar\n",
    "tqdm.pandas()\n",
    "mother_df['label_tokens'] = mother_df['Preferred Label'].progress_apply(process_text, replace = replace)\n",
    "mother_df[['Preferred Label','label_tokens','mesh_id','Semantic Types Label']].head(10)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T14:07:05.604937900Z",
     "start_time": "2024-11-15T14:06:50.189936800Z"
    }
   },
   "execution_count": 18
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Import topic-terms"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "88\n",
      "topic_terms_ng_xeno.csv\n"
     ]
    }
   ],
   "source": [
    "import re\n",
    "\n",
    "folder_path = \"ref-mesh/topic_terms\"\n",
    "file_pattern = os.path.join(folder_path, \"*.csv\")\n",
    "csv_files = glob.glob(file_pattern)\n",
    "csv_files_name = []\n",
    "for file in csv_files:\n",
    "    file_name = os.path.basename(file)\n",
    "    csv_files_name.append(file_name)\n",
    "\n",
    "#print('Available topic terms:')\n",
    "csv_files_df = pd.Series(csv_files_name)\n",
    "topic_terms_file = csv_files_df[csv_files_df.str.contains('topic_')]\n",
    "\n",
    "file_id = int(input('Available topic terms, select index:\\n'+str(pd.Series(topic_terms_file))))\n",
    "filename = topic_terms_file[file_id]\n",
    "topic_all_ser_df = pd.read_csv(folder_path+'/'+filename)\n",
    "\n",
    "topic_label = re.search('topic_terms_(.*).csv', filename).group(1)\n",
    "#-----------------------------------------------\n",
    "\n",
    "topic_all_ser = topic_all_ser_df.topic_terms.to_list()\n",
    "print(len(topic_all_ser))\n",
    "print(filename)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T14:08:18.340008700Z",
     "start_time": "2024-11-15T14:08:14.556922900Z"
    }
   },
   "execution_count": 19
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### cross 'topic-terms' with 'litvar_mesh'"
   ]
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0:00:04.852998\n",
      "topic_mesh_ng_xeno created\n",
      "   topic terms: 88\n",
      "   mesh found 270\n",
      "   semantic groups 49\n",
      "ref_mesh_ng_xeno.csv alreday exists\n",
      "\n",
      "(Please validate manually MeSH List to remove ambiguity)\n"
     ]
    }
   ],
   "source": [
    "#get mesh\n",
    "topic_mesh_all = pd.DataFrame(columns=mother_df.columns) #Empty DataFrame, same structure\n",
    "time1 =datetime.now()\n",
    "for term in topic_all_ser:\n",
    "    term_tokens = process_text(term, replace = replace)\n",
    "    mask = mother_df['label_tokens'].apply(lambda x: all(term in x for term in term_tokens))\n",
    "    meshy = mother_df[mask]\n",
    "\n",
    "    topic_mesh_all = pd.concat([meshy,topic_mesh_all])\n",
    "time2 =datetime.now()\n",
    "print(time2-time1)\n",
    "\n",
    "#analyze semantics\n",
    "topic_mesh_all_group = topic_mesh_all.groupby('Semantic Types Label').describe()\n",
    "topic_mesh_all_group_sem = topic_mesh_all_group.index.to_list()\n",
    "\n",
    "new_name = 'topic_mesh_'+topic_label\n",
    "exec(f'{new_name} = topic_mesh_all')\n",
    "print(new_name, 'created')\n",
    "print('   topic terms:', len(topic_all_ser))\n",
    "print('   mesh found', topic_mesh_all['Preferred Label'].nunique())\n",
    "print('   semantic groups',topic_mesh_all['Semantic Types Label'].nunique())\n",
    "\n",
    "overwrite = False\n",
    "# GET ref mesh list\n",
    "if os.path.exists('ref-mesh/ref_mesh_'+topic_label+'.csv'):\n",
    "    print('ref_mesh_'+topic_label+'.csv alreday exists')\n",
    "    if overwrite:\n",
    "        globals()[new_name][['Preferred Label', 'Semantic Types Label']].to_csv('ref-mesh/ref_mesh_'+topic_label+'.csv')\n",
    "        print('overwritten')\n",
    "else:\n",
    "    globals()[new_name][['Preferred Label', 'Semantic Types Label']].to_csv('ref-mesh/ref_mesh_'+topic_label+'.csv')\n",
    "    print('\\nref_mesh_'+topic_label+'.csv saved')\n",
    "\n",
    "topic_mesh_all[['Preferred Label','mesh_id','Semantic Types Label']].drop_duplicates()\n",
    "\n",
    "print(\"\\n(Please validate manually MeSH List to remove ambiguity)\")"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T14:09:59.203419100Z",
     "start_time": "2024-11-15T14:09:53.804386Z"
    }
   },
   "execution_count": 22
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Clean and save 'candidate_ref_mesh'"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "                           Preferred Label              Semantic Types Label  \\\n32921   Cytochrome P-450 CYP1A2 Inhibitors           Pharmacologic Substance   \n50219     Cytochrome P-450 CYP1A2 Inducers           Pharmacologic Substance   \n112029             Cytochrome P-450 CYP1A2   Amino Acid, Peptide, or Protein   \n402979             Cytochrome P-450 CYP1A2                            Enzyme   \n503807                      Food Additives                              Food   \n...                                    ...                               ...   \n514528            Metabolic Clearance Rate                Clinical Attribute   \n515280                          Metabolism                 Biologic Function   \n515695     Metabolic Networks and Pathways                Functional Concept   \n330242                         Xenobiotics     Biologically Active Substance   \n469836                         Xenobiotics  Hazardous or Poisonous Substance   \n\n                                                 Class ID  mesh_id  \\\n32921   http://purl.bioontology.org/ontology/MESH/D065609  D065609   \n50219   http://purl.bioontology.org/ontology/MESH/D065694  D065694   \n112029  http://purl.bioontology.org/ontology/MESH/D019388  D019388   \n402979  http://purl.bioontology.org/ontology/MESH/D019388  D019388   \n503807  http://purl.bioontology.org/ontology/MESH/D005503  D005503   \n...                                                   ...      ...   \n514528  http://purl.bioontology.org/ontology/MESH/D008657  D008657   \n515280  http://purl.bioontology.org/ontology/MESH/D008660  D008660   \n515695  http://purl.bioontology.org/ontology/MESH/D053858  D053858   \n330242  http://purl.bioontology.org/ontology/MESH/D015262  D015262   \n469836  http://purl.bioontology.org/ontology/MESH/D015262  D015262   \n\n       Semantic Types                                           Synonyms  \\\n32921            T121  Cytochrome P 450 CYP1A2 Inhibitors|P450 CYP1A2...   \n50219            T121  Cytochrome P 450 CYP1A2 Inducers|P450 CYP1A2 I...   \n112029           T116  Demethylase, Caffeine|Cytochrome P 450 CYP1A2|...   \n402979           T126  Demethylase, Caffeine|Cytochrome P 450 CYP1A2|...   \n503807           T168       Additives, Food|Additive, Food|Food Additive   \n...               ...                                                ...   \n514528           T201  Total Body Clearance Rate|Clearance Rates, Met...   \n515280           T038  Process, Metabolic|Metabolic Phenomenon|Anabol...   \n515695           T169  Metabolic Networks|Metabolic Pathway|Pathways,...   \n330242           T123                                         Xenobiotic   \n469836           T131                                         Xenobiotic   \n\n                                                  Parents  \\\n32921   http://purl.bioontology.org/ontology/MESH/D065607   \n50219   http://purl.bioontology.org/ontology/MESH/D065693   \n112029  http://purl.bioontology.org/ontology/MESH/D000...   \n402979  http://purl.bioontology.org/ontology/MESH/D000...   \n503807  http://purl.bioontology.org/ontology/MESH/D000...   \n...                                                   ...   \n514528  http://purl.bioontology.org/ontology/MESH/D019...   \n515280                http://www.w3.org/2002/07/owl#Thing   \n515695  http://purl.bioontology.org/ontology/MESH/D008660   \n330242  http://purl.bioontology.org/ontology/MESH/D004364   \n469836  http://purl.bioontology.org/ontology/MESH/D004364   \n\n                                        CUI  \\\n32921                              C3850068   \n50219                              C3850050   \n112029                             C0207509   \n402979                             C0207509   \n503807                             C0016453   \n...                                     ...   \n514528                    C0025515|C0087101   \n515280  C0025519|C2350543|C0699900|C0220781   \n515695           C1721099|C1291081|C1706062   \n330242                             C0043335   \n469836                             C0043335   \n\n                                                      AQL  \\\n32921   AD AE AN BL CF CH CL CS EC HI IM IP ME PD PK P...   \n50219   AD AE AN BL CF CH CL CS EC HI IM IP ME PD PK P...   \n112029  AD AE AN BI BL CF CH CL CS DE DF EC GE HI IM I...   \n402979  AD AE AN BI BL CF CH CL CS DE DF EC GE HI IM I...   \n503807  AD AE AN CH CL CS EC HI IP ME PD PK PO RE SD S...   \n...                                                   ...   \n514528                                     DE GE IM PH RE   \n515280                                     DE GE IM PH RE   \n515695                                  DE ES GE IM PH RE   \n330242  AD AE AG AI AN BL CF CH CL CS EC HI IM IP ME P...   \n469836  AD AE AG AI AN BL CF CH CL CS EC HI IM IP ME P...   \n\n                                                   TERMUI  \\\n32921                             T853455|T852927|T853454   \n50219                             T853458|T853457|T853456   \n112029  T057790|T057795|T057793|T057789|T057791|T05779...   \n402979  T057790|T057795|T057793|T057789|T057791|T05779...   \n503807                                            T016735   \n...                                                   ...   \n514528                                    T025744|T025745   \n515280  T718994|T721534|T721533|T025757|T025759|T71899...   \n515695                            T671142|T671140|T671141   \n330242                                 T044826|T000996442   \n469836                                 T044826|T000996442   \n\n                                  label_tokens  \n32921   [cytochrom, p, 450, cyp1a2, inhibitor]  \n50219       [cytochrom, p, 450, cyp1a2, induc]  \n112029             [cytochrom, p, 450, cyp1a2]  \n402979             [cytochrom, p, 450, cyp1a2]  \n503807                           [food, addit]  \n...                                        ...  \n514528               [metabol, clearanc, rate]  \n515280                               [metabol]  \n515695             [metabol, network, pathway]  \n330242                              [xenobiot]  \n469836                              [xenobiot]  \n\n[420 rows x 11 columns]",
      "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>Preferred Label</th>\n      <th>Semantic Types Label</th>\n      <th>Class ID</th>\n      <th>mesh_id</th>\n      <th>Semantic Types</th>\n      <th>Synonyms</th>\n      <th>Parents</th>\n      <th>CUI</th>\n      <th>AQL</th>\n      <th>TERMUI</th>\n      <th>label_tokens</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>32921</th>\n      <td>Cytochrome P-450 CYP1A2 Inhibitors</td>\n      <td>Pharmacologic Substance</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D065609</td>\n      <td>D065609</td>\n      <td>T121</td>\n      <td>Cytochrome P 450 CYP1A2 Inhibitors|P450 CYP1A2...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D065607</td>\n      <td>C3850068</td>\n      <td>AD AE AN BL CF CH CL CS EC HI IM IP ME PD PK P...</td>\n      <td>T853455|T852927|T853454</td>\n      <td>[cytochrom, p, 450, cyp1a2, inhibitor]</td>\n    </tr>\n    <tr>\n      <th>50219</th>\n      <td>Cytochrome P-450 CYP1A2 Inducers</td>\n      <td>Pharmacologic Substance</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D065694</td>\n      <td>D065694</td>\n      <td>T121</td>\n      <td>Cytochrome P 450 CYP1A2 Inducers|P450 CYP1A2 I...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D065693</td>\n      <td>C3850050</td>\n      <td>AD AE AN BL CF CH CL CS EC HI IM IP ME PD PK P...</td>\n      <td>T853458|T853457|T853456</td>\n      <td>[cytochrom, p, 450, cyp1a2, induc]</td>\n    </tr>\n    <tr>\n      <th>112029</th>\n      <td>Cytochrome P-450 CYP1A2</td>\n      <td>Amino Acid, Peptide, or Protein</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D019388</td>\n      <td>D019388</td>\n      <td>T116</td>\n      <td>Demethylase, Caffeine|Cytochrome P 450 CYP1A2|...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D000...</td>\n      <td>C0207509</td>\n      <td>AD AE AN BI BL CF CH CL CS DE DF EC GE HI IM I...</td>\n      <td>T057790|T057795|T057793|T057789|T057791|T05779...</td>\n      <td>[cytochrom, p, 450, cyp1a2]</td>\n    </tr>\n    <tr>\n      <th>402979</th>\n      <td>Cytochrome P-450 CYP1A2</td>\n      <td>Enzyme</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D019388</td>\n      <td>D019388</td>\n      <td>T126</td>\n      <td>Demethylase, Caffeine|Cytochrome P 450 CYP1A2|...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D000...</td>\n      <td>C0207509</td>\n      <td>AD AE AN BI BL CF CH CL CS DE DF EC GE HI IM I...</td>\n      <td>T057790|T057795|T057793|T057789|T057791|T05779...</td>\n      <td>[cytochrom, p, 450, cyp1a2]</td>\n    </tr>\n    <tr>\n      <th>503807</th>\n      <td>Food Additives</td>\n      <td>Food</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D005503</td>\n      <td>D005503</td>\n      <td>T168</td>\n      <td>Additives, Food|Additive, Food|Food Additive</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D000...</td>\n      <td>C0016453</td>\n      <td>AD AE AN CH CL CS EC HI IP ME PD PK PO RE SD S...</td>\n      <td>T016735</td>\n      <td>[food, addit]</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>514528</th>\n      <td>Metabolic Clearance Rate</td>\n      <td>Clinical Attribute</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D008657</td>\n      <td>D008657</td>\n      <td>T201</td>\n      <td>Total Body Clearance Rate|Clearance Rates, Met...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D019...</td>\n      <td>C0025515|C0087101</td>\n      <td>DE GE IM PH RE</td>\n      <td>T025744|T025745</td>\n      <td>[metabol, clearanc, rate]</td>\n    </tr>\n    <tr>\n      <th>515280</th>\n      <td>Metabolism</td>\n      <td>Biologic Function</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D008660</td>\n      <td>D008660</td>\n      <td>T038</td>\n      <td>Process, Metabolic|Metabolic Phenomenon|Anabol...</td>\n      <td>http://www.w3.org/2002/07/owl#Thing</td>\n      <td>C0025519|C2350543|C0699900|C0220781</td>\n      <td>DE GE IM PH RE</td>\n      <td>T718994|T721534|T721533|T025757|T025759|T71899...</td>\n      <td>[metabol]</td>\n    </tr>\n    <tr>\n      <th>515695</th>\n      <td>Metabolic Networks and Pathways</td>\n      <td>Functional Concept</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D053858</td>\n      <td>D053858</td>\n      <td>T169</td>\n      <td>Metabolic Networks|Metabolic Pathway|Pathways,...</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D008660</td>\n      <td>C1721099|C1291081|C1706062</td>\n      <td>DE ES GE IM PH RE</td>\n      <td>T671142|T671140|T671141</td>\n      <td>[metabol, network, pathway]</td>\n    </tr>\n    <tr>\n      <th>330242</th>\n      <td>Xenobiotics</td>\n      <td>Biologically Active Substance</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D015262</td>\n      <td>D015262</td>\n      <td>T123</td>\n      <td>Xenobiotic</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D004364</td>\n      <td>C0043335</td>\n      <td>AD AE AG AI AN BL CF CH CL CS EC HI IM IP ME P...</td>\n      <td>T044826|T000996442</td>\n      <td>[xenobiot]</td>\n    </tr>\n    <tr>\n      <th>469836</th>\n      <td>Xenobiotics</td>\n      <td>Hazardous or Poisonous Substance</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D015262</td>\n      <td>D015262</td>\n      <td>T131</td>\n      <td>Xenobiotic</td>\n      <td>http://purl.bioontology.org/ontology/MESH/D004364</td>\n      <td>C0043335</td>\n      <td>AD AE AG AI AN BL CF CH CL CS EC HI IM IP ME P...</td>\n      <td>T044826|T000996442</td>\n      <td>[xenobiot]</td>\n    </tr>\n  </tbody>\n</table>\n<p>420 rows × 11 columns</p>\n</div>"
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Clean and save 'candidate_ref_mesh'\n",
    "folder_path_ = \"ref-mesh/candidate_ref_mesh\"\n",
    "if not os.path.exists(folder_path_):\n",
    "    os.makedirs(folder_path_)\n",
    "\n",
    "mesh_rid_list = ['oma','Apop','Cell','gen Ty','Cilia','Cytokinesis','DNA','Fluorescein','Intercellular Signaling','Leukem','Lysergic Acid','Loeys-Dietz','Mitochondrial','Peptide Hormones','Toll-Like','Tumor', 'Neoplasm', 'RNA','Gene','Diethyl', 'Nucleos', 'Motif', 'Domain', 'Virus', 'virus','Infant', 'Child', 'Adolescent', 'Elder', 'Maternal', 'Youth', 'Man', 'Woman', 'National', 'Neoplasm', 'Reproductive', 'Sexual', 'Genome', 'Animal', 'Doping', 'Social', 'Urban','Health C','Health E','Health F','Health I','Health P','Health S','Health T','ty Health','le Health','ic Health','Polymor','Genetic','Risk F','Proteins','Protein','DNA','RNA', 'Polymor','Genetic', 'Pharma', 'Cosme', 'Drug', 'Distribution']\n",
    "\n",
    "topic_mesh_all = clean_mesh(mesh_rid_list, topic_mesh_all, save_clean=False)\n",
    "\n",
    "semantic_rip_list = ['Cell', 'Receptor', 'Hormone', 'Tissue', 'Body Part, Organ, or Organ Component', 'Congenital Abnormality', 'Anatomical Abnormality', 'Indicator, Reagent, or Diagnostic Aid', 'Neoplastic Process', 'Mammal', 'Gene or Genome', 'Element, Ion, or Isotope']\n",
    "\n",
    "topic_mesh_all = semantic_ripper(semantic_rip_list, topic_mesh_all, save=True)\n",
    "topic_mesh_all"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-11-15T14:11:22.168647800Z",
     "start_time": "2024-11-15T14:11:21.979613500Z"
    }
   },
   "execution_count": 26
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4. Ref-MeSH Operations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Check available ref-MeSH lists"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Check avalable refs:\n",
    "folder_path = \"ref-mesh\"  # Replace with the actual folder path\n",
    "#---------------------------\n",
    "\n",
    "# Create a file path pattern to match CSV files\n",
    "file_pattern = os.path.join(folder_path, \"*.csv\")\n",
    "\n",
    "# Use glob to get a list of file paths matching the pattern\n",
    "csv_files = glob.glob(file_pattern)\n",
    "csv_files_name = []\n",
    "# Print the list of CSV files\n",
    "for file in csv_files:\n",
    "    file_name = os.path.basename(file)\n",
    "    csv_files_name.append(file_name)\n",
    "\n",
    "print('Available reference mesh lists:')\n",
    "csv_files_df = pd.Series(csv_files_name)\n",
    "ref_files_df = csv_files_df[csv_files_df.str.contains('ref_mesh_')].reset_index(drop=True)\n",
    "ref_files_df_small = csv_files_df[csv_files_df.str.contains('small_ref_mesh_ng_')].reset_index(drop=True)\n",
    "\n",
    "file_names = [x.split('mesh_')[1].split('.csv')[0] for x in ref_files_df]\n",
    "file_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ref_names = ['ref_'] * 10\n",
    "ref_names = [string + str(i) for i, string in enumerate(ref_names)]\n",
    "\n",
    "ref_list = []\n",
    "for name, ref, save  in zip(ref_names, ref_files_df, ref_files_df):\n",
    "    df = pd.read_csv('ref-mesh/'+ref, index_col=0).reset_index(drop=True)\n",
    "    df = df.drop('Semantic Types Label', axis=1).drop_duplicates()\n",
    "    df = df.sort_values(by='mesh_id',ascending=True)\n",
    "    globals()[name] = df\n",
    "    #df.to_csv('ref-mesh/small_'+save) # 'small' for representation purposes only\n",
    "    ref_list.append(df)\n",
    "\n",
    "ref_list[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "## 1. Ref MeSH cleaner"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Mesh Cleaner"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "#Get rid from bias generating MeSH and run again GRPM Module 3\n",
    "\n",
    "# put here manually selected mesh to purge\n",
    "xeno = ['Polymorphism, Genetic', 'DNA Methylation','Liver Neoplasms','Energy Metabolism','Metabolism, Inborn Errors','X-Ray Absorption Spectroscopy','Blood-Testis Barrier']\n",
    "oxi_stress = ['Food Deprivation','Food Analysis','Food, Formulated','Food Microbiology','Food Intolerance','Food Quality','Food Industry','Foods, Specialized','Food Chain','Foods, Specialized','Food Chain']\n",
    "intol = ['Depression','Osteoporosis','Immunotherapy','Anxiety','Anti-Inflammatory Agents, Non-Steroidal','Immunotherapy, Adoptive','Desensitization, Immunologic','Peanut Hypersensitivity','Milk Hypersensitivity', 'Egg Hypersensitivity','Immunotherapy, Active','Neurologic Manifestations' ,'Infectious Anemia Virus, Equine','Nut and Peanut Hypersensitivity' ,'Diarrhea Virus 1, Bovine Viral','Eye Movement Desensitization Reprocessing','Diarrhea Virus 2, Bovine Viral']\n",
    "eat_taste = ['Blood Glucose','Mouth Neoplasms','Glucose Transporter Type 1','Glucose Transporter Type 2','Auditory Perception','Citric Acid Cycle','Glucose Transporter Type 4','Loeys-Dietz Syndrome','United States Food and Drug Administration','Sodium-Glucose Transporter 2','Sodium-Glucose Transporter 2 Inhibitors','Glucose-6-Phosphate','Glucose Transporter Type 3','Pitch Perception','Depth Perception','Glucose-1-Phosphate Adenylyltransferase','Glucose Dehydrogenases','Glucosephosphates']\n",
    "cvd = ['DNA, Mitochondrial','Cell Differentiation','Protein Transport','Mitochondrial Proteins','Toll-Like Receptors','Genome, Mitochondrial','Genes, Mitochondrial','Mitochondrial Precursor Protein Import Complex Proteins','Mitochondrial Precursor Protein Import Complex Proteins','Mitochondrial Proton-Translocating ATPases','Mitochondrial Dynamics','Mitochondrial Membranes','Leukemia, Plasma Cell','Heart Neoplasms']\n",
    "dmt2_ms = [ 'MicroRNAs', 'Mitochondria', 'DNA, Mitochondrial', 'Mitochondrial Proteins', 'Pancreatic Neoplasms','Autophagy','Mitochondrial Diseases','Genome, Mitochondrial','Genes, Mitochondrial','Mitochondrial Precursor Protein Import Complex Proteins','Mitochondrial Precursor Protein Import Complex Proteins','Mitochondrial Dynamics','Mitochondrial Membranes','Mitochondrial Myopathies','RNA, Mitochondrial', 'Mitochondrial Encephalomyopathies','AIDS-Associated Nephropathy','Familial Primary Pulmonary Hypertension','Mitochondria, Heart','Autophagy-Related Protein-1 Homolog','Ocular Hypertension','Autophagy-Related Protein 7','Mitochondrial Trifunctional Protein','Mitochondrial Permeability Transition Pore','Mitochondrial Uncoupling Proteins','Mitochondrial ADP, ATP Translocases','Autophagy-Related Protein 8 Family','Mitochondrial Ribosomes','Mitochondrial Trifunctional Protein, alpha Subunit','Autophagy-Related Protein 12','Mitochondrial Turnover','Lysergic Acid Diethylamide','Mitochondrial Trifunctional Protein, beta Subunit','Chaperone-Mediated Autophagy','Mitochondrial Swelling','Mitochondrial Transmembrane Permeability-Driven Necrosis','Mitochondrial Size']\n",
    "ob_bmi = ['Infant','Child','Adolescent','Elder','Maternal','Youth','Man','Woman','National','Neoplasm''Epidemiolo','Reproductive','Sexual','Genome','Animal','Doping','Social','Urban''Health C','Health E','Health F','Health I','Health P','Health S','Health T','ty Health','le Health','ic Health','Polymor','Genetic','Risk F']\n",
    "\n",
    "# note: to list above stands for examples (ref mesh already cleaned in archive)\n",
    "\n",
    "#clean mesh ref\n",
    "get_rid_list = xeno\n",
    "\n",
    "#check mesh ref\n",
    "topic_label = 'ng_' + 'xeno'\n",
    "path_ = 'ref-mesh/candidate_ref_mesh/'\n",
    "dff = pd.read_csv(path_+'ref_mesh_'+topic_label+'.csv', index_col=0)\n",
    "\n",
    "dff_rid = clean_mesh(get_rid_list, dff, path = 'ref-mesh/candidate_ref_mesh/', save_clean =True)\n",
    "\n",
    "print(dff.describe(),'\\n-----------------------------------------------------------------')\n",
    "print(dff_rid.describe())"
   ],
   "metadata": {
    "collapsed": false
   },
   "execution_count": null
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Semantic type ripper"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       Preferred Label                 Semantic Types Label     mesh_id\n",
      "count              609                                  609         609\n",
      "unique             243                                   58         243\n",
      "top      Diet, Healthy  Therapeutic or Preventive Procedure  D000072001\n",
      "freq                18                                   60          18 \n",
      "-----------------------------------------------------------------\n",
      "       Preferred Label                 Semantic Types Label     mesh_id\n",
      "count              543                                  543         543\n",
      "unique             232                                   51         232\n",
      "top      Diet, Healthy  Therapeutic or Preventive Procedure  D000072001\n",
      "freq                18                                   60          18\n"
     ]
    }
   ],
   "source": [
    "#Get rid from bias generating MeSH and run again GRPM Module 3\n",
    "\n",
    "semantic_rip_list = ['Cell', 'Receptor', 'Hormone', 'Tissue', 'Body Part, Organ, or Organ Component', 'Congenital Abnormality', 'Anatomical Abnormality', 'Indicator, Reagent, or Diagnostic Aid', 'Neoplastic Process', 'Mammal', 'Gene or Genome', 'Element, Ion, or Isotope']\n",
    "\n",
    "#check mesh ref\n",
    "topic_label = 'ng_' + 'ob_bmi'\n",
    "#path_ = 'ref-mesh/candidate_ref_mesh/'\n",
    "path_ = 'ref-mesh/'\n",
    "dff = pd.read_csv(path_+'ref_mesh_'+topic_label+'.csv', index_col=0)\n",
    "\n",
    "dff_rid = semantic_ripper(semantic_rip_list, dff, path = 'ref-mesh/', save=False)\n",
    "\n",
    "print(dff.describe(),'\\n-----------------------------------------------------------------')\n",
    "print(dff_rid.describe())"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-04-11T17:10:06.578508400Z",
     "start_time": "2024-04-11T17:10:06.334984500Z"
    }
   },
   "execution_count": 10
  },
  {
   "cell_type": "markdown",
   "source": [
    "## 2. create JSON sheet for ref-mesh lists\n",
    "https://jsoneditoronline.org/#left=local.lejobi&right=local.lejobi\n",
    "https://www.convertjson.com/json-to-html-table.htm"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "import json\n",
    "import csv\n",
    "\n",
    "folder_path = \"ref-mesh\"\n",
    "\n",
    "num_elements = len(ref_files_df)  # Number of elements you want in the list\n",
    "value_list = []\n",
    "for i in range(1, num_elements + 1):\n",
    "    element = \"file_pat\" + str(i)\n",
    "    value_list.append(element)\n",
    "\n",
    "jdata_list_n = ['jdata_'] * 10\n",
    "jdata_list_n = [string + str(i) for i, string in enumerate(jdata_list_n)]\n",
    "\n",
    "jdata_list = []\n",
    "for name in jdata_list_n:\n",
    "    empty =[]\n",
    "    globals()[name] = empty\n",
    "    jdata_list.append(empty)\n",
    "\n",
    "\n",
    "file_patterns = []\n",
    "choose_ref = ref_files_df # or ref_files_df_small\n",
    "for i in choose_ref:\n",
    "    file_patterns.append(os.path.join(folder_path, i))\n",
    "\n",
    "for i, j  in zip(range(len(ref_files_df)), jdata_list):\n",
    "\n",
    "    with open(file_patterns[i], 'r') as file1:\n",
    "        csv_reader = csv.DictReader(file1)\n",
    "        # for row in csv_reader:\n",
    "        #     j_data1.append(row)\n",
    "        for row in csv_reader:\n",
    "            modified_row = {key: value for key, value in row.items() if key != ''} #drop col\n",
    "            j.append(modified_row)"
   ],
   "metadata": {
    "collapsed": false
   },
   "execution_count": null
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#combined_data = {\n",
    "#    'data1': j_data1,\n",
    "#    'data2': j_data2 }\n",
    "\n",
    "combined_data = {}\n",
    "\n",
    "titles = [\n",
    "    \"Genaral Nutrition\",\n",
    "    \"Obesity, Weight Control and Compulsive Eating\",\n",
    "    \"Cardiovascular Health and Lipid Metabolism\",\n",
    "    \"Diabetes Mellitus Type II and Metabolic Syndrome\",\n",
    "    \"Vitamin and Micronutrients Metabolism and Deficiency-Related Diseases\",\n",
    "    \"Eating Behavior and Taste Sensation\",\n",
    "    \"Food Intolerances\",\n",
    "    \"Food Allergies\",\n",
    "    \"Diet-induced Oxidative Stress\",\n",
    "    \"Xenobiotics Metabolism\",\n",
    "]\n",
    "\n",
    "for tit, j_data in zip(titles, jdata_list):\n",
    "    key = tit\n",
    "    combined_data[key] = j_data\n",
    "\n",
    "#for i, j_data in enumerate(jdata_list, start=1):\n",
    "#    key = 'data' + str(i)\n",
    "#    combined_data[key] = j_data\n",
    "\n",
    "print(combined_data)\n",
    "#pd.DataFrame(combined_data)\n",
    "import pyperclip\n",
    "pyperclip.copy(str(combined_data))\n",
    "\n",
    "json_file = 'ref-mesh/nutrigenetis_referencemesh_small.json'\n",
    "with open(json_file, 'w') as outfile:\n",
    "    json.dump(combined_data, outfile, indent=4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. add mesh_id to MeSH list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# add mesh id to every mesh list\n",
    "label = 'ng_nutri'\n",
    "dff = pd.read_csv('ref-mesh/ref_mesh_'+label+'.csv',index_col=0).reset_index(drop=True)#.drop('Unnamed: 0', axis=1)\n",
    "\n",
    "dff_id = pd.merge(dff, mesh_id_ref_df, left_on='Preferred Label', right_on='Preferred Label')\n",
    "dff_id"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dff_id.to_csv('ref-mesh/ref_mesh_'+label+'.csv')\n",
    "#grpm_db_merge_id = grpm_db_merge_id.drop('Preferred Label', axis=1)\n",
    "check = pd.read_csv('ref-mesh/ref_mesh_'+label+'.csv', index_col=0)\n",
    "check"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "chack = check.drop('mesh_id_y',axis=1)\n",
    "check = check.rename(columns={'mesh_id_y': 'mesh_id'})\n",
    "check.to_csv('ref-mesh/ref_mesh_'+label+'.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5. Generate Random Mesh"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random grpm mesh list from MESH-STY"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "mesh_large_df_sty"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Generate Random mesh list from complete df\n",
    "print(mesh_large_df_sty['Preferred Label'].nunique())\n",
    "mesh_large_df_sty_mesh = mesh_large_df_sty['Preferred Label'].drop_duplicates()\n",
    "random_path = 'ref-mesh/random_lists/'\n",
    "\n",
    "size = 603\n",
    "sample_01 = mesh_large_df_sty_mesh.sample(n=size, random_state = 78954)\n",
    "sample_02 = mesh_large_df_sty_mesh.sample(n=size, random_state = 12245)\n",
    "sample_03 = mesh_large_df_sty_mesh.sample(n=size, random_state = 87498)\n",
    "sample_04 = mesh_large_df_sty_mesh.sample(n=size, random_state = 56798)\n",
    "sample_05 = mesh_large_df_sty_mesh.sample(n=size, random_state = 34565)\n",
    "sample_06 = mesh_large_df_sty_mesh.sample(n=size, random_state = 76523)\n",
    "sample_07 = mesh_large_df_sty_mesh.sample(n=size, random_state = 78968)\n",
    "sample_08 = mesh_large_df_sty_mesh.sample(n=size, random_state = 56845)\n",
    "sample_09 = mesh_large_df_sty_mesh.sample(n=size, random_state = 76624)\n",
    "sample_10 = mesh_large_df_sty_mesh.sample(n=size, random_state = 23845)\n",
    "\n",
    "sample_01.to_csv(random_path+'random_01.csv')\n",
    "sample_02.to_csv(random_path+'random_02.csv')\n",
    "sample_03.to_csv(random_path+'random_03.csv')\n",
    "sample_04.to_csv(random_path+'random_04.csv')\n",
    "sample_05.to_csv(random_path+'random_05.csv')\n",
    "sample_06.to_csv(random_path+'random_06.csv')\n",
    "sample_07.to_csv(random_path+'random_07.csv')\n",
    "sample_08.to_csv(random_path+'random_08.csv')\n",
    "sample_09.to_csv(random_path+'random_09.csv')\n",
    "sample_10.to_csv(random_path+'random_10.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random grpm mesh list from MESH-STY-LITVAR\n",
    "Generate Random Mesh list from grpm df\n",
    "\n",
    "    grpm_db_mesh = pd.read_csv('ref-mesh/grpm_db_mesh.csv')\n",
    "    grpm_db_mesh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import os\n",
    "if os.path.exists('ref-mesh/grpm_db_mesh.csv'):\n",
    "    grpm_db_mesh = pd.read_csv('ref-mesh/grpm_db_mesh.csv',index_col=0)\n",
    "    print('grpm mesh imported from archive')\n",
    "else:\n",
    "    grpmdb_table = pd.read_csv('grpm_dataset\\grpm_db_pcg\\grpm_table_output.csv')\n",
    "    grpm_db_mesh = pd.DataFrame(grpmdb_table.mesh.drop_duplicates())\n",
    "    type(grpm_db_mesh)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "rand = pd.read_csv('ref-mesh/random_grpm_03.csv')\n",
    "mask= grpm_db_mesh.mesh.isin(rand['mesh'])\n",
    "grpm_db_mesh[mask]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# randomize 10 samples\n",
    "size = 450\n",
    "sample_grpm_01 = grpm_db_mesh.mesh.sample(n=size, random_state = 78954)\n",
    "sample_grpm_02 = grpm_db_mesh.mesh.sample(n=size, random_state = 12245)\n",
    "sample_grpm_03 = grpm_db_mesh.mesh.sample(n=size, random_state = 87498)\n",
    "sample_grpm_04 = grpm_db_mesh.mesh.sample(n=size, random_state = 56798)\n",
    "sample_grpm_05 = grpm_db_mesh.mesh.sample(n=size, random_state = 34565)\n",
    "sample_grpm_06 = grpm_db_mesh.mesh.sample(n=size, random_state = 76523)\n",
    "sample_grpm_07 = grpm_db_mesh.mesh.sample(n=size, random_state = 78968)\n",
    "sample_grpm_08 = grpm_db_mesh.mesh.sample(n=size, random_state = 56845)\n",
    "sample_grpm_09 = grpm_db_mesh.mesh.sample(n=size, random_state = 76624)\n",
    "sample_grpm_00 = grpm_db_mesh.mesh.sample(n=size, random_state = 23845)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#check fidelity\n",
    "mask= grpm_db_mesh.mesh.isin(sample_grpm_01)\n",
    "grpm_db_mesh[mask]\n",
    "#type(sample_grpm_10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# save random samples\n",
    "random_path = 'ref-mesh/random_lists/'\n",
    "\n",
    "if os.path.exists('ref-mesh/random_grpm_00.csv'):\n",
    "    sample_grpm_01 = pd.read_csv(random_path+'random_grpm_01.csv',index_col=0)\n",
    "    sample_grpm_02 = pd.read_csv(random_path+'random_grpm_02.csv',index_col=0)\n",
    "    sample_grpm_03 = pd.read_csv(random_path+'random_grpm_03.csv',index_col=0)\n",
    "    sample_grpm_04 = pd.read_csv(random_path+'random_grpm_04.csv',index_col=0)\n",
    "    sample_grpm_05 = pd.read_csv(random_path+'random_grpm_05.csv',index_col=0)\n",
    "    sample_grpm_06 = pd.read_csv(random_path+'random_grpm_06.csv',index_col=0)\n",
    "    sample_grpm_07 = pd.read_csv(random_path+'random_grpm_07.csv',index_col=0)\n",
    "    sample_grpm_08 = pd.read_csv(random_path+'random_grpm_08.csv',index_col=0)\n",
    "    sample_grpm_09 = pd.read_csv(random_path+'random_grpm_09.csv',index_col=0)\n",
    "    sample_grpm_00 = pd.read_csv(random_path+'random_grpm_00.csv',index_col=0)\n",
    "\n",
    "save_rand = False\n",
    "if save_rand == True:\n",
    "    sample_grpm_01.to_csv(random_path+'random_grpm_01.csv')\n",
    "    sample_grpm_02.to_csv(random_path+'random_grpm_02.csv')\n",
    "    sample_grpm_03.to_csv(random_path+'random_grpm_03.csv')\n",
    "    sample_grpm_04.to_csv(random_path+'random_grpm_04.csv')\n",
    "    sample_grpm_05.to_csv(random_path+'random_grpm_05.csv')\n",
    "    sample_grpm_06.to_csv(random_path+'random_grpm_06.csv')\n",
    "    sample_grpm_07.to_csv(random_path+'random_grpm_07.csv')\n",
    "    sample_grpm_08.to_csv(random_path+'random_grpm_08.csv')\n",
    "    sample_grpm_09.to_csv(random_path+'random_grpm_09.csv')\n",
    "    sample_grpm_00.to_csv(random_path+'random_grpm_00.csv')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "type(sample_grpm_01)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "sample_grpm_01_list = sample_grpm_01.to_list()\n",
    "sample_grpm_02_list = sample_grpm_02.to_list()\n",
    "sample_grpm_03_list = sample_grpm_03.to_list()\n",
    "sample_grpm_04_list = sample_grpm_04.to_list()\n",
    "sample_grpm_05_list = sample_grpm_05.to_list()\n",
    "sample_grpm_06_list = sample_grpm_06.to_list()\n",
    "sample_grpm_07_list = sample_grpm_07.to_list()\n",
    "sample_grpm_08_list = sample_grpm_08.to_list()\n",
    "sample_grpm_09_list = sample_grpm_09.to_list()\n",
    "sample_grpm_00_list = sample_grpm_00.to_list()\n",
    "\n",
    "lists= [sample_grpm_01_list,\n",
    "        sample_grpm_02_list,\n",
    "        sample_grpm_03_list,\n",
    "        sample_grpm_04_list,\n",
    "        sample_grpm_05_list,\n",
    "        sample_grpm_06_list,\n",
    "        sample_grpm_07_list,\n",
    "        sample_grpm_08_list,\n",
    "        sample_grpm_09_list,\n",
    "        sample_grpm_00_list]\n",
    "\n",
    "num_lists = len(lists)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Coocuurance method"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#(1) Initialize a 2D list of zeros with dimensions equal to the number of lists\n",
    "cooccur_matrix = [[0] * num_lists for i in range(num_lists)]\n",
    "cooccur_matrix = np.zeros((len(lists), len(lists)), dtype=int)\n",
    "\n",
    "type(cooccur_matrix)\n",
    "print(len(set(lists[1]) & set(lists[1])))\n",
    "print(len(set(lists[1]) & set(lists[3])))\n",
    "\n",
    "# Loop over all pairs of lists and count the number of co-occurring elements\n",
    "for i in range(num_lists):\n",
    "    for j in range(num_lists):\n",
    "        if i == j:\n",
    "            cooccur_matrix[i][j] = len(set(lists[i]))\n",
    "        else:\n",
    "            cooccur_matrix[i][j] = len(set(lists[i]) & set(lists[j])) # use set() to extract unique elements\n",
    "\n",
    "print(type(cooccur_matrix))\n",
    "\n",
    "# Convert the 2D list to a Pandas DataFrame\n",
    "cooccur_df = pd.DataFrame(cooccur_matrix,\n",
    "                          columns=['random{}'.format(i+1) for i in range(num_lists)],\n",
    "                          index=['random{}'.format(i+1) for i in range(num_lists)])\n",
    "\n",
    "# Print the resulting DataFrame\n",
    "cooccur_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Read Random grpm mesh list without rep"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# COOCCURANCE MATRIX MODULE------------\n",
    "# second matrix build check\n",
    "sample_grpm_01_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep1.csv')\n",
    "sample_grpm_02_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep2.csv')\n",
    "sample_grpm_03_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep3.csv')\n",
    "sample_grpm_04_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep4.csv')\n",
    "sample_grpm_05_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep5.csv')\n",
    "sample_grpm_06_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep6.csv')\n",
    "sample_grpm_07_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep7.csv')\n",
    "sample_grpm_08_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep8.csv')\n",
    "sample_grpm_09_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep9.csv')\n",
    "sample_grpm_00_norep = pd.read_csv(random_path+'random_grpm_mesh_norep/random_grpm_norep10.csv')\n",
    "\n",
    "lists= [sample_grpm_01_norep.mesh.to_list(),\n",
    "        sample_grpm_02_norep.mesh.to_list(),\n",
    "        sample_grpm_03_norep.mesh.to_list(),\n",
    "        sample_grpm_04_norep.mesh.to_list(),\n",
    "        sample_grpm_05_norep.mesh.to_list(),\n",
    "        sample_grpm_06_norep.mesh.to_list(),\n",
    "        sample_grpm_07_norep.mesh.to_list(),\n",
    "        sample_grpm_08_norep.mesh.to_list(),\n",
    "        sample_grpm_09_norep.mesh.to_list(),\n",
    "        sample_grpm_00_norep.mesh.to_list()]\n",
    "\n",
    "num_lists = len(lists)\n",
    "\n",
    "# Initialize a 2D list of zeros with dimensions equal to the number of lists\n",
    "cooccur_matrix = [[0] * num_lists for i in range(num_lists)]\n",
    "type(cooccur_matrix)\n",
    "print(len(set(lists[1]) & set(lists[1])))\n",
    "print(len(set(lists[1]) & set(lists[3])))\n",
    "\n",
    "# Loop over all pairs of lists and count the number of co-occurring elements\n",
    "for i in range(num_lists):\n",
    "    for j in range(num_lists):\n",
    "        if i == j:\n",
    "            cooccur_matrix[i][j] = len(lists[i])\n",
    "        else:\n",
    "            cooccur_matrix[i][j] = len(set(lists[i]) & set(lists[j]))\n",
    "\n",
    "# Convert the 2D list to a Pandas DataFrame\n",
    "cooccur_df = pd.DataFrame(cooccur_matrix, columns=['random{}'.format(i+1) for i in range(num_lists)], index=['random{}'.format(i+1) for i in range(num_lists)])\n",
    "\n",
    "# Print the resulting DataFrame\n",
    "cooccur_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Print the resulting matrix with row and column headers\n",
    "print(', '.join([''] + ['list{}'.format(i+1) for i in range(num_lists)]))\n",
    "\n",
    "for i in range(num_lists):\n",
    "    row = ['list{}'.format(i+1)]\n",
    "    row.extend(cooccur_matrix[i])\n",
    "    print(', '.join(str(x) for x in row))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "crea 2 set da 5 random mesh senza ripetizioni"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create Random grpm mesh list without rep from MESH-STY-LITVAR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# randomize\n",
    "sample_grpm_full = grpm_db_mesh.mesh.sample(n=len(grpm_db_mesh), random_state = 782954)\n",
    "sample_grpm_full"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "# chunk size\n",
    "size = 450\n",
    "# Split the DataFrame into smaller DataFrames of chunk length 450\n",
    "chunks = np.array_split(sample_grpm_full, len(sample_grpm_full) // 450 + 1)\n",
    "\n",
    "# Print the number of chunks and the length of each chunk\n",
    "for i, chunk in enumerate(chunks):\n",
    "    print('Chunk {}: {} rows'.format(i+1, len(chunk)))\n",
    "    pass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Save each DataFrame chunk to a separate CSV file\n",
    "for i, chunk in enumerate(chunks):\n",
    "    chunk.to_csv('ref-mesh/random_grpm_norep{}.csv'.format(i+1), index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Extras"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Analyze Reference MeSH list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import ast\n",
    "if os.path.exists('ref-mesh/ref_mesh_nest.csv'):\n",
    "    ref = pd.read_csv('ref-mesh/ref_mesh_nest.csv', sep=\";\")\n",
    "    print('nested ref mesh', ref['Class ID'].nunique())\n",
    "\n",
    "    ref['Semantic Types'] = ref['Semantic Types'].apply(ast.literal_eval)\n",
    "    #otherwise= dfg['col1'] = df['col1'].str.strip('[]').str.split(', ').apply(lambda x: [int(i) for i in x])\n",
    "    sty = pd.read_csv('ref-mesh/MeshSTY-code.csv',sep=';')\n",
    "    sty = sty.rename(columns={'ID':'Semantic Types'})\n",
    "    print(ref)\n",
    "else:\n",
    "    print('ref_mesh_nest.csv not available')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "if 'ref' in globals():\n",
    "    ref_large = []\n",
    "    for i in range(len(ref)):\n",
    "        for sem in ref['Semantic Types'][i]: #dfrspost = mother table\n",
    "            out = ref['Preferred Label'][i],ref['Class ID'][i],sem,ref['Synonyms'][i],ref['Parents'][i],ref['CUI'][i],ref['AQL'][i],ref['TERMUI'][i]\n",
    "            ref_large.append(out)\n",
    "\n",
    "    ref_large_df = pd.DataFrame(ref_large)\n",
    "    new_col_names = ['Preferred Label','Class ID','Semantic Types','Synonyms','Parents','CUI','AQL','TERMUI']\n",
    "    ref_large_df.columns = new_col_names\n",
    "    display(ref_large_df)\n",
    "    #ref_large_df.to_csv('ref-mesh/ref_mesh_large.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Mesh semantics analyze on Ref list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## set df on ref mesh sty large\n",
    "\n",
    "if os.path.exists('ref-mesh/ref_mesh_sty_large.csv'):\n",
    "    ref_sty = pd.read_csv('ref-mesh/ref_mesh_sty_large.csv', index_col=0)\n",
    "else:\n",
    "    ref_sty = pd.merge(ref_large_df, sty, on='Semantic Types', how='inner').reset_index(drop=True)\n",
    "    #Add rsid coulmn con merge\n",
    "    #rspmidmesh_merge = pd.merge(pmidmesh, rsidpmid, on= 'pmid', how='inner').drop_duplicates().reindex(columns=['pmid', 'rsid', 'mesh'])\n",
    "    ref_sty = ref_sty.rename(columns={'Preferred Label_y':'Semantic Types Label','Preferred Label_x':'Preferred Label'})\n",
    "    #ref_sty = ref_sty.drop('Column1',axis=1)\n",
    "    #ref_sty.to_csv('ref-mesh/ref_mesh_sty_large.csv')\n",
    "\n",
    "ref_sty = ref_sty[['Preferred Label', 'Semantic Types Label', 'Class ID', 'Semantic Types', 'Synonyms', 'Parents', 'CUI', 'AQL', 'TERMUI']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "print(ref_sty['Preferred Label'].nunique(), 'mesh total')\n",
    "ref_sty"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "if 'ref_sty' in globals():\n",
    "    #modulo groupby and bar\n",
    "    ref_sty_less = ref_sty[['Preferred Label','Semantic Types Label']]\n",
    "\n",
    "    ### groupby.describe analysis by rsid--------------------\n",
    "    ref_sty_less_count = ref_sty_less.groupby('Semantic Types Label').describe().reset_index()\n",
    "    ref_sty_less_count.columns = ref_sty_less_count.columns.to_flat_index()\n",
    "    new_column_names = ['Semantic Types Label', 'mesh-count', 'mesh-unique','mesh-top','mesh-freq']\n",
    "    ref_sty_less_count.columns = new_column_names\n",
    "    ref_sty_less_count_sort = ref_sty_less_count.sort_values(by='mesh-count',ascending=False).reset_index(drop=True)\n",
    "\n",
    "    x = ref_sty_less_count_sort['Semantic Types Label'].iloc[:]\n",
    "    y = ref_sty_less_count_sort['mesh-count'].iloc[:]\n",
    "    plt.figure(figsize=(4, len(ref_sty_less_count_sort)*0.25))\n",
    "    plt.title('Reference Mesh- Semantic Type enrichment', loc='center',pad=10)\n",
    "\n",
    "    plt.barh(x,y)\n",
    "    plt.gca().invert_yaxis()\n",
    "    plt.tick_params(axis='x', which='both', top=True, bottom=False, labeltop=True, labelbottom=False)\n",
    "    #plt.xlabel('pmid count', position=(0.5, 1.08))\n",
    "    plt.ylabel('Semantic Types')\n",
    "    plt.xlabel('mesh', position=(0.5, 1.08))\n",
    "    ax = plt.gca()\n",
    "    ax.xaxis.set_label_position('top')\n",
    "    # use log scale\n",
    "    plt.gca().set_xscale('log')\n",
    "    #plt.savefig('Reference Mesh- Semantic Type enrichment.png',dpi=300, bbox_inches = \"tight\")\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ref_sty_less_count_sort#['Semantic Types Label']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# analizing abundance\n",
    "#count = [12, 21]\n",
    "#remove_value = ref_sty_less_count_sort[ref_sty_less_count_sort['mesh-count'].isin(count)]['Semantic Types Label'].reset_index(drop=True).to_list()\n",
    "remove_value = ref_sty_less_count_sort[ref_sty_less_count_sort['mesh-count']> 56]['Semantic Types Label'].reset_index(drop=True).to_list()\n",
    "ref_sty[ref_sty['Semantic Types Label'].isin(remove_value)]['Semantic Types Label'].drop_duplicates()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### STY Ripper form complete ref mesh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "## Rip form complete ref mesh\n",
    "save = False\n",
    "rip_list = ['Cell', 'Receptor', 'Hormone', 'Tissue', 'Body Part, Organ, or Organ Component', 'Congenital Abnormality', 'Anatomical Abnormality', 'Indicator, Reagent, or Diagnostic Aid']\n",
    "\n",
    "remove_value = ref_sty_less_count_sort[ref_sty_less_count_sort['Semantic Types Label'].isin(rip_list)]['Semantic Types Label'].reset_index(drop=True).to_list()\n",
    "remove_value = ref_sty_less_count_sort[ref_sty_less_count_sort['mesh-count']> 56]['Semantic Types Label'].reset_index(drop=True).to_list()\n",
    "\n",
    "#new_ref_sty = ref_sty[ref_sty['Semantic Types Label'].isin(remove_value)]\n",
    "mask = ref_sty['Semantic Types Label'].isin(rip_list)\n",
    "new_ref_sty = ref_sty[-mask]\n",
    "\n",
    "if save == True:\n",
    "    new_ref_sty.to_csv('ref-mesh/new_ref_mesh_large_corrected.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "new_ref_sty['Preferred Label'].drop_duplicates().to_csv('ref-mesh/new_ref_mesh_corrected.csv')\n",
    "new_ref_sty['Preferred Label'].drop_duplicates()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Global Mesh df break through:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Search for all sons from a parent ID (trial)\n",
    "    goal: merge all child mesh together"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Search for all brothers from a parent ID:\n",
    "var4 = 'D000066888' #'Diet, Food, and Nutrition'\n",
    "var5 = 'D044623' #'Nutrition Therapy'\n",
    "var6 = 'D014808' #Nutritional and Metabolic Diseases\n",
    "var7 = 'D002318'#Cardiovascular Diseases\n",
    "var8 = 'D004066'#Digestive System Diseases\n",
    "var9 = 'D004700'#Endocrine System Diseases\n",
    "var10 = 'D006967'#Hypersensitivity\n",
    "\n",
    "sub = df[df['Parents'].str.contains(var4).fillna(False)]\n",
    "# use .dropna(inplace=True)   OR   df.fillna(0, inplace=True)\n",
    "sub"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Concatenate the results\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dfgb = pd.DataFrame(columns=df.columns) #Empty DataFrame, same structure\n",
    "#dfgg = pd.DataFrame()\n",
    "dfgg = sub\n",
    "# I MUST search in 'Parents' for the top category ID.\n",
    "\n",
    "#Parents Class ID list:\n",
    "kids = sub['Class ID'].tolist() #search for nested childs\n",
    "\n",
    "# Multiple search:\n",
    "for i in kids:\n",
    "    child = df[df['Parents'].str.contains(i).fillna(False)]\n",
    "    dfgg = dfgg.append([child])\n",
    "    dfgb = pd.concat([child, dfgb.loc[:]])\n",
    "    kids = child['Class ID'].tolist()\n",
    "    for i in kids:\n",
    "        child = df[df['Parents'].str.contains(i).fillna(False)]\n",
    "        dfgg = dfgg.append([child])\n",
    "        #dfgb = pd.concat([child, dfgb.loc[:]])\n",
    "        kids = child['Class ID'].tolist()\n",
    "        for i in kids:\n",
    "            child = df[df['Parents'].str.contains(i).fillna(False)]\n",
    "            dfgg = dfgg.append([child])\n",
    "            #dfgb = pd.concat([child, dfgb.loc[:]])\n",
    "            kids = child['Class ID'].tolist()\n",
    "            for i in kids:\n",
    "                child = df[df['Parents'].str.contains(i).fillna(False)]\n",
    "                dfgg = dfgg.append([child])\n",
    "                #dfgb = pd.concat([child, dfgb.loc[:]])\n",
    "                kids = child['Class ID'].tolist()\n",
    "                for i in kids:\n",
    "                    child = df[df['Parents'].str.contains(i).fillna(False)]\n",
    "                    dfgg = dfgg.append([child])\n",
    "                    #dfgb = pd.concat([child, dfgb.loc[:]])\n",
    "                    kids = child['Class ID'].tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dfgg.loc[:].drop_duplicates()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "chat_mesh_nutri#.drop_duplicates()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "len(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Create new empty dataframe\n",
    "dfg = pd.DataFrame(columns=df.columns)\n",
    "\n",
    "ds1 = df[df['Class ID'].str.contains(var3)]\n",
    "ds2 = df[df['Class ID'].str.contains(var4)]\n",
    "#dfg.append(arr, ignore_index=True)\n",
    "#dfg2 = dfg.append(ds1)\n",
    "dfg2 = pd.concat([ds2, dfg2.loc[:]])\n",
    "dfg2\n",
    "#USE CONCAT!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dd= df[['Preferred Label','MeSH Frequency']]\n",
    "ddf= dd.dropna(subset='Preferred Label')\n",
    "print(ddf.shape)\n",
    "print(dd.shape)\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "#plt.scatter(dd['Preferred Label'], dd['MeSH Frequency'])\n",
    "ddf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#df[\"Preferred Label\"].__contains__(\"human\")\n",
    "df2=df[df[\"Preferred Label\"].str.contains('Polymorphism')]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df3= df2[['Preferred Label','MeSH Frequency']]\n",
    "df3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df2=df[df[\"Preferred Label\"].str.contains('Diabetes')]\n",
    "df3= df2[['Preferred Label','MeSH Frequency']]\n",
    "df3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df2=df[df[\"Preferred Label\"].str.contains('Fabry')]\n",
    "df3= df2[['Preferred Label','MeSH Frequency']]\n",
    "df3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# links\n",
    " #https://www.nlm.nih.gov/databases/download/data_distrib_main.html\n",
    " #https://www.nlm.nih.gov/databases/download/mesh.html\n",
    " #https://id.nlm.nih.gov/mesh/swagger/ui"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# previous trials"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Comparison of specific MESH list with MESH big data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dn = pd.read_csv('ref-mesh/nbdb1-mesh.csv')\n",
    "\n",
    "print(len(dn['mesh'].drop_duplicates()))\n",
    "dga= df[df['Preferred Label'].isin(dn['mesh'])]\n",
    "\n",
    "#dgb = dn[-df['Preferred Label'].isin(dn['mesh'])].drop_duplicates()\n",
    "\n",
    "#see not contained in all-MESH\n",
    "dgaa= dga['Preferred Label']\n",
    "dgb = dn['mesh'].drop_duplicates()\n",
    "dgg = dgb[dgb.isin(dgaa)==False]\n",
    "dgg"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Interpolation with mesh selected form semantic type"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dn = pd.read_csv('ref-mesh/MeshEX-meshsorted.csv',sep=';')\n",
    "print(len(dn['Preferred Label'].drop_duplicates()))\n",
    "dnmask = dn.Column2.isin(['green'])\n",
    "dnfilter = dn[dnmask]\n",
    "dnfilter\n",
    "# moldulo 'Allineator'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 'Allineator' module for list value\n",
    "dnallign = []\n",
    "for i in range(len(dnfilter)):\n",
    "    for pmid in dnfilter['Semantic Types'][i]: #dfrspost = mother table\n",
    "        #out = (dfrspost['rsid'][i], pmid)\n",
    "        out = dnfilter['Preferred Label'][i], pmid\n",
    "        dnallign.append(out)\n",
    "print(type(dnallign))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### table explosion Module:\n",
    "\n",
    "data.csv = mesh; 'Semantic Types'; ID\n",
    "Bone Density,\"T201,T081\",D015519\n",
    "Ideal Body Weight,\"T201,T074,T081\",D056865ù\n",
    "\n",
    "split the SEM column by comma and explode the values to create a new row for each SEM value\n",
    "df = df.assign(SEM=df['SEM'].str.split(',')).explode('SEM')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# table explosion Module:\n",
    "\n",
    "# split the SEM column by comma and explode the values to create a new row for each SEM value\n",
    "dnfilter = dnfilter.assign(SEM=dnfilter['Semantic Types'].str.split(',')).explode('SEM')\n",
    "\n",
    "# print the updated table\n",
    "dnfilerstimple = dnfilter[['Preferred Label','Class ID','Semantic Types','SEM']]\n",
    "dnfiltsupersimple = dnfilter[['Preferred Label','SEM']].drop_duplicates()\n",
    "dnfiltsupersimple.to_clipboard()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Modulo analisi groupby:\n",
    "dnfilercount= dnfiltsupersimple.groupby('SEM').describe().reset_index()\n",
    "dnfilercount.columns = dnfilercount.columns.to_flat_index()\n",
    "new_column_names = ['SEM', 'Labes_count', 'Labes_count_unique','Labes_top','Labels_freq']\n",
    "dnfilercount.columns = new_column_names\n",
    "dnfilercountsort = dnfilercount.sort_values(by= 'Labes_count' ,ascending = False)\n",
    "dnfilercountsortsmall= dnfilercountsort[['SEM','Labes_count']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Add semantic name column #Use Merge\n",
    "semcode = pd.read_csv('ref-mesh/MeshSTY-code.csv',sep=';')\n",
    "new_column_names = ['Label', 'SEM']\n",
    "semcode.columns = new_column_names\n",
    "dnfilercountsortsmall2 = pd.merge(dnfilercountsortsmall, semcode, on='SEM', how='inner')\n",
    "print(len(dnfilercountsortsmall.SEM))\n",
    "dnfilercountsortsmall2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "w = dnfilercountsortsmall2.Labes_count[dnfilercountsortsmall2['Labes_count'] >1]\n",
    "u = dnfilercountsortsmall2.Label[dnfilercountsortsmall2['Labes_count'] >1]\n",
    "\n",
    "#plt.figure(figsize=(15, 12))\n",
    "plt.figure(figsize=(12, 6))\n",
    "plt.scatter(u,w)\n",
    "\n",
    "plt.title('Scatter Plot: \"nutritional physiology\" reference mesh plot')\n",
    "plt.xticks(rotation=90)\n",
    "\n",
    "plt.show()\n",
    "print(len(dnfilercountsortsmall2.Label[dnfilercountsortsmall2['Labes_count'] >1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dnfilercountsortsmall2[dnfilercountsortsmall2['Labes_count'] >1].head(8)"
   ]
  }
 ],
 "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": "3.11.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}