{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "negative-stranger",
   "metadata": {},
   "source": [
    "# Use hormone2cell to predict hormone producing/receiving cell types."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "north-account",
   "metadata": {},
   "source": [
    "This notebook showcases how to predict hormone producing cell types using hormone2cell."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "collectible-crest",
   "metadata": {},
   "source": [
    "**Steps**:\n",
    "- 0. **Download and process the dataset**\n",
    "     \n",
    "\n",
    "- 1. **Calculate average expression per cell type per tissue**\n",
    "     - Calculate the average gene expression for each cell type within each tissue.\n",
    "- 2. **Predict hormone producing strength**\n",
    "  - 2.1 Identify hormone-producing cell types using data-derived thresholds.\n",
    "  - 2.2 Identify hormone-producing cell types using precomputed thresholds from the Hormone Cell Atlas.\n",
    "  - Notes:\n",
    "    - The precomputed thresholds are derived from the pan-tissue Hormone Cell Atlas, while the data-derived thresholds are calculated directly from the user-provided dataset. Depending on your analytical goals, you may choose either 2.1 (data-derived) or 2.2 (precomputed).\n",
    "    - Without adjustment: Predictions of hormone-producing cell types and their strengths are made directly using the default function `hormone_strength`, without applying any adjustment to gene expression thresholds.\n",
    "    - With adjustment: Predictions of hormone-producing cell types and their strengths are performed using the `hormone_strength` function with `adjustment =  True`. Here, adjustment refers to modifying thresholds for hormone-related genes based on the overall specificity score and the degree of tissue coverage.\n",
    "    - When both single-cell and single-nucleus data are available, predictions should be performed separately for each dataset, because expression levels can differ substantially due to gene length (Lake et al., 2017: https://www.nature.com/articles/s41598-017-04426-w\n",
    "). The results can then be integrated by taking the maximum strength observed across the two assays.    \n",
    "- 3. **Predict hormone receiving strength**\n",
    "  - 3.1 Identify hormone-receiving cell types using data-derived thresholds.\n",
    "  - 3.2 Identify hormone-receiving cell types using precomputed thresholds from the Hormone Cell Atlas.\n",
    "  - Notes:\n",
    "     - It is recommended not to adjust the thresholds for hormone-receiving-related genes, since hormone sensitization is generally broader than hormone production.\n",
    "    \n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "starting-hollywood",
   "metadata": {},
   "outputs": [],
   "source": [
    "import scanpy as sc\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import time\n",
    "from collections import Counter\n",
    "import hormone2cell\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib_venn import venn2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "proper-firmware",
   "metadata": {},
   "outputs": [],
   "source": [
    "sc.settings.verbosity = 3  # verbosity: errors (0), warnings (1), info (2), hints (3)\n",
    "sc.set_figure_params(dpi=80, facecolor=\"white\")\n",
    "#sc.logging.print_header()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "destroyed-confusion",
   "metadata": {},
   "source": [
    "## 0. Download and process the dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "existing-issue",
   "metadata": {},
   "source": [
    "We used the pancreas as the query dataset, which includes both single-cell and single-nucleus data. The dataset was downloaded from the CellxGene website (https://datasets.cellxgene.cziscience.com/d7191e85-30c6-41ba-a439-e1d1fe5a3227.h5ad). For demonstration purposes, we sampled 200 cells per cell type per suspension type.\n",
    "\n",
    "If you would like to reproduce the hormone-producing and hormone-receiving results from our manuscript, please download the per-tissue h5ad files from https://hormonecellatlas.org.uk/downloads/\n",
    " and run the analysis using the precomputed thresholds provided by the Hormone Cell Atlas."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "august-equality",
   "metadata": {},
   "outputs": [],
   "source": [
    "# adata=sc.read('data/pancreas_downsample200CT.h5ad')\n",
    "adata=hormone2cell.data.load_pancreas_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "appropriate-dispute",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3274, 33053)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "adata.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "derived-infrared",
   "metadata": {},
   "source": [
    "Raw count data with the full gene set are required for hormone2cell."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "postal-appointment",
   "metadata": {},
   "outputs": [],
   "source": [
    "adata=adata.raw.to_adata()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "duplicate-ballet",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "np.float32(450742.0)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "adata.X.data.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "yellow-snowboard",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "constitutional-audio",
   "metadata": {},
   "source": [
    "Check and modify the obs to meet the requirements of hormone2cell. The `.obs` must contain columns for the tissue, cell type and suspension type (i.e. cell or nucleus) for downstream analysis."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "joint-curtis",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "suspension_type\n",
       "nucleus    1674\n",
       "cell       1600\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "adata.obs['suspension_type'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "selected-there",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "cell_type\n",
       "endothelial cell            400\n",
       "pancreatic A cell           400\n",
       "pancreatic D cell           400\n",
       "pancreatic ductal cell      400\n",
       "pancreatic acinar cell      400\n",
       "pancreatic stellate cell    400\n",
       "type B pancreatic cell      400\n",
       "pancreatic PP cell          274\n",
       "macrophage                  200\n",
       "Name: count, dtype: int64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "adata.obs['cell_type'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "portable-headset",
   "metadata": {},
   "outputs": [],
   "source": [
    "adata.obs['Tissue']='Pancreas'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "monetary-story",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "spanish-muslim",
   "metadata": {},
   "source": [
    "Ensure that `.var_names` are gene symbols, this particular object comes with them set to ENSG IDs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "featured-pennsylvania",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>exist_in_Tritschler2022</th>\n",
       "      <th>exist_in_Fasolino2022</th>\n",
       "      <th>exist_in_Muraro2016</th>\n",
       "      <th>exist_in_Tosti2021</th>\n",
       "      <th>feature_reference</th>\n",
       "      <th>feature_biotype</th>\n",
       "      <th>feature_length</th>\n",
       "      <th>feature_type</th>\n",
       "      <th>gene_id</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>feature_name</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>A1BG</th>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>NCBITaxon:9606</td>\n",
       "      <td>gene</td>\n",
       "      <td>2134</td>\n",
       "      <td>protein_coding</td>\n",
       "      <td>ENSG00000121410</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>A1BG-AS1</th>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>NCBITaxon:9606</td>\n",
       "      <td>gene</td>\n",
       "      <td>1667</td>\n",
       "      <td>lncRNA</td>\n",
       "      <td>ENSG00000268895</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              exist_in_Tritschler2022  exist_in_Fasolino2022  \\\n",
       "feature_name                                                   \n",
       "A1BG                             True                   True   \n",
       "A1BG-AS1                         True                   True   \n",
       "\n",
       "              exist_in_Muraro2016  exist_in_Tosti2021 feature_reference  \\\n",
       "feature_name                                                              \n",
       "A1BG                         True                True    NCBITaxon:9606   \n",
       "A1BG-AS1                     True                True    NCBITaxon:9606   \n",
       "\n",
       "             feature_biotype feature_length    feature_type          gene_id  \n",
       "feature_name                                                                  \n",
       "A1BG                    gene           2134  protein_coding  ENSG00000121410  \n",
       "A1BG-AS1                gene           1667          lncRNA  ENSG00000268895  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "adata.var['gene_id']=adata.var.index\n",
    "adata.var['feature_name']=adata.var['feature_name'].astype('str')\n",
    "adata.var = adata.var.set_index('feature_name')\n",
    "adata.var_names_make_unique()\n",
    "adata.var.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "excellent-basics",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "aerial-stretch",
   "metadata": {},
   "source": [
    "## 1. Calculate average expression per cell type per tissue\n",
    "\n",
    "Along with the actual average expression, which is computed after internally running `sc.pp.normalize_total()` with a target sum of 10,000 and log1p-transformed prior to being returned, the resulting data frame carries additional information on the gene's presence in any given tissue/celltype/suspension combination which is used in downstream analysis."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "dying-commodity",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "363 hormone-related genes are found.\n",
      "normalizing counts per cell\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/rfs/project/rfs-iCNyzSAaucw/lf529/miniforge3/envs/hormone2cell_env2/lib/python3.10/site-packages/scanpy/preprocessing/_normalization.py:269: UserWarning: Received a view of an AnnData. Making a copy.\n",
      "  view_to_actual(adata)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    finished (0:00:02)\n",
      "Computing average gene expression per cell type in the cell dataset.\n",
      "filtered out 130 genes that are detected in less than 3 cells\n",
      "Tissues in input: ['Pancreas'] — average calculation done.\n",
      "Computing average gene expression per cell type in the nucleus dataset.\n",
      "filtered out 125 genes that are detected in less than 3 cells\n",
      "Tissues in input: ['Pancreas'] — average calculation done.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/rfs/project/rfs-iCNyzSAaucw/lf529/miniforge3/envs/hormone2cell_env2/lib/python3.10/site-packages/legacy_api_wrap/__init__.py:88: UserWarning: Some cells have zero counts\n",
      "  return fn(*args_all, **kw)\n"
     ]
    }
   ],
   "source": [
    "ave_all=hormone2cell.compute_aveExp_by_category(adata,sc_sn_col='suspension_type',\n",
    "                                                celltype_col= \"cell_type\",\n",
    "                                               tissue_col='Tissue')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "equipped-default",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Tissue</th>\n",
       "      <th>cell_type</th>\n",
       "      <th>Cluster</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Gene</th>\n",
       "      <th>ExpressedCellNumber</th>\n",
       "      <th>TotalCellNumber</th>\n",
       "      <th>Percentage</th>\n",
       "      <th>logExpression</th>\n",
       "      <th>suspension_type</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Pancreas</td>\n",
       "      <td>pancreatic A cell</td>\n",
       "      <td>2</td>\n",
       "      <td>Pancreas____pancreatic A cell</td>\n",
       "      <td>IGF2</td>\n",
       "      <td>13</td>\n",
       "      <td>200</td>\n",
       "      <td>6.5</td>\n",
       "      <td>0.732021</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Pancreas</td>\n",
       "      <td>pancreatic D cell</td>\n",
       "      <td>3</td>\n",
       "      <td>Pancreas____pancreatic D cell</td>\n",
       "      <td>IGF2</td>\n",
       "      <td>24</td>\n",
       "      <td>200</td>\n",
       "      <td>12.0</td>\n",
       "      <td>0.668118</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Tissue          cell_type Cluster                Celltype_tissue  Gene  \\\n",
       "1  Pancreas  pancreatic A cell       2  Pancreas____pancreatic A cell  IGF2   \n",
       "2  Pancreas  pancreatic D cell       3  Pancreas____pancreatic D cell  IGF2   \n",
       "\n",
       "   ExpressedCellNumber  TotalCellNumber  Percentage  logExpression  \\\n",
       "1                   13              200         6.5       0.732021   \n",
       "2                   24              200        12.0       0.668118   \n",
       "\n",
       "  suspension_type  \n",
       "1            cell  \n",
       "2            cell  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ave_all.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "architectural-purse",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2822, 10)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ave_all.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "blessed-partition",
   "metadata": {},
   "source": [
    "## 2. Predict hormone producing strength"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "experienced-species",
   "metadata": {},
   "source": [
    "Load the definitions of hormone production, as specified in the hormone–receptor reference database of the Hormone Cell Atlas project.\n",
    "\n",
    "In this dataframe, `hormogene_include` columns represent genes involved in the induction of hormone production, while `hormogene_exclude` columns represent genes involved in its repression. The columns represent an `AND` relationship, i.e. all of the genes in a row need to fulfil subsequent filtering conditions for the hormone to qualify as produced or repressed in a cell type. `OR` relationships, indicative of multiple independent production pathways, are represented by multiple entries for the hormone in the data frame.\n",
    "\n",
    "\n",
    "Genes excluded from strength calculations but required to meet expression thresholds are hard-coded in utils.py (see Methods in Fei & Huang-Doran (in press)) and may you need to adjust these definitions for your own study."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "equal-mozambique",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_figures</th>\n",
       "      <th>Tier</th>\n",
       "      <th>hpc_include1</th>\n",
       "      <th>hpc_include2</th>\n",
       "      <th>hpc_include3</th>\n",
       "      <th>hpc_include4</th>\n",
       "      <th>hpc_include5</th>\n",
       "      <th>hpc_include6</th>\n",
       "      <th>hpc_include7</th>\n",
       "      <th>hpc_include8</th>\n",
       "      <th>hpc_exclude1</th>\n",
       "      <th>hpc_exclude2</th>\n",
       "      <th>receptor_known</th>\n",
       "      <th>hormone_type_broad</th>\n",
       "      <th>hormone_type_fine</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>acth</td>\n",
       "      <td>ACTH</td>\n",
       "      <td>ACTH</td>\n",
       "      <td>1</td>\n",
       "      <td>POMC</td>\n",
       "      <td>PCSK1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>PCSK2</td>\n",
       "      <td>NaN</td>\n",
       "      <td>yes</td>\n",
       "      <td>peptide_protein</td>\n",
       "      <td>peptide</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>activin_a</td>\n",
       "      <td>Activin A</td>\n",
       "      <td>Activin A</td>\n",
       "      <td>2</td>\n",
       "      <td>INHBA</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>yes</td>\n",
       "      <td>peptide_protein</td>\n",
       "      <td>protein_multimer</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  hormone_short hormone_display hormone_figures Tier hpc_include1  \\\n",
       "0          acth            ACTH            ACTH    1         POMC   \n",
       "1     activin_a       Activin A       Activin A    2        INHBA   \n",
       "\n",
       "  hpc_include2 hpc_include3 hpc_include4 hpc_include5 hpc_include6  \\\n",
       "0        PCSK1          NaN          NaN          NaN          NaN   \n",
       "1          NaN          NaN          NaN          NaN          NaN   \n",
       "\n",
       "  hpc_include7 hpc_include8 hpc_exclude1 hpc_exclude2 receptor_known  \\\n",
       "0          NaN          NaN        PCSK2          NaN            yes   \n",
       "1          NaN          NaN          NaN          NaN            yes   \n",
       "\n",
       "  hormone_type_broad hormone_type_fine  \n",
       "0    peptide_protein           peptide  \n",
       "1    peptide_protein  protein_multimer  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hormone_definition=hormone2cell.data.load_hormone_producing_file()\n",
    "hormone_definition.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "coral-above",
   "metadata": {},
   "source": [
    "### 2.1 Identify hormone-producing cell types using data-derived thresholds.\n",
    "\n",
    "At this point, we're ready to perform data-driven hormone2cell inference. The function will:\n",
    "- Filter the average expression data frame, removing entries with fewer than 10 cells in a category, fewer than 3 cells expressing the gene, or where the log expression is too low (`thresh_expr_low`) or has too low a percentage of expressing cells (`thresh_pct`).\n",
    "- For each gene, identify its highest observed expression in a cell type across the whole data.\n",
    "- For genes inducing hormone production (`include_cols` from the loaded hormone definition data frame), require at least `thresh_included_initial` of the max expression to be kept in the expression matrix. For values below that threshold, the expression will be set to 0.\n",
    "- For genes repressing hormone production (`exclude_cols`), require at least `thresh_excluded_initial` of the max expression to be kept in the expression matrix. For values below that threshold, the expression will be set to 0.\n",
    "- For a hormone to be deemed present in a cell type, any of its independent induction mechanisms (individual lines in the reference data frame, OR logic) must be fulfilled. An induction mechanism is deemed fulfilled if all of its induction genes have nonzero expression after the filtering (AND logic). Any induction gene having zero expression, or all repression genes having nonzero expression, has the induction mechanism set to not fulfilled.\n",
    "- Hormone activity is screened for specificity. If hormone production is present in at least `coverage_threshold` of the cell types of a tissue, or a Tau specificty score below `specificity_threshold` (with the input computed as the mean expression of a hormone's AND genes, summed up across its OR induction mechanisms, for instances of hormone production being considered active in a cell type; 0 otherwise), the hormone is deemed to be non-specific. Causes of this may include ambient RNA contamination.\n",
    "- For non-specific hormones, the expression filtering thresholds are adjusted to `thresh_included_adjusted` and `thresh_excluded_adjusted` and the analysis is repeated."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "db5cf1ea-06d2-438b-b0af-59f96efa893e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Tissue</th>\n",
       "      <th>cell_type</th>\n",
       "      <th>Cluster</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Gene</th>\n",
       "      <th>ExpressedCellNumber</th>\n",
       "      <th>TotalCellNumber</th>\n",
       "      <th>Percentage</th>\n",
       "      <th>logExpression</th>\n",
       "      <th>suspension_type</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Pancreas</td>\n",
       "      <td>pancreatic A cell</td>\n",
       "      <td>2</td>\n",
       "      <td>Pancreas____pancreatic A cell</td>\n",
       "      <td>IGF2</td>\n",
       "      <td>13</td>\n",
       "      <td>200</td>\n",
       "      <td>6.5</td>\n",
       "      <td>0.732021</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Pancreas</td>\n",
       "      <td>pancreatic D cell</td>\n",
       "      <td>3</td>\n",
       "      <td>Pancreas____pancreatic D cell</td>\n",
       "      <td>IGF2</td>\n",
       "      <td>24</td>\n",
       "      <td>200</td>\n",
       "      <td>12.0</td>\n",
       "      <td>0.668118</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Pancreas</td>\n",
       "      <td>pancreatic PP cell</td>\n",
       "      <td>4</td>\n",
       "      <td>Pancreas____pancreatic PP cell</td>\n",
       "      <td>IGF2</td>\n",
       "      <td>4</td>\n",
       "      <td>200</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.246472</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Pancreas</td>\n",
       "      <td>pancreatic ductal cell</td>\n",
       "      <td>6</td>\n",
       "      <td>Pancreas____pancreatic ductal cell</td>\n",
       "      <td>IGF2</td>\n",
       "      <td>2</td>\n",
       "      <td>200</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.092293</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Pancreas</td>\n",
       "      <td>type B pancreatic cell</td>\n",
       "      <td>8</td>\n",
       "      <td>Pancreas____type B pancreatic cell</td>\n",
       "      <td>IGF2</td>\n",
       "      <td>114</td>\n",
       "      <td>200</td>\n",
       "      <td>57.0</td>\n",
       "      <td>1.943938</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     Tissue               cell_type Cluster  \\\n",
       "1  Pancreas       pancreatic A cell       2   \n",
       "2  Pancreas       pancreatic D cell       3   \n",
       "3  Pancreas      pancreatic PP cell       4   \n",
       "5  Pancreas  pancreatic ductal cell       6   \n",
       "7  Pancreas  type B pancreatic cell       8   \n",
       "\n",
       "                      Celltype_tissue  Gene  ExpressedCellNumber  \\\n",
       "1       Pancreas____pancreatic A cell  IGF2                   13   \n",
       "2       Pancreas____pancreatic D cell  IGF2                   24   \n",
       "3      Pancreas____pancreatic PP cell  IGF2                    4   \n",
       "5  Pancreas____pancreatic ductal cell  IGF2                    2   \n",
       "7  Pancreas____type B pancreatic cell  IGF2                  114   \n",
       "\n",
       "   TotalCellNumber  Percentage  logExpression suspension_type  \n",
       "1              200         6.5       0.732021            cell  \n",
       "2              200        12.0       0.668118            cell  \n",
       "3              200         2.0       0.246472            cell  \n",
       "5              200         1.0       0.092293            cell  \n",
       "7              200        57.0       1.943938            cell  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ave_all.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "wrapped-kennedy",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "I. Compute initial strengths\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from current data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from current data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "II. Identify hormones to adjust and recompute only the adjusted hormones\n",
      "   1). Identify hormones to adjust based on the specificity\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from current data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from current data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "   2). Identify hormones to adjust based on the tissue coverage\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from current data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from current data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "III. Merge: replace original rows with adjusted hormone rows\n",
      "(512, 4)\n",
      "    Hormones that have been adjusted includes: dict_keys(['activin_a', 'activins_all', 'adm', 'adm_all', 'angiotensinogen', 'chemerin', 'endothelin_3', 'endothelins_all', 'gdf15', 'glp1', 'glp2', 'npy', 'pacap', 'pomc', 'pp', 'pro_amylin_iapp', 'pro_ghrelin', 'pro_glucagon', 'pro_insulin', 'pro_npy', 'pro_somatostatin', 'rbp4', 'salusin', 'somatostatin', 'visfatin'])\n",
      "------hormone_long_adj_coverage-------\n",
      "        Hormone               Celltype_tissue  Strength    Tissue\n",
      "0     activin_a  Pancreas____endothelial cell  0.000000  Pancreas\n",
      "1  activins_all  Pancreas____endothelial cell  3.586259  Pancreas\n",
      "------pairs-------\n",
      "        Hormone    Tissue\n",
      "0     activin_a  Pancreas\n",
      "1  activins_all  Pancreas\n",
      "(512, 4)\n",
      "Before removing duplicates: (230, 7)\n",
      "After removing duplicates: (173, 7)\n",
      "Finally hormone2cell finds: 173 hormone-celltype pairs.\n",
      "CPU times: user 370 ms, sys: 3.61 ms, total: 374 ms\n",
      "Wall time: 383 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(173, 8)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "mask=ave_all['suspension_type']=='cell'\n",
    "ave_all_cell=ave_all.loc[mask]\n",
    "dt_cell=hormone2cell.hormone_strength(\n",
    "    ave_all = ave_all_cell,\n",
    "    geneset_definition = hormone_definition,\n",
    "    celltype_column = \"Celltype_tissue\", # must be unique\n",
    "    tissue_col = \"Tissue\",\n",
    "    adjustment =  True,\n",
    "    assay = \"cell\",\n",
    "    include_cols = ['hpc_include1', 'hpc_include2', 'hpc_include3',\n",
    "       'hpc_include4', 'hpc_include5', 'hpc_include6', 'hpc_include7',\n",
    "       'hpc_include8'],\n",
    "    exclude_cols =['hpc_exclude1', 'hpc_exclude2'],\n",
    "    # initial thresholds\n",
    "    thresh_expr_low = 0.01,\n",
    "    thresh_pct = 5,\n",
    "    # specificity / coverage thresholds\n",
    "    specificity_threshold = 0.8,\n",
    "    coverage_threshold = 0.5,\n",
    "    # filtering thresholds (first pass)\n",
    "    thresh_included_initial = 0.1,\n",
    "    thresh_excluded_initial = 0.15,\n",
    "    # filtering thresholds (adjusted pass)\n",
    "    thresh_included_adjusted = 0.5,\n",
    "    thresh_excluded_adjusted = 0.15,\n",
    "    max_expression_file='max_expression_cell.csv'\n",
    ")\n",
    "dt_cell.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2b38e210-b2bf-4c4b-a53b-73cdc585a01b",
   "metadata": {},
   "source": [
    "The resulted `dt_cell` contains the hormones expressed in each cell type and their expression strengths, along with the corresponding annotations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "ed247272-ab8b-4502-b223-edc73d01bf81",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Hormone</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Strength</th>\n",
       "      <th>Tissue</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>Tier</th>\n",
       "      <th>assay</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>glp1</td>\n",
       "      <td>Pancreas____pancreatic stellate cell</td>\n",
       "      <td>7.070757</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Glucagon-like peptide 1</td>\n",
       "      <td>glp1</td>\n",
       "      <td>2</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>glp1</td>\n",
       "      <td>Pancreas____type B pancreatic cell</td>\n",
       "      <td>5.734788</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Glucagon-like peptide 1</td>\n",
       "      <td>glp1</td>\n",
       "      <td>2</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Hormone                       Celltype_tissue  Strength    Tissue  \\\n",
       "0    glp1  Pancreas____pancreatic stellate cell  7.070757  Pancreas   \n",
       "1    glp1    Pancreas____type B pancreatic cell  5.734788  Pancreas   \n",
       "\n",
       "           hormone_display hormone_short Tier assay  \n",
       "0  Glucagon-like peptide 1          glp1    2  cell  \n",
       "1  Glucagon-like peptide 1          glp1    2  cell  "
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dt_cell.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "unlikely-authority",
   "metadata": {},
   "source": [
    "We can perform the analysis on nuclear rather than cellular data as well."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "substantial-discussion",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "I. Compute initial strengths\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from current data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from current data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "II. Identify hormones to adjust and recompute only the adjusted hormones\n",
      "   1). Identify hormones to adjust based on the specificity\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from current data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from current data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "   2). Identify hormones to adjust based on the tissue coverage\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from current data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from current data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "III. Merge: replace original rows with adjusted hormone rows\n",
      "(324, 4)\n",
      "    Hormones that have been adjusted includes: dict_keys(['chemerin', 'visfatin'])\n",
      "------hormone_long_adj_coverage-------\n",
      "    Hormone               Celltype_tissue  Strength    Tissue\n",
      "0  chemerin  Pancreas____endothelial cell  4.995571  Pancreas\n",
      "1  visfatin  Pancreas____endothelial cell  5.121772  Pancreas\n",
      "------pairs-------\n",
      "    Hormone    Tissue\n",
      "0  chemerin  Pancreas\n",
      "1  visfatin  Pancreas\n",
      "(320, 4)\n",
      "Before removing duplicates: (94, 7)\n",
      "After removing duplicates: (76, 7)\n",
      "Finally hormone2cell finds: 76 hormone-celltype pairs.\n"
     ]
    }
   ],
   "source": [
    "mask=ave_all['suspension_type']=='nucleus'\n",
    "ave_all_nucleus=ave_all.loc[mask]\n",
    "dt_nucleus=hormone2cell.hormone_strength(\n",
    "    ave_all = ave_all_nucleus,\n",
    "    geneset_definition = hormone_definition,\n",
    "    celltype_column = \"Celltype_tissue\",  # must be unique\n",
    "    tissue_col = \"Tissue\",\n",
    "    adjustment =  True,\n",
    "    assay = \"nucleus\",\n",
    "    include_cols = ['hpc_include1', 'hpc_include2', 'hpc_include3',\n",
    "       'hpc_include4', 'hpc_include5', 'hpc_include6', 'hpc_include7',\n",
    "       'hpc_include8'],\n",
    "    exclude_cols =['hpc_exclude1', 'hpc_exclude2'],\n",
    "    # initial thresholds\n",
    "    thresh_expr_low = 0.01,\n",
    "    thresh_pct = 5,\n",
    "    # specificity / coverage thresholds\n",
    "    specificity_threshold = 0.8,\n",
    "    coverage_threshold = 0.5,\n",
    "    # filtering thresholds (first pass)\n",
    "    thresh_included_initial = 0.1,\n",
    "    thresh_excluded_initial = 0.15,\n",
    "    # filtering thresholds (adjusted pass)\n",
    "    thresh_included_adjusted = 0.5,\n",
    "    thresh_excluded_adjusted = 0.15,\n",
    "    max_expression_file='max_expression_nucleus.csv'\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "d4882e30-c806-43db-9c96-b8f83362dc7f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Hormone</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Strength</th>\n",
       "      <th>Tissue</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>Tier</th>\n",
       "      <th>assay</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>crh</td>\n",
       "      <td>Pancreas____pancreatic D cell</td>\n",
       "      <td>2.320317</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>CRH</td>\n",
       "      <td>crh</td>\n",
       "      <td>1</td>\n",
       "      <td>nucleus</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>somatostatin</td>\n",
       "      <td>Pancreas____pancreatic D cell</td>\n",
       "      <td>8.120015</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Somatostatin</td>\n",
       "      <td>somatostatin</td>\n",
       "      <td>2</td>\n",
       "      <td>nucleus</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        Hormone                Celltype_tissue  Strength    Tissue  \\\n",
       "0           crh  Pancreas____pancreatic D cell  2.320317  Pancreas   \n",
       "2  somatostatin  Pancreas____pancreatic D cell  8.120015  Pancreas   \n",
       "\n",
       "  hormone_display hormone_short Tier    assay  \n",
       "0             CRH           crh    1  nucleus  \n",
       "2    Somatostatin  somatostatin    2  nucleus  "
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dt_nucleus.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "infrared-genome",
   "metadata": {},
   "source": [
    "#### 2.1.1 Combine results from single cell and single nucleus data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "modular-winner",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Hormone</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Strength</th>\n",
       "      <th>Tissue</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>Tier</th>\n",
       "      <th>assay</th>\n",
       "      <th>Hormone_CT</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>glp1</td>\n",
       "      <td>Pancreas____pancreatic stellate cell</td>\n",
       "      <td>7.070757</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Glucagon-like peptide 1</td>\n",
       "      <td>glp1</td>\n",
       "      <td>2</td>\n",
       "      <td>cell_only</td>\n",
       "      <td>glp1___Pancreas____pancreatic stellate cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>glp1</td>\n",
       "      <td>Pancreas____pancreatic PP cell</td>\n",
       "      <td>4.827841</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Glucagon-like peptide 1</td>\n",
       "      <td>glp1</td>\n",
       "      <td>2</td>\n",
       "      <td>cell_only</td>\n",
       "      <td>glp1___Pancreas____pancreatic PP cell</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Hormone                       Celltype_tissue  Strength    Tissue  \\\n",
       "0    glp1  Pancreas____pancreatic stellate cell  7.070757  Pancreas   \n",
       "1    glp1        Pancreas____pancreatic PP cell  4.827841  Pancreas   \n",
       "\n",
       "           hormone_display hormone_short Tier      assay  \\\n",
       "0  Glucagon-like peptide 1          glp1    2  cell_only   \n",
       "1  Glucagon-like peptide 1          glp1    2  cell_only   \n",
       "\n",
       "                                    Hormone_CT  \n",
       "0  glp1___Pancreas____pancreatic stellate cell  \n",
       "1        glp1___Pancreas____pancreatic PP cell  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "(188, 9)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hpc_datadriven=hormone2cell.combine_assay(cell=dt_cell, nucleus=dt_nucleus,celltype_column='Celltype_tissue')\n",
    "display(hpc_datadriven.head(2))\n",
    "hpc_datadriven.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "macro-drink",
   "metadata": {},
   "source": [
    "### 2.2 Identify hormone-producing cell types using precomputed thresholds from the Hormone Cell Atlas.\n",
    "A key strength of the accompanying manuscript is its analysis of a large integrated dataset spanning 47 tissues and 14 million cells. The large scale integration allowed for defining maximum values in a manner more representative of what is likely to be occurring at the level of the entire body, which should lead to more informed thresholding."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "velvet-congo",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "I. Compute initial strengths\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "II. Identify hormones to adjust and recompute only the adjusted hormones\n",
      "   1). Identify hormones to adjust based on the specificity\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "   2). Identify hormones to adjust based on the tissue coverage\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "III. Merge: replace original rows with adjusted hormone rows\n",
      "(512, 4)\n",
      "    Hormones that have been adjusted includes: dict_keys(['activin_a', 'activins_all', 'adm', 'adm_all', 'angiotensinogen', 'chemerin', 'endothelin_3', 'endothelins_all', 'follistatin_like_1', 'gdf15', 'glp1', 'glp2', 'npy', 'pacap', 'pomc', 'pp', 'pro_amylin_iapp', 'pro_ghrelin', 'pro_glucagon', 'pro_insulin', 'pro_npy', 'pro_somatostatin', 'rbp4', 'salusin', 'somatostatin', 'visfatin'])\n",
      "------hormone_long_adj_coverage-------\n",
      "        Hormone               Celltype_tissue  Strength    Tissue\n",
      "0     activin_a  Pancreas____endothelial cell  1.969141  Pancreas\n",
      "1  activins_all  Pancreas____endothelial cell  5.555400  Pancreas\n",
      "------pairs-------\n",
      "        Hormone    Tissue\n",
      "0     activin_a  Pancreas\n",
      "1  activins_all  Pancreas\n",
      "(512, 4)\n",
      "Before removing duplicates: (257, 7)\n",
      "After removing duplicates: (194, 7)\n",
      "Finally hormone2cell finds: 194 hormone-celltype pairs.\n",
      "CPU times: user 377 ms, sys: 2.01 ms, total: 379 ms\n",
      "Wall time: 424 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(194, 8)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "mask=ave_all['suspension_type']=='cell'\n",
    "ave_all_cell=ave_all.loc[mask]\n",
    "dt_cell=hormone2cell.hormone_strength(\n",
    "    ave_all = ave_all_cell,\n",
    "    geneset_definition = hormone_definition,\n",
    "    celltype_column = \"Celltype_tissue\",  # must be unique\n",
    "    tissue_col = \"Tissue\",\n",
    "    adjustment =  True,\n",
    "    assay = \"cell\",\n",
    "    include_cols = ['hpc_include1', 'hpc_include2', 'hpc_include3',\n",
    "       'hpc_include4', 'hpc_include5', 'hpc_include6', 'hpc_include7',\n",
    "       'hpc_include8'],\n",
    "    exclude_cols =['hpc_exclude1', 'hpc_exclude2'],\n",
    "    # initial thresholds\n",
    "    thresh_expr_low = 0.01,\n",
    "    thresh_pct = 5,\n",
    "    # specificity / coverage thresholds\n",
    "    specificity_threshold = 0.8,\n",
    "    coverage_threshold = 0.5,\n",
    "    # filtering thresholds (first pass)\n",
    "    thresh_included_initial = 0.1,\n",
    "    thresh_excluded_initial = 0.15,\n",
    "    # filtering thresholds (adjusted pass)\n",
    "    thresh_included_adjusted = 0.5,\n",
    "    thresh_excluded_adjusted = 0.15,\n",
    "    use_precomputed='cell',\n",
    "    max_expression_file='max_expression_cell_pre.csv'\n",
    ")\n",
    "dt_cell.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "supported-circuit",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "I. Compute initial strengths\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "II. Identify hormones to adjust and recompute only the adjusted hormones\n",
      "   1). Identify hormones to adjust based on the specificity\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "   2). Identify hormones to adjust based on the tissue coverage\n",
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "III. Merge: replace original rows with adjusted hormone rows\n",
      "(324, 4)\n",
      "    Hormones that have been adjusted includes: dict_keys(['chemerin', 'visfatin'])\n",
      "------hormone_long_adj_coverage-------\n",
      "    Hormone               Celltype_tissue  Strength    Tissue\n",
      "0  chemerin  Pancreas____endothelial cell  4.995571  Pancreas\n",
      "1  visfatin  Pancreas____endothelial cell  5.121772  Pancreas\n",
      "------pairs-------\n",
      "    Hormone    Tissue\n",
      "0  chemerin  Pancreas\n",
      "1  visfatin  Pancreas\n",
      "(320, 4)\n",
      "Before removing duplicates: (94, 7)\n",
      "After removing duplicates: (76, 7)\n",
      "Finally hormone2cell finds: 76 hormone-celltype pairs.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(76, 8)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mask=ave_all['suspension_type']=='nucleus'\n",
    "ave_all_nucleus=ave_all.loc[mask]\n",
    "dt_nucleus=hormone2cell.hormone_strength(\n",
    "    ave_all = ave_all_nucleus,\n",
    "    geneset_definition = hormone_definition,\n",
    "    celltype_column = \"Celltype_tissue\",  # must be unique\n",
    "    tissue_col = \"Tissue\",\n",
    "    adjustment =  True,\n",
    "    assay = \"nucleus\",\n",
    "    include_cols = ['hpc_include1', 'hpc_include2', 'hpc_include3',\n",
    "       'hpc_include4', 'hpc_include5', 'hpc_include6', 'hpc_include7',\n",
    "       'hpc_include8'],\n",
    "    exclude_cols =['hpc_exclude1', 'hpc_exclude2'],\n",
    "    # initial thresholds\n",
    "    thresh_expr_low = 0.01,\n",
    "    thresh_pct = 5,\n",
    "    # specificity / coverage thresholds\n",
    "    specificity_threshold = 0.8,\n",
    "    coverage_threshold = 0.5,\n",
    "    # filtering thresholds (first pass)\n",
    "    thresh_included_initial = 0.1,\n",
    "    thresh_excluded_initial = 0.15,\n",
    "    # filtering thresholds (adjusted pass)\n",
    "    thresh_included_adjusted = 0.5,\n",
    "    thresh_excluded_adjusted = 0.15,\n",
    "    use_precomputed='nucleus',\n",
    "    max_expression_file='max_expression_nucleus_pre.csv'\n",
    ")\n",
    "dt_nucleus.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "challenging-hayes",
   "metadata": {},
   "source": [
    "#### 2.2.1 Combine results from single cell and single nucleus data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "collaborative-lawrence",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Hormone</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Strength</th>\n",
       "      <th>Tissue</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>Tier</th>\n",
       "      <th>assay</th>\n",
       "      <th>Hormone_CT</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>glp1</td>\n",
       "      <td>Pancreas____pancreatic stellate cell</td>\n",
       "      <td>7.070757</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Glucagon-like peptide 1</td>\n",
       "      <td>glp1</td>\n",
       "      <td>2</td>\n",
       "      <td>cell_only</td>\n",
       "      <td>glp1___Pancreas____pancreatic stellate cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>glp1</td>\n",
       "      <td>Pancreas____pancreatic PP cell</td>\n",
       "      <td>4.827841</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Glucagon-like peptide 1</td>\n",
       "      <td>glp1</td>\n",
       "      <td>2</td>\n",
       "      <td>cell_only</td>\n",
       "      <td>glp1___Pancreas____pancreatic PP cell</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Hormone                       Celltype_tissue  Strength    Tissue  \\\n",
       "0    glp1  Pancreas____pancreatic stellate cell  7.070757  Pancreas   \n",
       "1    glp1        Pancreas____pancreatic PP cell  4.827841  Pancreas   \n",
       "\n",
       "           hormone_display hormone_short Tier      assay  \\\n",
       "0  Glucagon-like peptide 1          glp1    2  cell_only   \n",
       "1  Glucagon-like peptide 1          glp1    2  cell_only   \n",
       "\n",
       "                                    Hormone_CT  \n",
       "0  glp1___Pancreas____pancreatic stellate cell  \n",
       "1        glp1___Pancreas____pancreatic PP cell  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "(208, 9)"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hpc_precomputed=hormone2cell.combine_assay(cell=dt_cell, nucleus=dt_nucleus,celltype_column='Celltype_tissue')\n",
    "display(hpc_datadriven.head(2))\n",
    "hpc_precomputed.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "enhanced-incidence",
   "metadata": {},
   "source": [
    "Visulization of the hormone producing cell types from two methods."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "international-accident",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 480x480 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 386,
       "width": 389
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# import matplotlib.pyplot as plt\n",
    "# from matplotlib_venn import venn2\n",
    "\n",
    "set1 = set(hpc_datadriven['Hormone_CT'])\n",
    "set2 = set(hpc_precomputed['Hormone_CT'])\n",
    "\n",
    "plt.figure(figsize=(6,6))\n",
    "venn2([set1, set2], set_labels=(\"Data-driven\", \"Precomputated\"))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "dimensional-membrane",
   "metadata": {},
   "outputs": [],
   "source": [
    "## save the files if needed\n",
    "# hpc_datadriven.to_csv('hpc_datadriven.csv')\n",
    "# hpc_precomputed.to_csv('hpc_precomputed.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "wired-boutique",
   "metadata": {},
   "source": [
    "## 3. Predict hormone receiving strength"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "designed-blame",
   "metadata": {},
   "source": [
    "Load the hormone receiving definitions from the hormone-receptor reference database. The file is formatted similarly to the earlier one, the pertinent columns being `receptorgene` this time.\n",
    "\n",
    "Genes excluded from strength calculations but required to meet expression thresholds are hard-coded in utils.py (see Methods in Fei & Huang-Doran (in press)) and may you need to adjust these definitions for your own study."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "convenient-matthew",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_figures</th>\n",
       "      <th>Tier</th>\n",
       "      <th>receptorgene1</th>\n",
       "      <th>receptorgene2</th>\n",
       "      <th>gene1_name</th>\n",
       "      <th>gene2_name</th>\n",
       "      <th>receptor_type_broad</th>\n",
       "      <th>receptor_type_fine</th>\n",
       "      <th>receptor_status</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>asprosin</td>\n",
       "      <td>Asprosin</td>\n",
       "      <td>Asprosin</td>\n",
       "      <td>3</td>\n",
       "      <td>OR4M1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>olfactory receptor family 4 subfamily M member 1</td>\n",
       "      <td>NaN</td>\n",
       "      <td>gpcr</td>\n",
       "      <td>NaN</td>\n",
       "      <td>established</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>asprosin</td>\n",
       "      <td>Asprosin</td>\n",
       "      <td>Asprosin</td>\n",
       "      <td>3</td>\n",
       "      <td>PTPRD</td>\n",
       "      <td>NaN</td>\n",
       "      <td>protein tyrosine phosphatase receptor delta</td>\n",
       "      <td>NaN</td>\n",
       "      <td>enzyme_linked</td>\n",
       "      <td>NaN</td>\n",
       "      <td>established</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  hormone_short hormone_display hormone_figures Tier receptorgene1  \\\n",
       "0      asprosin        Asprosin        Asprosin    3         OR4M1   \n",
       "1      asprosin        Asprosin        Asprosin    3         PTPRD   \n",
       "\n",
       "  receptorgene2                                        gene1_name gene2_name  \\\n",
       "0           NaN  olfactory receptor family 4 subfamily M member 1        NaN   \n",
       "1           NaN       protein tyrosine phosphatase receptor delta        NaN   \n",
       "\n",
       "  receptor_type_broad receptor_type_fine receptor_status  \n",
       "0                gpcr                NaN     established  \n",
       "1       enzyme_linked                NaN     established  "
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "receptor_definition=hormone2cell.data.load_hormone_receptor_file()\n",
    "receptor_definition.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "distributed-greene",
   "metadata": {},
   "source": [
    "### 3.1 Identify hormone-receiving cell types using data-derived thresholds."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "waiting-generation",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from current data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from current data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "      Hormone                       Celltype_tissue   Strength    Tissue\n",
      "612  resistin  Pancreas____pancreatic stellate cell  14.868078  Pancreas\n",
      "52       igf2          Pancreas____endothelial cell  12.381264  Pancreas\n",
      "Before removing duplicates: (837, 7)\n",
      "After removing duplicates: (311, 7)\n",
      "Finally hormone2cell finds: 311 hormone-celltype pairs.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Hormone</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Strength</th>\n",
       "      <th>Tissue</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>Tier</th>\n",
       "      <th>assay</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>gastrin</td>\n",
       "      <td>Pancreas____pancreatic D cell</td>\n",
       "      <td>0.422587</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Gastrin</td>\n",
       "      <td>gastrin</td>\n",
       "      <td>2</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>gastrin</td>\n",
       "      <td>Pancreas____pancreatic PP cell</td>\n",
       "      <td>0.303184</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Gastrin</td>\n",
       "      <td>gastrin</td>\n",
       "      <td>2</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Hormone                 Celltype_tissue  Strength    Tissue  \\\n",
       "0  gastrin   Pancreas____pancreatic D cell  0.422587  Pancreas   \n",
       "1  gastrin  Pancreas____pancreatic PP cell  0.303184  Pancreas   \n",
       "\n",
       "  hormone_display hormone_short Tier assay  \n",
       "0         Gastrin       gastrin    2  cell  \n",
       "1         Gastrin       gastrin    2  cell  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 389 ms, sys: 3.74 ms, total: 392 ms\n",
      "Wall time: 414 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(311, 8)"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "mask=ave_all['suspension_type']=='cell'\n",
    "ave_all_cell=ave_all.loc[mask]\n",
    "dt_cell=hormone2cell.hormone_strength(\n",
    "    ave_all = ave_all_cell,\n",
    "    geneset_definition = receptor_definition,\n",
    "    celltype_column = \"Celltype_tissue\",\n",
    "    tissue_col = \"Tissue\",\n",
    "    adjustment =  False,\n",
    "    assay = \"cell\",\n",
    "    include_cols = ['receptorgene1', 'receptorgene2'],\n",
    "    exclude_cols =[],\n",
    "    # initial thresholds\n",
    "    thresh_expr_low = 0.01,\n",
    "    thresh_pct = 5,\n",
    "    # filtering thresholds (first pass)\n",
    "    thresh_included_initial = 0.1,\n",
    "    thresh_excluded_initial = 0.15,\n",
    "    max_expression_file='max_expression_cell_receptor.csv'\n",
    ")\n",
    "display(dt_cell.head(2))\n",
    "dt_cell.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "harmful-words",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from current data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from current data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "    Hormone                     Celltype_tissue   Strength    Tissue\n",
      "481    igf2  Pancreas____pancreatic acinar cell  18.891066  Pancreas\n",
      "568    igf2  Pancreas____pancreatic ductal cell  16.461430  Pancreas\n",
      "Before removing duplicates: (744, 7)\n",
      "After removing duplicates: (291, 7)\n",
      "Finally hormone2cell finds: 291 hormone-celltype pairs.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Hormone</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Strength</th>\n",
       "      <th>Tissue</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>Tier</th>\n",
       "      <th>assay</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>gastrin</td>\n",
       "      <td>Pancreas____pancreatic D cell</td>\n",
       "      <td>0.422587</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Gastrin</td>\n",
       "      <td>gastrin</td>\n",
       "      <td>2</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>gastrin</td>\n",
       "      <td>Pancreas____pancreatic PP cell</td>\n",
       "      <td>0.303184</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Gastrin</td>\n",
       "      <td>gastrin</td>\n",
       "      <td>2</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Hormone                 Celltype_tissue  Strength    Tissue  \\\n",
       "0  gastrin   Pancreas____pancreatic D cell  0.422587  Pancreas   \n",
       "1  gastrin  Pancreas____pancreatic PP cell  0.303184  Pancreas   \n",
       "\n",
       "  hormone_display hormone_short Tier assay  \n",
       "0         Gastrin       gastrin    2  cell  \n",
       "1         Gastrin       gastrin    2  cell  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "(291, 8)"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mask=ave_all['suspension_type']=='nucleus'\n",
    "ave_all_nucleus=ave_all.loc[mask]\n",
    "dt_nucleus=hormone2cell.hormone_strength(\n",
    "    ave_all = ave_all_nucleus,\n",
    "    geneset_definition = receptor_definition,\n",
    "    celltype_column = \"Celltype_tissue\",\n",
    "    tissue_col = \"Tissue\",\n",
    "    adjustment =  False,\n",
    "    assay = \"cell\",\n",
    "    include_cols = ['receptorgene1', 'receptorgene2'],\n",
    "    exclude_cols =[],\n",
    "    # initial thresholds\n",
    "    thresh_expr_low = 0.01,\n",
    "    thresh_pct = 5,\n",
    "    # filtering thresholds (first pass)\n",
    "    thresh_included_initial = 0.1,\n",
    "    thresh_excluded_initial = 0.15,\n",
    "    max_expression_file='max_expression_nucleus_receptor.csv'\n",
    ")\n",
    "display(dt_cell.head(2))\n",
    "dt_nucleus.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "alike-corrections",
   "metadata": {},
   "source": [
    "#### 3.2.1 Combine results from single cell and single nucleus data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "advanced-example",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(424, 9)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hrc_datadriven=hormone2cell.combine_assay(cell=dt_cell, nucleus=dt_nucleus,celltype_column='Celltype_tissue')\n",
    "hrc_datadriven.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "upset-interim",
   "metadata": {},
   "source": [
    "### 3.2 Identify hormone-receiving cell types using precomputed thresholds from the Hormone Cell Atlas."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "antique-sentence",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "      Hormone                       Celltype_tissue   Strength    Tissue\n",
      "612  resistin  Pancreas____pancreatic stellate cell  14.868078  Pancreas\n",
      "52       igf2          Pancreas____endothelial cell  12.381264  Pancreas\n",
      "Before removing duplicates: (848, 7)\n",
      "After removing duplicates: (311, 7)\n",
      "Finally hormone2cell finds: 311 hormone-celltype pairs.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Hormone</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Strength</th>\n",
       "      <th>Tissue</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>Tier</th>\n",
       "      <th>assay</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>gastrin</td>\n",
       "      <td>Pancreas____pancreatic D cell</td>\n",
       "      <td>0.422587</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Gastrin</td>\n",
       "      <td>gastrin</td>\n",
       "      <td>2</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>gastrin</td>\n",
       "      <td>Pancreas____pancreatic PP cell</td>\n",
       "      <td>0.303184</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Gastrin</td>\n",
       "      <td>gastrin</td>\n",
       "      <td>2</td>\n",
       "      <td>cell</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Hormone                 Celltype_tissue  Strength    Tissue  \\\n",
       "0  gastrin   Pancreas____pancreatic D cell  0.422587  Pancreas   \n",
       "1  gastrin  Pancreas____pancreatic PP cell  0.303184  Pancreas   \n",
       "\n",
       "  hormone_display hormone_short Tier assay  \n",
       "0         Gastrin       gastrin    2  cell  \n",
       "1         Gastrin       gastrin    2  cell  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 390 ms, sys: 2.1 ms, total: 392 ms\n",
      "Wall time: 394 ms\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(311, 8)"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "mask=ave_all['suspension_type']=='cell'\n",
    "ave_all_cell=ave_all.loc[mask]\n",
    "dt_cell=hormone2cell.hormone_strength(\n",
    "    ave_all = ave_all_cell,\n",
    "    geneset_definition = receptor_definition,\n",
    "    celltype_column = \"Celltype_tissue\",\n",
    "    tissue_col = \"Tissue\",\n",
    "    adjustment =  False,\n",
    "    assay = \"cell\",\n",
    "    include_cols = ['receptorgene1', 'receptorgene2'],\n",
    "    exclude_cols =[],\n",
    "    # initial thresholds\n",
    "    thresh_expr_low = 0.01,\n",
    "    thresh_pct = 5,\n",
    "    # filtering thresholds (first pass)\n",
    "    thresh_included_initial = 0.1,\n",
    "    thresh_excluded_initial = 0.15,\n",
    "    use_precomputed='cell',\n",
    "    max_expression_file='max_expression_cell_receptor2.csv'\n",
    ")\n",
    "display(dt_cell.head(2))\n",
    "dt_cell.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "genetic-cooking",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "    1) Filter the input expression table\n",
      "    2) Build gene–celltype matrix for included genes (lower threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    3) Build gene–celltype matrix for excluded genes (higher threshold)\n",
      "            Use Max_logExp from precomputed data\n",
      "    4) Apply hormone rules to compute the final table\n",
      "    Hormone                     Celltype_tissue   Strength    Tissue\n",
      "481    igf2  Pancreas____pancreatic acinar cell  18.891066  Pancreas\n",
      "568    igf2  Pancreas____pancreatic ductal cell  16.461430  Pancreas\n",
      "Before removing duplicates: (744, 7)\n",
      "After removing duplicates: (291, 7)\n",
      "Finally hormone2cell finds: 291 hormone-celltype pairs.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Hormone</th>\n",
       "      <th>Celltype_tissue</th>\n",
       "      <th>Strength</th>\n",
       "      <th>Tissue</th>\n",
       "      <th>hormone_display</th>\n",
       "      <th>hormone_short</th>\n",
       "      <th>Tier</th>\n",
       "      <th>assay</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>gastrin</td>\n",
       "      <td>Pancreas____pancreatic acinar cell</td>\n",
       "      <td>5.188416</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Gastrin</td>\n",
       "      <td>gastrin</td>\n",
       "      <td>2</td>\n",
       "      <td>nucleus</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>insulin</td>\n",
       "      <td>Pancreas____pancreatic acinar cell</td>\n",
       "      <td>7.339443</td>\n",
       "      <td>Pancreas</td>\n",
       "      <td>Insulin</td>\n",
       "      <td>insulin</td>\n",
       "      <td>1</td>\n",
       "      <td>nucleus</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Hormone                     Celltype_tissue  Strength    Tissue  \\\n",
       "0  gastrin  Pancreas____pancreatic acinar cell  5.188416  Pancreas   \n",
       "1  insulin  Pancreas____pancreatic acinar cell  7.339443  Pancreas   \n",
       "\n",
       "  hormone_display hormone_short Tier    assay  \n",
       "0         Gastrin       gastrin    2  nucleus  \n",
       "1         Insulin       insulin    1  nucleus  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "(291, 8)"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mask=ave_all['suspension_type']=='nucleus'\n",
    "ave_all_nucleus=ave_all.loc[mask]\n",
    "dt_nucleus=hormone2cell.hormone_strength(\n",
    "    ave_all = ave_all_nucleus,\n",
    "    geneset_definition = receptor_definition,\n",
    "    celltype_column = \"Celltype_tissue\",\n",
    "    tissue_col = \"Tissue\",\n",
    "    adjustment =  False,\n",
    "    assay = \"nucleus\",\n",
    "    include_cols = ['receptorgene1', 'receptorgene2'],\n",
    "    exclude_cols =[],\n",
    "    # initial thresholds\n",
    "    thresh_expr_low = 0.01,\n",
    "    thresh_pct = 5,\n",
    "    # filtering thresholds (first pass)\n",
    "    thresh_included_initial = 0.1,\n",
    "    thresh_excluded_initial = 0.15,\n",
    "    use_precomputed='nucleus',\n",
    "    max_expression_file='max_expression_nucleus_receptor2.csv'\n",
    ")\n",
    "display(dt_nucleus.head(2))\n",
    "dt_nucleus.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "lasting-butterfly",
   "metadata": {},
   "source": [
    "#### 3.2.1 Combine results from single cell and single nucleus data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "portuguese-frost",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(427, 9)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hrc_precomputed=hormone2cell.combine_assay(cell=dt_cell, nucleus=dt_nucleus,celltype_column='Celltype_tissue')\n",
    "hrc_precomputed.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fifteen-separation",
   "metadata": {},
   "source": [
    "Visulization of the hormone producing cell types from two methods."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "olympic-fairy",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 480x480 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 387,
       "width": 381
      }
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "from matplotlib_venn import venn2, venn3\n",
    "\n",
    "set1 = set(hrc_datadriven['Hormone_CT'])\n",
    "set2 = set(hrc_precomputed['Hormone_CT'])\n",
    "\n",
    "plt.figure(figsize=(6,6))\n",
    "venn2([set1, set2], set_labels=(\"Data-driven\", \"Precomputated\"))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "funky-trouble",
   "metadata": {},
   "outputs": [],
   "source": [
    "# hrc_datadriven.to_csv('hrc_datadriven.csv')\n",
    "# hrc_precomputed.to_csv('hrc_precomputed.csv')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "hormone2cell_env2",
   "language": "python",
   "name": "hormone2cell_env2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.19"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}