3A.[omics_revamped]Analyze_gprofiler_annotation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## g:profiler enrichment analysis evaluation \n",
    "\n",
    "updated the gene name. \n",
    "    \n",
    "1. pick the smallest p val\n",
    "    * pick the smallest p , then add JI requirement\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "LLM_score_thresh = 0.01 # only drop the ones LLM does not assign a name \n",
    "adj_pval_thresh = 0.05\n",
    "JI_thresh = 0.05"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "plt.rcParams['svg.fonttype'] = 'none'\n",
    "plt.rcParams.update({'font.size': 7, 'font.family': 'sans-serif'})\n",
    "plt.rcParams['xtick.labelsize'] = 7\n",
    "plt.rcParams['ytick.labelsize'] = 7\n",
    "plt.rcParams['axes.labelsize'] = 7\n",
    "plt.rcParams['axes.titlesize'] = 7\n",
    "\n",
    "from utils.analyze_enrichment_utils import *\n",
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data files from other analyses:\n",
    " - inputfile: LLM analyzed + gProfiler enrichment analysis on updated gene symbols\n",
    " - LLM coverage analysis - used GPT-4 to review the analysis text to determine which genes met the criteria that they are \"definite\" supports, not conjecture.\n",
    " - Best matching GO terms to LLM names, used to estimate how many genes match the name, enabling the computation of a JI"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>Source</th>\n",
       "      <th>GeneSetID</th>\n",
       "      <th>GeneSetName</th>\n",
       "      <th>GeneList</th>\n",
       "      <th>updated GeneList</th>\n",
       "      <th>n_Genes</th>\n",
       "      <th>LLM Name</th>\n",
       "      <th>LLM Analysis</th>\n",
       "      <th>Score</th>\n",
       "      <th>Term</th>\n",
       "      <th>GO ID</th>\n",
       "      <th>Adjusted P-value</th>\n",
       "      <th>intersection_size</th>\n",
       "      <th>term_size</th>\n",
       "      <th>query_size</th>\n",
       "      <th>intersections</th>\n",
       "      <th>gprofiler_JI</th>\n",
       "      <th>gprofiler_coverage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NeST</td>\n",
       "      <td>Cluster1-10</td>\n",
       "      <td>Cluster1-10</td>\n",
       "      <td>CTRL HSD17B14 KIAA0232 PAQR8 PLA2G1B RNF145 SG...</td>\n",
       "      <td>CTRL HSD17B14 KIAA0232 PAQR8 PLA2G1B RNF145 SG...</td>\n",
       "      <td>12</td>\n",
       "      <td>Lipid Metabolism and Membrane Dynamics</td>\n",
       "      <td>1. CTRL, or carboxyl-terminal esterase/lipase,...</td>\n",
       "      <td>0.85</td>\n",
       "      <td>L-leucine import across plasma membrane</td>\n",
       "      <td>GO:1903801</td>\n",
       "      <td>0.080467</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>10</td>\n",
       "      <td>SLC7A8</td>\n",
       "      <td>0.071429</td>\n",
       "      <td>0.1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  Source    GeneSetID  GeneSetName  \\\n",
       "0   NeST  Cluster1-10  Cluster1-10   \n",
       "\n",
       "                                            GeneList  \\\n",
       "0  CTRL HSD17B14 KIAA0232 PAQR8 PLA2G1B RNF145 SG...   \n",
       "\n",
       "                                    updated GeneList  n_Genes  \\\n",
       "0  CTRL HSD17B14 KIAA0232 PAQR8 PLA2G1B RNF145 SG...       12   \n",
       "\n",
       "                                 LLM Name  \\\n",
       "0  Lipid Metabolism and Membrane Dynamics   \n",
       "\n",
       "                                        LLM Analysis  Score  \\\n",
       "0  1. CTRL, or carboxyl-terminal esterase/lipase,...   0.85   \n",
       "\n",
       "                                      Term       GO ID  Adjusted P-value  \\\n",
       "0  L-leucine import across plasma membrane  GO:1903801          0.080467   \n",
       "\n",
       "   intersection_size  term_size  query_size intersections  gprofiler_JI  \\\n",
       "0                  1          5          10        SLC7A8      0.071429   \n",
       "\n",
       "   gprofiler_coverage  \n",
       "0                 0.1  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "import pandas as pd\n",
    "from utils.analyze_enrichment_utils import cal_JI_coverage\n",
    "\n",
    "group_col = [\"Source\", \"GeneSetID\", \"GeneSetName\", \"GeneList\"] # group by these columns in future steps\n",
    "\n",
    "## input file: LLM analyzed + gProfiler enrichment\n",
    "updated_name_llm_file = 'data/omics_revamped_LLM_gprofiler_new_gene_name_DF.tsv'\n",
    "\n",
    "# Read the TSV file into a Pandas DataFrame\n",
    "LLM_genes_DF = pd.read_csv(updated_name_llm_file, delimiter=\"\\t\")\n",
    "# make sure intersection_size, term_size, query_size are integers\n",
    "LLM_genes_DF['intersection_size'] = LLM_genes_DF['intersection_size'].apply(lambda x: int(x))\n",
    "LLM_genes_DF['term_size'] = LLM_genes_DF['term_size'].apply(lambda x: int(x))\n",
    "LLM_genes_DF['query_size'] = LLM_genes_DF['query_size'].apply(lambda x: int(x))\n",
    "#LLM_genes_DF.shape\n",
    "LLM_genes_DF = cal_JI_coverage(LLM_genes_DF)\n",
    "LLM_genes_DF.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "## load the LLM coverage data as a common df to merge\n",
    "llm_coverage_df = pd.read_csv(\"data/omics_revamped_LLM_genecounts_DF.tsv\", sep=\"\\t\")\n",
    "\n",
    "llm_JI_file = 'data/omics_revamped_LLM_w_best_matching_GO_terms_for_JI.tsv'\n",
    "llm_ji_df = pd.read_csv(llm_JI_file, sep=\"\\t\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(53452, 23)\n",
      "(53452, 28)\n",
      "Index(['Source', 'GeneSetID', 'GeneSetName', 'GeneList', 'updated GeneList',\n",
      "       'n_Genes', 'LLM Name', 'LLM Analysis', 'Score', 'Term', 'GO ID',\n",
      "       'Adjusted P-value', 'intersection_size', 'term_size', 'query_size',\n",
      "       'intersections', 'gprofiler_JI', 'gprofiler_coverage',\n",
      "       'Supporting Genes', 'Supporting Count', 'LLM Support Analysis',\n",
      "       'GenesMentionedInText', 'LLM_coverage', 'LLM_best_matching_GO',\n",
      "       'best_matching_GO_ID', 'best_matching_GO_term_genes',\n",
      "       'best_matching_GO_term_size', 'LLM_JI'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "# merge with the LLM coverage data\n",
    "merged_genes_DF = pd.merge(LLM_genes_DF, llm_coverage_df, on=['Source','GeneSetID','GeneSetName', 'GeneList', 'n_Genes'], how='left')\n",
    "print(merged_genes_DF.shape)\n",
    "# print(merged_genes_DF.columns)\n",
    "\n",
    "\n",
    "# merge with the LLM JI data\n",
    "\n",
    "merged_genes_DF = pd.merge(merged_genes_DF, llm_ji_df, on=['Source','GeneSetID','GeneList','n_Genes', 'LLM Name','Supporting Count'], how='left')\n",
    "print(merged_genes_DF.shape)\n",
    "print(merged_genes_DF.columns)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Pick Best APV\n",
    "\n",
    " - Select the best GO by the smallest p-value\n",
    " - Filter by p-value. \n",
    "\n",
    "If we later apply a coverage threshold to generate this curve, the curve will not change if the threshold is changed because the enriched term is not picked based on the coverage threshold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LLM_success_TF        True  False  Total\n",
      "gprofiler_success_TF                    \n",
      "True                   124    105    229\n",
      "False                   11     60     71\n",
      "Total                  135    165    300\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>n_Genes</th>\n",
       "      <th>GeneList</th>\n",
       "      <th>LLM Name</th>\n",
       "      <th>Score</th>\n",
       "      <th>Term</th>\n",
       "      <th>GO ID</th>\n",
       "      <th>Adjusted P-value</th>\n",
       "      <th>intersection_size</th>\n",
       "      <th>term_size</th>\n",
       "      <th>intersections</th>\n",
       "      <th>gprofiler_coverage</th>\n",
       "      <th>LLM_coverage</th>\n",
       "      <th>gprofiler_JI</th>\n",
       "      <th>LLM_success_TF</th>\n",
       "      <th>gprofiler_success_TF</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>24</td>\n",
       "      <td>CITED2 COL5A1 CRABP2 KCTD12 MDFIC MMP2 NRP1 OR...</td>\n",
       "      <td>Cellular Matrix Remodeling and Tissue Development</td>\n",
       "      <td>0.85</td>\n",
       "      <td>collagen metabolic process</td>\n",
       "      <td>GO:0032963</td>\n",
       "      <td>0.005761</td>\n",
       "      <td>4</td>\n",
       "      <td>101</td>\n",
       "      <td>COL5A1,MMP2,VIM,CST3</td>\n",
       "      <td>0.181818</td>\n",
       "      <td>0.21</td>\n",
       "      <td>0.033613</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>47</td>\n",
       "      <td>1060P11.3 ADM AHR AMIGO2 ARL4C ATP10D CAV2 CD4...</td>\n",
       "      <td>Cellular Adhesion and Extracellular Matrix Int...</td>\n",
       "      <td>0.85</td>\n",
       "      <td>inflammatory response</td>\n",
       "      <td>GO:0006954</td>\n",
       "      <td>0.001123</td>\n",
       "      <td>11</td>\n",
       "      <td>829</td>\n",
       "      <td>1060P11.3,ADM,CAV2,CDC42EP3,EMP1,LPAR6,TRIB2,A...</td>\n",
       "      <td>0.250000</td>\n",
       "      <td>0.13</td>\n",
       "      <td>0.012761</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>59</td>\n",
       "      <td>ABAT ASS1 CHI3L1 CHST2 CLDN3 EIF5B FRZB GAL HE...</td>\n",
       "      <td>System of unrelated proteins</td>\n",
       "      <td>0.00</td>\n",
       "      <td>actomyosin structure organization</td>\n",
       "      <td>GO:0031032</td>\n",
       "      <td>0.005933</td>\n",
       "      <td>6</td>\n",
       "      <td>204</td>\n",
       "      <td>AKR1B1,CAPN3,DKK1,MARCKS,MTUS1,OPN3</td>\n",
       "      <td>0.113208</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.023904</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>9</td>\n",
       "      <td>AMIGO2 AREG GAS6 GPR37 IFT57 PELI1 SQLE AKAP12...</td>\n",
       "      <td>System of unrelated proteins</td>\n",
       "      <td>0.00</td>\n",
       "      <td>regulation of oligodendrocyte apoptotic process</td>\n",
       "      <td>GO:1900141</td>\n",
       "      <td>0.000143</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>GAS6,AKAP12</td>\n",
       "      <td>0.222222</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.222222</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>15</td>\n",
       "      <td>1060P11.3 CPE EFNB2 HIST1H2AC IL1R2 INHBB LYPD...</td>\n",
       "      <td>System of unrelated proteins</td>\n",
       "      <td>0.00</td>\n",
       "      <td>cellular response to interleukin-1</td>\n",
       "      <td>GO:0071347</td>\n",
       "      <td>0.000712</td>\n",
       "      <td>4</td>\n",
       "      <td>106</td>\n",
       "      <td>H2AC6,IL1R2,NUP93,SOX9</td>\n",
       "      <td>0.285714</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.034483</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   n_Genes                                           GeneList  \\\n",
       "0       24  CITED2 COL5A1 CRABP2 KCTD12 MDFIC MMP2 NRP1 OR...   \n",
       "1       47  1060P11.3 ADM AHR AMIGO2 ARL4C ATP10D CAV2 CD4...   \n",
       "2       59  ABAT ASS1 CHI3L1 CHST2 CLDN3 EIF5B FRZB GAL HE...   \n",
       "3        9  AMIGO2 AREG GAS6 GPR37 IFT57 PELI1 SQLE AKAP12...   \n",
       "4       15  1060P11.3 CPE EFNB2 HIST1H2AC IL1R2 INHBB LYPD...   \n",
       "\n",
       "                                            LLM Name  Score  \\\n",
       "0  Cellular Matrix Remodeling and Tissue Development   0.85   \n",
       "1  Cellular Adhesion and Extracellular Matrix Int...   0.85   \n",
       "2                       System of unrelated proteins   0.00   \n",
       "3                       System of unrelated proteins   0.00   \n",
       "4                       System of unrelated proteins   0.00   \n",
       "\n",
       "                                              Term       GO ID  \\\n",
       "0                       collagen metabolic process  GO:0032963   \n",
       "1                            inflammatory response  GO:0006954   \n",
       "2                actomyosin structure organization  GO:0031032   \n",
       "3  regulation of oligodendrocyte apoptotic process  GO:1900141   \n",
       "4               cellular response to interleukin-1  GO:0071347   \n",
       "\n",
       "   Adjusted P-value  intersection_size  term_size  \\\n",
       "0          0.005761                  4        101   \n",
       "1          0.001123                 11        829   \n",
       "2          0.005933                  6        204   \n",
       "3          0.000143                  2          2   \n",
       "4          0.000712                  4        106   \n",
       "\n",
       "                                       intersections  gprofiler_coverage  \\\n",
       "0                               COL5A1,MMP2,VIM,CST3            0.181818   \n",
       "1  1060P11.3,ADM,CAV2,CDC42EP3,EMP1,LPAR6,TRIB2,A...            0.250000   \n",
       "2                AKR1B1,CAPN3,DKK1,MARCKS,MTUS1,OPN3            0.113208   \n",
       "3                                        GAS6,AKAP12            0.222222   \n",
       "4                             H2AC6,IL1R2,NUP93,SOX9            0.285714   \n",
       "\n",
       "   LLM_coverage  gprofiler_JI  LLM_success_TF  gprofiler_success_TF  \n",
       "0          0.21      0.033613            True                  True  \n",
       "1          0.13      0.012761            True                  True  \n",
       "2           NaN      0.023904           False                  True  \n",
       "3           NaN      0.222222           False                  True  \n",
       "4           NaN      0.034483           False                  True  "
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from utils.analyze_enrichment_utils import get_min_adj_p_value\n",
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "coverage_thresh = 0.2\n",
    "\n",
    "# Grouping the DataFrame by the specified columns\n",
    "grouped = merged_genes_DF.groupby(group_col)\n",
    "\n",
    "# Apply the function to each group and reset the index\n",
    "reduced_LLM_genes_APV_only_DF = grouped.apply(lambda x: get_min_adj_p_value(x, sortCol='gprofiler_JI')).reset_index(drop=True)\n",
    "\n",
    "# Add the 'LLM_success_TF' and 'gprofiler_success_TF' columns\n",
    "reduced_LLM_genes_APV_only_DF['LLM_success_TF'] = ((reduced_LLM_genes_APV_only_DF['Score'] >= LLM_score_thresh))\n",
    "reduced_LLM_genes_APV_only_DF['gprofiler_success_TF'] = ~((reduced_LLM_genes_APV_only_DF['Adjusted P-value'] > adj_pval_thresh)) # enrichment criteria\n",
    "\n",
    "print(create_success_contingency_table(reduced_LLM_genes_APV_only_DF))\n",
    "\n",
    "result_df = select_rows_and_columns(reduced_LLM_genes_APV_only_DF, 5)\n",
    "result_df\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# save the DataFrame to a TSV file\n",
    "reduced_LLM_genes_APV_only_DF.to_csv(\"data/omics_revamped_LLM_gprofiler_new_gene_name_DF_APV_only.tsv\", sep=\"\\t\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "coverage threshold: 0.0, enrichment: 229.0, LLM: 135.0\n",
      "coverage threshold: 0.05, enrichment: 84.0, LLM: 62.0\n",
      "coverage threshold: 0.1, enrichment: 33.0, LLM: 42.0\n",
      "coverage threshold: 0.2, enrichment: 4.0, LLM: 21.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/cellar/users/mhu/Projects/llm_evaluation_for_gene_set_interpretation/utils/analyze_enrichment_utils.py:190: UserWarning: Attempt to set non-positive ylim on a log-scaled axis will be ignored.\n",
      "  plt.ylim(-8, 300)\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 300x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from utils.analyze_enrichment_utils import plot_thresholding_res\n",
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "plot_thresholding_res(reduced_LLM_genes_APV_only_DF, 'gprofiler_JI', 'LLM_JI',coverage_thresh_list = np.arange(0.0, 1.1, 0.05), enrich_adj_pval_thresh = adj_pval_thresh, LLM_score_thresh = LLM_score_thresh, highlight_coverage = [0.0, 0.05, 0.1, 0.2],figsize=(3,3), ax_label_keyword= 'JI', log_scale=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Pick GO term with best APV then filter by JI thresholds \n",
    " - Select the best GO term by the smallest p-value\n",
    "      - If there are multiple GO terms with the same smallest p-value, then pick the term with the highest JI.\n",
    " - Filter by both the p-value and JI thresholds.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "JI threshold: 0.0\n",
      "LLM_success_TF        True  False  Total\n",
      "gprofiler_success_TF                    \n",
      "True                   124    105    229\n",
      "False                   11     60     71\n",
      "Total                  135    165    300\n",
      "JI threshold: 0.05\n",
      "LLM_success_TF        True  False  Total\n",
      "gprofiler_success_TF                    \n",
      "True                    33     51     84\n",
      "False                   29    187    216\n",
      "Total                   62    238    300\n",
      "JI threshold: 0.1\n",
      "LLM_success_TF        True  False  Total\n",
      "gprofiler_success_TF                    \n",
      "True                    11     22     33\n",
      "False                   31    236    267\n",
      "Total                   42    258    300\n",
      "JI threshold: 0.2\n",
      "LLM_success_TF        True  False  Total\n",
      "gprofiler_success_TF                    \n",
      "True                     0      4      4\n",
      "False                   21    275    296\n",
      "Total                   21    279    300\n"
     ]
    }
   ],
   "source": [
    "from utils.analyze_enrichment_utils import get_min_adj_p_value\n",
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "JI_thresh_list = [0.0, 0.05, 0.1, 0.2]\n",
    "for JI_thresh in JI_thresh_list:\n",
    "    print(f\"JI threshold: {JI_thresh}\")\n",
    "    # Grouping the DataFrame by the specified columns\n",
    "    grouped = merged_genes_DF.groupby(group_col)\n",
    "\n",
    "    # Apply the function to each group and reset the index\n",
    "    reduced_LLM_genes_APV_thred_JI_DF = grouped.apply(lambda x: get_min_adj_p_value(x, sortCol='gprofiler_JI')).reset_index(drop=True)\n",
    "\n",
    "\n",
    "    # Add the 'LLM_success_TF' and 'gprofiler_success_TF' columns\n",
    "    reduced_LLM_genes_APV_thred_JI_DF['LLM_success_TF'] = ((reduced_LLM_genes_APV_thred_JI_DF ['Score'] >= LLM_score_thresh)& (reduced_LLM_genes_APV_thred_JI_DF ['LLM_JI'] >= JI_thresh)) # LLM creteria\n",
    "    reduced_LLM_genes_APV_thred_JI_DF['gprofiler_success_TF'] = ~((reduced_LLM_genes_APV_thred_JI_DF ['Adjusted P-value'] > adj_pval_thresh) | (reduced_LLM_genes_APV_thred_JI_DF['gprofiler_JI'] < JI_thresh)) # enrichment criteria\n",
    "\n",
    "\n",
    "    print(create_success_contingency_table(reduced_LLM_genes_APV_thred_JI_DF))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Heatmap for example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 320x60 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# find the GeneSetID: Cluster2-105 from the full table\n",
    "example = merged_genes_DF.loc[\n",
    "merged_genes_DF['GeneSetID'] == 'Cluster2-105'].iloc[[0,2]] # one is the most significant, one is more similar name\n",
    "# example\n",
    "geneList = example['GeneList'].values[0].split(' ')\n",
    "\n",
    "new_cols = ['Name', 'Supporting Genes','Named by']\n",
    "new_df = pd.DataFrame(columns=new_cols)\n",
    "\n",
    "# LLM supporting genes\n",
    "llm_supporting_genes = example['Supporting Genes'].values[0].split(' ')\n",
    "# Join the LLM supporting genes list into a single string\n",
    "llm_supporting_genes_str = ', '.join(llm_supporting_genes)\n",
    "llm_name_row = pd.DataFrame({'Name': [example['LLM Name'].values[0]], 'Supporting Genes': llm_supporting_genes_str, 'Named by': ['GPT-4']})\n",
    "new_df = pd.concat([new_df, llm_name_row], ignore_index=True)\n",
    "\n",
    "for go_term in example['Term'].values:\n",
    "    intersection = example.loc[example['Term'] == go_term, 'intersections'].values[0].split(',')\n",
    "    # Join the intersection list into a single string\n",
    "    intersection_str = ', '.join(intersection)\n",
    "    new_row = pd.DataFrame({'Name': [go_term], 'Supporting Genes': intersection_str, 'Named by': ['g:Profiler']})\n",
    "    new_df = pd.concat([new_df, new_row], ignore_index=True)\n",
    "\n",
    "\n",
    "new_df = new_df.reset_index(drop=True)\n",
    "# new_df\n",
    "# geneList\n",
    "# plot a heatmap, where columns are the genes and rows are the gene sets, heat at the intersections\n",
    "\n",
    "# column is all the genes in the target gene set \n",
    "genes = set(geneList)\n",
    "genes = sorted(genes)\n",
    "\n",
    "# Create a binary matrix\n",
    "matrix = pd.DataFrame(0, index=new_df['Name'], columns=genes)\n",
    "for idx, row in new_df.iterrows():\n",
    "    supporting_genes = row['Supporting Genes'].split(', ')\n",
    "    matrix.loc[row['Name'], supporting_genes] = 1\n",
    "\n",
    "# Calculate appropriate figure size based on the number of genes and gene sets\n",
    "fig_height =len(new_df['Name']) * 0.2\n",
    "fig_width= len(genes) * 0.2\n",
    "\n",
    "# Plot the heatmap with square cells and adjusted figure size\n",
    "plt.figure(figsize=(fig_width, fig_height))\n",
    "ax = sns.heatmap(matrix, cmap=['#c0c0c0', '#ffbd00'], cbar=False, linewidths=.5, annot=False, square=True)\n",
    "# Move the x-axis labels to the top\n",
    "ax.xaxis.set_ticks_position('top')\n",
    "ax.xaxis.set_label_position('top')\n",
    "\n",
    "# Adjust the font size of the labels\n",
    "ax.set_xticklabels(ax.get_xticklabels(), rotation=45, ha='left')\n",
    "ax.set_yticklabels(ax.get_yticklabels(), rotation=0, ha='right')\n",
    "\n",
    "\n",
    "plt.xlabel('')\n",
    "plt.ylabel('')\n",
    "plt.savefig('figures/omics_LLM_gprofiler_CULLIN_example_gene_intersection_heatmap.svg', dpi=300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "llm_eval",
   "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": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}