3A.[model compare]compare_semantic_similarity.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib as mpl\n",
    "import seaborn as sns\n",
    "from glob import glob\n",
    "import re\n",
    "from scipy.stats import wilcoxon, mannwhitneyu\n",
    "\n",
    "# Set default configuration for all plots\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",
    "plt.rcParams['axes.linewidth'] = 0.5\n",
    "plt.rcParams['hatch.linewidth'] = 0.5\n",
    "plt.rcParams['xtick.major.width'] = 0.4\n",
    "plt.rcParams['ytick.major.width'] = 0.4\n",
    "plt.rcParams['xtick.minor.width'] = 0.3\n",
    "plt.rcParams['ytick.minor.width'] = 0.3\n",
    "plt.rcParams['legend.frameon'] = False\n",
    "plt.rcParams['xtick.major.size'] = 3\n",
    "plt.rcParams['ytick.major.size'] = 3\n",
    "plt.rcParams['xtick.minor.size'] = 2\n",
    "plt.rcParams['ytick.minor.size'] = 2\n",
    "plt.rcParams['xtick.major.pad'] = 1\n",
    "plt.rcParams['ytick.major.pad'] = 1\n",
    "plt.rcParams['axes.labelpad'] = 1\n",
    "plt.rcParams['patch.linewidth'] = 0.25\n",
    "plt.figsize = (2,2)    \n",
    "\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# open the file \n",
    "files= glob('data/GO_term_analysis/model_compare/sim_rank_*.tsv')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['data/GO_term_analysis/model_compare/sim_rank_LLM_processed_model_compare_100set_gpt_4.tsv',\n",
       " 'data/GO_term_analysis/model_compare/sim_rank_LLM_processed_model_compare_100set_mixtral_instruct.tsv',\n",
       " 'data/GO_term_analysis/model_compare/sim_rank_LLM_processed_model_compare_100set_llama2_70b.tsv',\n",
       " 'data/GO_term_analysis/model_compare/sim_rank_LLM_processed_model_compare_100set_gemini_pro.tsv',\n",
       " 'data/GO_term_analysis/model_compare/sim_rank_LLM_processed_model_compare_100set_gpt_35.tsv']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(100, 6)\n",
      "(200, 6)\n",
      "(300, 6)\n",
      "(400, 6)\n",
      "(500, 6)\n"
     ]
    }
   ],
   "source": [
    "## combine files and sort by semantic similarity score\n",
    "all_names_df = pd.DataFrame()\n",
    "for file in files:\n",
    "    df = pd.read_csv(file, sep='\\t')\n",
    "    model_name = '_'.join(file.split('/')[-1].split('.')[0].split('_')[-2:])\n",
    "    # print(model_name)\n",
    "    if model_name == 'gpt_35':\n",
    "        prefix = 'gpt_3.5'\n",
    "    else:\n",
    "        prefix = model_name\n",
    "        \n",
    "    #rename the column\n",
    "    df.rename(columns={f'{prefix}_default Name': 'LLM name', f'{prefix}_default Score': 'LLM Score'}, inplace=True)\n",
    "    df = df.loc[:,['GO', 'Term_Description','LLM name', 'LLM Score','LLM_name_GO_term_sim']]\n",
    "    df.loc[:,'Model'] = prefix\n",
    "    all_names_df = pd.concat([all_names_df, df])\n",
    "    print(all_names_df.shape)\n",
    "all_names_df.sort_values(by='LLM_name_GO_term_sim', ascending=False, inplace=True)\n",
    "\n",
    "output_df = pd.concat([all_names_df.head(5), all_names_df.tail(5)])\n",
    "output_df.to_csv('data/GO_term_analysis/model_compare/all_model_top_bottom_5_GO_terms.tsv', sep='\\t', index=False) # Table 1 in the manuscript"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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>GO</th>\n",
       "      <th>LLM_name_GO_term_sim</th>\n",
       "      <th>model</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>GO:0014744</td>\n",
       "      <td>0.483668</td>\n",
       "      <td>GPT-4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>GO:0000338</td>\n",
       "      <td>0.391335</td>\n",
       "      <td>GPT-4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>GO:0060696</td>\n",
       "      <td>0.479061</td>\n",
       "      <td>GPT-4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>GO:0019885</td>\n",
       "      <td>0.752696</td>\n",
       "      <td>GPT-4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>GO:0003166</td>\n",
       "      <td>0.548072</td>\n",
       "      <td>GPT-4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           GO  LLM_name_GO_term_sim  model\n",
       "0  GO:0014744              0.483668  GPT-4\n",
       "1  GO:0000338              0.391335  GPT-4\n",
       "2  GO:0060696              0.479061  GPT-4\n",
       "3  GO:0019885              0.752696  GPT-4\n",
       "4  GO:0003166              0.548072  GPT-4"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "combined_df = pd.DataFrame()\n",
    "for file in files:\n",
    "    df = pd.read_csv(file, sep='\\t')\n",
    "    #if the default Name is 'system of unrelated proteins' remove it from the dataframe\n",
    "\n",
    "    model_name = '_'.join(file.split('/')[-1].split('.')[0].split('_')[-2:])\n",
    "    # print(model_name)\n",
    "    if model_name == 'gpt_35':\n",
    "        prefix = 'gpt_3.5'\n",
    "    else:\n",
    "        prefix = model_name\n",
    "    df = df[df[f'{prefix}_default Name'] != 'System of unrelated proteins']\n",
    "    # print(df.shape)\n",
    "    sem_sim = df.loc[:, ['GO', 'LLM_name_GO_term_sim']]\n",
    "    # sem_sim = df.loc[:, ['GO', 'true_GO_term_sim_percentile']]\n",
    "    sem_sim['model'] = model_name\n",
    "    combined_df = pd.concat([combined_df, sem_sim])\n",
    "combined_df['model'] = combined_df['model'].map({'gpt_4': 'GPT-4', 'gemini_pro': 'Gemini Pro','gpt_35':'GPT-3.5', 'mixtral_instruct': 'Mixtral Instruct', 'llama2_70b': 'Llama2 70b'})\n",
    "combined_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_666107/1754759280.py:18: FutureWarning: Passing `palette` without assigning `hue` is deprecated.\n",
      "  sns.swarmplot(x='model', y='LLM_name_GO_term_sim', data=combined_df, ax=ax, order=order, palette=['#6CBE4C', '#9C80BB', '#65AFAF', '#E07A5F', '#4579BD'], size=1.75, alpha=1)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model:  GPT-4 median:  0.49406657499999995\n",
      "model:  Gemini Pro median:  0.45542836\n",
      "model:  GPT-3.5 median:  0.502566115\n",
      "model:  Mixtral Instruct median:  0.45095879\n",
      "model:  Llama2 70b median:  0.4035816\n",
      "GPT-4 vs Llama2 70b: p = 0.0023234923501029865\n",
      "GPT-3.5 vs Llama2 70b: p = 0.0008411283006340667\n",
      "Mixtral Instruct vs Llama2 70b: p = 0.01350122051667856\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_666107/1754759280.py:53: UserWarning: FixedFormatter should only be used together with FixedLocator\n",
      "  ax.set_xticklabels(['GPT-4', 'Gemini\\nPro', 'GPT-3.5','Mixtral\\nInstruct', 'Llama2\\n70b'],rotation=0, ha='center')\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 250x250 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(2.5,2.5))\n",
    "order = ['GPT-4', 'Gemini Pro', 'GPT-3.5','Mixtral Instruct', 'Llama2 70b']\n",
    "# Perform Mann-Whitney U tests\n",
    "p_values = {}\n",
    "for i in range(len(order)):\n",
    "    for j in range(i+1, len(order)):\n",
    "        model1 = order[i]\n",
    "        model2 = order[j]\n",
    "        data1 = combined_df[combined_df['model'] == model1]['LLM_name_GO_term_sim']\n",
    "        data2 = combined_df[combined_df['model'] == model2]['LLM_name_GO_term_sim']\n",
    "        stat, p = mannwhitneyu(data1, data2)\n",
    "        p_values[(model1, model2)] = p\n",
    "        \n",
    "#855C75,#D9AF6B,#AF6458,#736F4C,#526A83,#625377,#68855C,#9C9C5E,#A06177,#8C785D,#467378,#7C7C7C\n",
    "# sns.violinplot(x='model', y='LLM_name_GO_term_sim', data=combined_df, ax=ax, order=order, inner='quartiles', linewidth=0.5, width=0.8,cut=0, palette=['#d88c9a','#f2d0a9','#f1e3d3','#99c1b9','#d2b7e5'], edgecolor='black')\n",
    "\n",
    "# Add swarm plot\n",
    "sns.swarmplot(x='model', y='LLM_name_GO_term_sim', data=combined_df, ax=ax, order=order, palette=['#6CBE4C', '#9C80BB', '#65AFAF', '#E07A5F', '#4579BD'], size=1.75, alpha=1)\n",
    "# add median line \n",
    "medians = combined_df.groupby(['model'])['LLM_name_GO_term_sim'].median()\n",
    "for i in range(len(order)):\n",
    "    print('model: ', order[i], 'median: ', medians[order[i]])\n",
    "    ax.plot([i-0.25, i+0.25], [medians[order[i]], medians[order[i]]], lw=1, color='black')\n",
    "\n",
    "\n",
    "# Determine the maximum y-value\n",
    "max_y = max(combined_df['LLM_name_GO_term_sim']) + 0.03  # Adjust the 0.03 if needed\n",
    "\n",
    "# Offset for each bar to prevent overlap\n",
    "bar_offset = max_y * 0.05  \n",
    "i = 0\n",
    "# Loop through your p-values and add bars for each significant pair\n",
    "for index, ((model1, model2), p_value) in enumerate(p_values.items()):\n",
    "    if p_value < 0.05:  # Only plot bars for significant differences\n",
    "        bar_start = order.index(model1)  # x-coordinate for model1\n",
    "        bar_end = order.index(model2)    # x-coordinate for model2\n",
    "\n",
    "        # Adjust y-coordinate for the bar to avoid overlap\n",
    "        y_coord = max_y + bar_offset + (i * bar_offset)\n",
    "\n",
    "        # Plot the horizontal line for the significance bar\n",
    "        ax.plot([bar_start, bar_end], [y_coord, y_coord], color='black', lw=0.5)\n",
    "\n",
    "        # Add text for significance level\n",
    "        significance_text = \"****\" if p_value < 0.0001 else \"***\" if p_value < 0.001 else \"**\" if p_value < 0.01 else \"*\"\n",
    "        ax.text((bar_start + bar_end) / 2, y_coord - 0.01, significance_text, ha='center')\n",
    "        print(f'{model1} vs {model2}: p = {p_value}')\n",
    "        i += 1\n",
    "\n",
    "\n",
    "ax.set_ylabel('Semantic similarity \\n between LLM name and GO term name')\n",
    "ax.set_xlabel('')\n",
    "ax.set_xticklabels(['GPT-4', 'Gemini\\nPro', 'GPT-3.5','Mixtral\\nInstruct', 'Llama2\\n70b'],rotation=0, ha='center')\n",
    "sns.despine()\n",
    "# plt.savefig('figures/model_compare/compare_raw_semantic_similarity_swamp_only.svg', bbox_inches='tight')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Compare the semantic similarity vs confidence score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 250x250 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from scipy.stats import pearsonr\n",
    "\n",
    "# Load the data\n",
    "input_file = 'data/GO_term_analysis/model_compare/sim_rank_LLM_processed_model_compare_100set_gpt_4.tsv'\n",
    "df = pd.read_csv(input_file, sep='\\t')\n",
    "\n",
    "# Separate the data\n",
    "zero_df = df[df['gpt_4_default Name'] == 'System of unrelated proteins']\n",
    "non_zero_df = df[df['gpt_4_default Name'] != 'System of unrelated proteins']\n",
    "\n",
    "\n",
    "# Create subplots\n",
    "fig, (ax0, ax1) = plt.subplots(1, 2, sharey=True, figsize=(2.5,2.5), gridspec_kw={'width_ratios': [1, 3]})\n",
    "fig.subplots_adjust(wspace=0.01)\n",
    "\n",
    "# First plot for zeros\n",
    "sns.scatterplot(x = 'gpt_4_default Score', y = 'LLM_name_GO_term_sim',data=zero_df, alpha=1, s=8, ax=ax0, color = '#A5AA99')\n",
    "ax0.set_xlim(-0.05, 0.05)\n",
    "ax0.set_xlabel('')\n",
    "ax0.spines['right'].set_visible(False)\n",
    "ax0.yaxis.tick_left()\n",
    "#remove y-axis label\n",
    "ax0.set_ylabel('')\n",
    "sns.despine(ax=ax0)\n",
    "\n",
    "# Define the color palette\n",
    "palette = {\n",
    "    'Low Confidence': '#bc6c25',\n",
    "    'Medium Confidence': '#0077b6',\n",
    "    'High Confidence': '#e63946'\n",
    "}\n",
    "\n",
    "# Second plot for non-zeros\n",
    "sns.scatterplot(x = 'gpt_4_default Score', y = 'LLM_name_GO_term_sim',data=non_zero_df, hue = 'gpt_4_default Score Bin',palette=palette, alpha=1, s=8, ax=ax1, legend=False)\n",
    "# plot best fit line\n",
    "m, b = np.polyfit(non_zero_df['gpt_4_default Score'], non_zero_df['LLM_name_GO_term_sim'], 1)\n",
    "ax1.plot(non_zero_df['gpt_4_default Score'], m*non_zero_df['gpt_4_default Score'] + b, color='grey', lw=0.75)\n",
    "\n",
    "ax1.set_xlim(0.80, 1.0)\n",
    "ax1.set_xlabel('')\n",
    "ax1.tick_params(labelright=False)  # don't put tick labels at the right\n",
    "# remove ticks\n",
    "ax1.yaxis.set_ticks_position('none')\n",
    "sns.despine(ax=ax1, left=True)\n",
    "\n",
    "\n",
    "# Add correlation coefficient\n",
    "r, p = pearsonr(non_zero_df['gpt_4_default Score'], non_zero_df['LLM_name_GO_term_sim'])\n",
    "fig.text(0.15, 0.9, f'y = {m:.2f}x + {b:.2f}', ha='left', va='center', fontsize=7)\n",
    "fig.text(0.15, 0.95, f'R = {r:.2f}', ha='left', va='center', fontsize=7)\n",
    "fig.text(0.15, 0.85, f'p = {p:.2e}', ha='left', va='center', fontsize=7)\n",
    "\n",
    "# Add common x-label and y-label\n",
    "fig.text(0.5, 0.01, 'GPT-4 confidence score', ha='center', va='center')\n",
    "fig.text(0, 0.5, 'Semantic similarity\\nbetween LLM name and GO name', ha='center', va='center', rotation='vertical')\n",
    "\n",
    "# Add the slanted lines to indicate the break at the bottom\n",
    "d = .25  # proportion of vertical to horizontal extent of the slanted line\n",
    "kwargs = dict(marker=[(-d, -1), (d, 1)], markersize=7,\n",
    "              linestyle=\"none\", color='black', mec='k', mew=1, clip_on=False)\n",
    "ax0.plot([1, 1], [0, 0], transform=ax0.transAxes, **kwargs)\n",
    "ax1.plot([0, 0], [0, 0], transform=ax1.transAxes, **kwargs)\n",
    "\n",
    "# Adjust the layout\n",
    "plt.tight_layout(w_pad=0.05)\n",
    "# remove legend, will add it later\n",
    "plt.legend('') \n",
    "# save \n",
    "plt.savefig('figures/model_compare/compare_semantic_similarity_vs_raw_confidence_score_scatter_GPT4_split.svg', bbox_inches='tight')\n",
    "# Show the plot\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
}