jupyter/Evaluation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "63d9d4f4-2246-4e94-8c4f-b60337d4a26b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import pandas as pd\n",
    "import os\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "293cba7e-b622-4c1d-a8ba-94de56503320",
   "metadata": {},
   "outputs": [],
   "source": [
    "from dotenv import load_dotenv\n",
    "load_dotenv()\n",
    "api_key = os.getenv(\"MISTRAL_API_KEY\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "63cae255-5643-482a-9472-3f303f778772",
   "metadata": {},
   "source": [
    "# Evaluation of fine-tuned models\n",
    "\n",
    "## Generation of evaluation set\n",
    "\n",
    "Evaluation set will be composed of test items created with the same logic as **dataset examples**.\n",
    "Test items are combinations of : list of genes + list of compounds + matching pathway (=target) and are drawn from KEGG dataset.\n",
    "I created these random combinations in [testitemgeneration.py](../testitemgeneration.py). I made 200 test items.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fbc02975-8d5a-4c55-8f09-2fd7a8843326",
   "metadata": {},
   "source": [
    "## Interrogation of the models with the test dataset\n",
    "\n",
    "To be fair with the different models, a system prompt is added, in order to set the following guidelines :\n",
    "* Find the altered pathway\n",
    "* <200 words\n",
    "\n",
    "Following models were used :"
   ]
  },
  {
   "cell_type": "raw",
   "id": "d0385ce9-049a-468c-b91a-f35eb997842a",
   "metadata": {},
   "source": [
    "open-mistral-7b (LLM1)\n",
    "ft:open-mistral-7b:5aebfd1c:20240627:e63dc09f (LLM 2 - first model trained on 500 exemples)\n",
    "ft:open-mistral-7b:5aebfd1c:20240628:de14ae4a (LLM 3 - second model trained on 1000 exemples)\n",
    "mistral-small-latest (LLM 4)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "da403ebf-ecf0-49c8-a02e-f568334fc42c",
   "metadata": {},
   "source": [
    "I had the models pass this test using [evaluation.py](evaluation.py) script. Below is an example of how the results look like."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "bd179dc0-2c00-4125-9ccf-f506f23211fa",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "200"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "entries = []\n",
    "with open(\"test/test_pass.jsonl\", 'r') as infile:\n",
    "    for line in infile:\n",
    "        try:\n",
    "            entry = json.loads(line)\n",
    "            entries.append(entry)\n",
    "        except json.JSONDecodeError:\n",
    "            print(f\"Ignoring invalid JSON: {line}\")\n",
    "len(entries)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "e159ee5c-57b4-4d14-a019-a7018a2ef577",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'input': {'genes': ['PCCA', 'PGAM1'],\n",
       "  'cpd': ['Tetrahydrosarcinapterin',\n",
       "   'D-Xylulose 5-phosphate',\n",
       "   'CO2',\n",
       "   'L-erythro-3-Methylmalyl-CoA',\n",
       "   '2-Phosphoglycolate',\n",
       "   '5-Formyl-5,6,7,8-tetrahydromethanopterin',\n",
       "   'L-Serine',\n",
       "   'D-Glucono-1,5-lactone 6-phosphate',\n",
       "   'Glycolate',\n",
       "   'D-Glucono-1,5-lactone',\n",
       "   'D-Fructose 6-phosphate',\n",
       "   'D-Gluconic acid',\n",
       "   'Glycerone',\n",
       "   'D-Glycerate',\n",
       "   '4-Hydroxybutyryl-CoA']},\n",
       " 'target': 'Carbon metabolism',\n",
       " 'output': {'open-mistral-7b': 'The altered pathway involves the genes PCCA and PGAM1, which are associated with the glyoxylate cycle and the glycolysis pathway, respectively.\\n\\nThe glyoxylate cycle is a metabolic pathway that occurs in some microorganisms, plants, and animals. It allows these organisms to convert acetyl-CoA derived from fatty acids into glucose-6-phosphate, providing a means of glucose production when glucose is limited. The cycle involves two key enzymes, ISOCITRATE LYASE (ICL) and MALATE SYNTHASE (MLS), which are not present in the listed compounds.\\n\\nPCCA (Propionyl-CoA Carboxylase Alpha Chain) is an enzyme involved in the propionate metabolism pathway, which is closely related to the glyoxylate cycle. Propionate is a three-carbon compound that can enter the glyoxylate cycle through its conversion to malate.\\n\\nPGAM1 (Phosphoglycerate Mutase 1) is an enzyme involved in glycolysis, specifically in the conversion of 3-phosphoglycerate to 2-phosphoglycerate.\\n\\nIn the altered pathway, Tetrahydrosarcinapterin, D-Xylulose 5-phosphate, CO2, L-erythro-3-Methylmalyl-CoA, 2-Phosphoglycolate, 5-Formyl-5,6,7,8-tetrahydromethanopterin, L-Serine, D-Glucono-1,5-lactone 6-phosphate, Glycolate, D-Glucono-1,5-lactone, D-Fructose 6-phosphate, D-Gluconic acid, Glycerone, D-Glycerate, and 4-Hydroxybutyryl-CoA are involved. However, without more specific information, it is difficult to determine exactly how these compounds are altered in the pathway. It is possible that mutations in PCCA or PGAM1, or other genes in these pathways, are causing the alterations.',\n",
       "  'ft:open-mistral-7b:5aebfd1c:20240627:e63dc09f': 'The altered pathway identified is the Glyoxylate and Dicarboxylate Metabolism. The genes PCCA and PGAM1 play crucial roles in this pathway, with PCCA involved in the formation of succinyl-CoA and PGAM1 in the conversion of 2-oxoglutarate to succinate. The compounds, including Tetrahydrosarcinapterin, D-Xylulose 5-phosphate, CO2, and L-erythro-3-Methylmalyl-CoA, are intermediates or end products of this pathway. The alteration in these compounds, such as the increased levels of CO2 and Glycolate, suggests a disruption in the metabolism of these compounds. The presence of 2-Phosphoglycolate, a compound indicating photorespiratory metabolism, also supports the involvement of the Glyoxylate and Dicarboxylate Metabolism pathway. The alteration in this pathway may contribute to the observed changes in cellular metabolism and energy production.',\n",
       "  'ft:open-mistral-7b:5aebfd1c:20240628:de14ae4a': 'The altered pathway identified in this experiment is the Pentose Phosphate Pathway. This is supported by the alterations in the genes PCCA and PGAM1, which are key enzymes in this pathway. PCCA is involved in the oxidative branch of the pathway, specifically in the conversion of D-ribulose 5-phosphate to D-ribose 5-phosphate. PGAM1 is involved in the non-oxidative branch, catalyzing the conversion of 1,3-bisphosphoglycerate to glyceraldehyde 3-phosphate. The presence of metabolites such as D-Xylulose 5-phosphate, D-Fructose 6-phosphate, D-Gluconic acid, and D-Glucono-1,5-lactone, which are intermediates in the Pentose Phosphate Pathway, further supports this finding. The alteration in Tetrahydrosarcinapterin, a cofactor in folate metabolism, and L-Serine, which can be synthesized via the Pentose Phosphate Pathway, also supports this conclusion. Therefore, the alterations in these genes and compounds suggest that the Pentose Phosphate Pathway is significantly involved in the condition under study.',\n",
       "  'mistral-small-latest': 'The altered pathway involves the genes PCCA and PGAM1, and the compounds listed. The PCCA gene encodes propionyl-CoA carboxylase alpha chain, which is involved in the propanoate metabolism pathway. The PGAM1 gene encodes phosphoglycerate mutase 1, which is involved in the glycolysis/gluconeogenesis pathway.\\n\\nThe altered pathway starts with the compound Tetrahydrosarcinapterin, which is involved in the propanoate metabolism pathway. It is converted to 5-Formyl-5,6,7,8-tetrahydromethanopterin by the action of PCCA. This compound is then involved in the glycolysis/gluconeogenesis pathway, where it is converted to D-Xylulose 5-phosphate by PGAM1. This compound is further converted to D-Fructose 6-phosphate, D-Glucono-1,5-lactone 6-phosphate, D-Glucono-1,5-lactone, D-Gluconic acid, Glycerone, D-Glycerate, and finally CO2.\\n\\nThe compound L-erythro-3-Methylmalyl-CoA, which is a product of the propanoate metabolism pathway, is also involved in the altered pathway. It is converted to 4-Hydroxybutyryl-CoA, which is then involved in the butanoate metabolism pathway. Additionally, the compound 2-Phosphoglycolate is also involved in the altered pathway, but its exact role and conversion is not clear from the given information.'}}"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "entries[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "291b79ca-8dcb-4c70-9638-18bedfed91f2",
   "metadata": {},
   "source": [
    "## Scoring of the models : LLM as a judge\n",
    "\n",
    "In order to score models performance on test dataset, I used mistral-large-latest model together with a list of metrics and rubrics. Large model was instructed to send back a json file containing the scores. Following metrics were used :"
   ]
  },
  {
   "cell_type": "raw",
   "id": "d853d673-cc8b-4e04-87f3-ddb75ea44d9f",
   "metadata": {},
   "source": [
    "- metric: accuracy\n",
    "  rubrics: |\n",
    "    Score 1: The pathway found in output is not even a metabolic pathway.\n",
    "    Score 2: The pathway found in output is a metabolic pathway.\n",
    "    Score 3: The pathway found in output is clearly stated and almost identical to the target pathway (ex: synthesis instead of degradation of the same compound).\n",
    "    Score 4: The pathway found in output is clearly stated and exactly the same as the target pathway (perfect match).\n",
    "\n",
    "- metric: informativity\n",
    "  rubrics: |\n",
    "    Score 1: No or incomplete citation of genes and compounds, or no justification\n",
    "    Score 2: Genes and compounds are cited but are not properly used to justify pathway choice, useless information\n",
    "    Score 3: Information about genes and compounds are exact and used properly to justify pathway choice in output.\n",
    "\n",
    "- metric: format\n",
    "  rubrics: |\n",
    "    Score 1: Output is longer than 200 words and does not include a metabolic pathway.\n",
    "    Score 2: Output is longer than 200 words or does not include a metabolic pathway.\n",
    "    Score 3: Output is shorter than 200 words and include one clearly named metabolic pathway."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "77b354c9-f7d6-4a94-bd14-710e05e1705c",
   "metadata": {},
   "source": [
    "Scoring was performed using [scoring.py](../scoring.py) script."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4621e77f-0379-4ffc-80ea-0a2c7a86aa21",
   "metadata": {},
   "source": [
    "## Visualisation of the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "26ed86a5-536c-492e-b84c-499b23ec6177",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       LLM  accuracy  informativity  format\n",
      "0    LLM 1         2              3       2\n",
      "1    LLM 2         2              3       3\n",
      "2    LLM 3         2              3       3\n",
      "3    LLM 4         2              3       3\n",
      "4    LLM 1         2              3       3\n",
      "..     ...       ...            ...     ...\n",
      "795  LLM 4         3              3       3\n",
      "796  LLM 1         2              3       2\n",
      "797  LLM 2         2              3       3\n",
      "798  LLM 3         2              3       3\n",
      "799  LLM 4         2              3       3\n",
      "\n",
      "[800 rows x 4 columns]\n"
     ]
    }
   ],
   "source": [
    "# Initialize a list to store the data\n",
    "data_list = []\n",
    "\n",
    "# Read the JSONL file\n",
    "with open('../test/test_eval.jsonl', 'r') as file:\n",
    "    for line in file:\n",
    "        data = json.loads(line)\n",
    "        for llm, scores in data.items():\n",
    "            scores['LLM'] = llm\n",
    "            data_list.append(scores)\n",
    "\n",
    "# Create the DataFrame\n",
    "df = pd.DataFrame(data_list)\n",
    "\n",
    "# Reorder columns\n",
    "df = df[['LLM', 'accuracy', 'informativity', 'format']]\n",
    "\n",
    "# Display the DataFrame\n",
    "print(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "e32bc9c1-839e-449b-a241-2fe4b7a8c7ee",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.replace('LLM 1', 'mistral-7b')\n",
    "df = df.replace('LLM 2', 'fine-tune-1')\n",
    "df = df.replace('LLM 3', 'fine-tune-2')\n",
    "df = df.replace('LLM 4', 'mistral-small')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "f149fc16-ce49-4336-b658-1b24e4a2bdc7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             LLM  accuracy  informativity  format\n",
      "0    fine-tune-1       565            561     595\n",
      "1    fine-tune-2       553            579     598\n",
      "2     mistral-7b       480            529     524\n",
      "3  mistral-small       537            540     585\n"
     ]
    }
   ],
   "source": [
    "sums = df.groupby(by=df[\"LLM\"]).sum()\n",
    "sums = sums.reset_index()\n",
    "print(sums)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "792136d3-fcec-41df-8c9d-2f858b8a93b2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x1000 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10, 10))\n",
    "\n",
    "metrics = ['accuracy', 'informativity', 'format']\n",
    "\n",
    "for i, metric in enumerate(metrics):\n",
    "    plt.subplot(3, 1, i+1)\n",
    "    sns.barplot(x=metric, y='LLM', data=sums, legend=False)\n",
    "    plt.title(f'{metric.capitalize()}')\n",
    "    plt.xlabel('')\n",
    "    plt.ylabel('')\n",
    "\n",
    "plt.savefig('metrics_comparison.png', dpi=300)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "8029e056-05a3-4384-a682-6dccc42ac610",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_noft2 = df.loc[df['LLM'] != 'fine-tune-2']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "bf21d889-e1c8-41ee-986c-eb6d125667ef",
   "metadata": {},
   "outputs": [
    {
     "ename": "ValueError",
     "evalue": "cannot insert LLM, already exists",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32m~\\AppData\\Local\\Temp\\ipykernel_4220\\303103801.py\u001b[0m in \u001b[0;36m?\u001b[1;34m()\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[0msums\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mdf_noft2\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mgroupby\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mby\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mdf\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m\"LLM\"\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msum\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 2\u001b[1;33m \u001b[0msums\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msums\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mreset_index\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      3\u001b[0m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msums\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\WORKER\\Documents\\ProjetsPerso\\HackathonMistral\\metabo-assistant\\.venv\\Lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36m?\u001b[1;34m(self, level, drop, inplace, col_level, col_fill, allow_duplicates, names)\u001b[0m\n\u001b[0;32m   6468\u001b[0m                     level_values = algorithms.take(\n\u001b[0;32m   6469\u001b[0m                         \u001b[0mlevel_values\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlab\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mallow_fill\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfill_value\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mlev\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_na_value\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   6470\u001b[0m                     \u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   6471\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 6472\u001b[1;33m                 new_obj.insert(\n\u001b[0m\u001b[0;32m   6473\u001b[0m                     \u001b[1;36m0\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   6474\u001b[0m                     \u001b[0mname\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   6475\u001b[0m                     \u001b[0mlevel_values\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mD:\\WORKER\\Documents\\ProjetsPerso\\HackathonMistral\\metabo-assistant\\.venv\\Lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36m?\u001b[1;34m(self, loc, column, value, allow_duplicates)\u001b[0m\n\u001b[0;32m   5154\u001b[0m                 \u001b[1;34m\"'self.flags.allows_duplicate_labels' is False.\"\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   5155\u001b[0m             \u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   5156\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0mallow_duplicates\u001b[0m \u001b[1;32mand\u001b[0m \u001b[0mcolumn\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   5157\u001b[0m             \u001b[1;31m# Should this be a different kind of error??\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 5158\u001b[1;33m             \u001b[1;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33mf\"\u001b[0m\u001b[1;33mcannot insert \u001b[0m\u001b[1;33m{\u001b[0m\u001b[0mcolumn\u001b[0m\u001b[1;33m}\u001b[0m\u001b[1;33m, already exists\u001b[0m\u001b[1;33m\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   5159\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0mis_integer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mloc\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   5160\u001b[0m             \u001b[1;32mraise\u001b[0m \u001b[0mTypeError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"loc must be int\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   5161\u001b[0m         \u001b[1;31m# convert non stdlib ints to satisfy typing checks\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mValueError\u001b[0m: cannot insert LLM, already exists"
     ]
    }
   ],
   "source": [
    "sums = df_noft2.groupby(by=df[\"LLM\"]).sum()\n",
    "sums = sums.reset_index()\n",
    "print(sums)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "682872d6-579e-4679-9a39-f0435900e628",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 5))\n",
    "sums_melted = sums.melt(id_vars=['LLM'], var_name='Metric', value_name='Score')\n",
    "sns.barplot(x='LLM', y='Score', hue='Metric', data=sums_melted)\n",
    "plt.title('Metrics per LLM', fontsize=14)\n",
    "plt.xlabel('LLM', fontsize=12)\n",
    "plt.ylabel('Score', fontsize=12)\n",
    "plt.legend(title='Metric', title_fontsize='10', fontsize='8')\n",
    "plt.xticks(fontsize=10)\n",
    "plt.yticks(fontsize=10)\n",
    "plt.tight_layout()\n",
    "plt.savefig('combined_graph.png', dpi=300, bbox_inches='tight')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "8a044026-8c29-4a89-bd95-baf0b5cb56c7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 5))\n",
    "sums_melted = sums.melt(id_vars=['LLM'], var_name='Metric', value_name='Score')\n",
    "sns.barplot(x='LLM', y='Score', hue='Metric', data=sums_melted)\n",
    "plt.title('Metrics per LLM', fontsize=14)\n",
    "plt.xlabel('LLM', fontsize=12)\n",
    "plt.ylabel('Score', fontsize=12)\n",
    "plt.legend(title='Metric', title_fontsize='10', fontsize='8')\n",
    "plt.xticks(fontsize=10)\n",
    "plt.yticks(fontsize=10)\n",
    "plt.tight_layout()\n",
    "plt.savefig('combined_graph.png', dpi=300, bbox_inches='tight')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "13f1b265-5c14-4abf-8467-57efd9de9b63",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 500x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(5, 4))\n",
    "sns.histplot(df,\n",
    "    x=\"LLM\", hue=\"accuracy\",\n",
    "    multiple=\"stack\")\n",
    "plt.title('Accuracy values per LLM', fontsize=14)\n",
    "plt.savefig('accuracy_values.png', dpi=300, bbox_inches='tight')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "151e817b-0871-47ee-9e1e-3082b9135ce6",
   "metadata": {},
   "source": [
    "The fine tuned model clearly have more \"exact match\" answers than original model mistral-7b.\n",
    "However, they also seem to have more scores of 1, associated with an answer not containing a metabolic pathway.\n",
    "I then checked the answers corresponding to these scores of 1 (see below). \n",
    "It seems that they mostly mention metabolic pathway, and sometimes it is even the right metabolic pathway.\n",
    "Likely there was a scoring problem with these answers, highlighting a need for a manual scoring.\n",
    "Considering this represents 800 answers to score, it would be interesting to first implement a scoring interface (future perspective)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "f38603d5-e2ff-4f5f-bf07-b10204f16ccd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Line 4, target pathway Valine, leucine and isoleucine degradation, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Butanoate metabolism. This is supported by the alteration of genes associated with this pathway, such as AACS and HMGCS1, which are involved in the biosynthesis of fatty acids. Additionally, the presence of compounds like 3-Methyl-1-hydroxybutyl-ThPP and L-Leucine, which are intermediates in this pathway, further strengthens this conclusion. The metabolites 2-Methylpropanoyl-CoA and 2-Methylbut-2-enoyl-CoA are also indicative of the involvement of this pathway, as they are intermediates in the butanoate metabolism. The compound [Dihydrolipoyllysine-residue (2-methylpropanoyl)transferase] S-(3-methylbutanoyl)dihydrolipoyllysine also suggests the involvement of this pathway, as it is a key enzyme in the Butanoate metabolism. Therefore, the Butanoate metabolism pathway is the likely altered pathway.\n",
      "\n",
      "Line 9, target pathway Mucin type O-glycan biosynthesis, model fine-tune-2:\n",
      "The altered pathway identified in the experiment is the Glycan Biosynthesis - N-Glycan pathway. This is supported by the alterations observed in multiple genes and compounds associated with this pathway. The genes GALNT3, GCNT1, GALNT8, GALNT11, B3GNT3, GALNT1, GALNT5, ST6GALNAC3, ST6GALNAC4, ST3GAL2, POC1B-GALNT4, GALNT9, GALNT10, GALNT16, C1GALT1C1, and GALNT6 are all involved in the initiation, elongation, and modification of N-glycan chains. The compounds [Protein]-L-serine, UDP-N-acetyl-D-galactosamine, Tn antigen, Sialyl-Tn antigen, and T antigen are intermediates or products of this pathway. Therefore, the alterations in these genes and compounds suggest that the Glycan Biosynthesis - N-Glycan pathway is involved in the condition under study.\n",
      "\n",
      "Line 11, target pathway Glyoxylate and dicarboxylate metabolism, model fine-tune-2:\n",
      "The altered pathway identified in the experiment is the Oxidative pentose phosphate pathway. This pathway was altered due to the change in the genes AGXT and AMT, which are involved in the metabolism of amino acids, particularly aspartate and asparagine. These genes, along with the observed compounds such as Oxaloacetate, Glycine, and Succinate, suggest a disruption in the metabolism of these amino acids. The presence of Hydrogen peroxide and 2-Hydroxy-3-oxopropanoate further indicate oxidative stress, which could be a result of the alteration in this pathway. The pathway is directly related to the pentose phosphate pathway, which plays a role in the production of NADPH, a key antioxidant. The altered pathway consequently leads to the accumulation of Oxalate, a compound linked to this pathway, which may contribute to the observed condition.\n",
      "\n",
      "Line 15, target pathway Alanine, aspartate and glutamate metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Aminoacyl-tRNA biosynthesis. This pathway was identified due to the alteration of several genes and compounds directly involved in this process. The genes GPT, NIT2, GLS2, ALDH5A1, ADSS1, IL4I1, ASNS, RIMKLA, ASPA, GLUD2, ASS1, FOLH1, PPAT, ALDH4A1, ASRGL1, and ADSS2 are all associated with amino acid metabolism and tRNA synthesis. The alterations in these genes suggest that the amino acid synthesis pathway, specifically the aminoacyl-tRNA biosynthesis, is affected. The compounds L-Glutamine, N-Acetylaspartylglutamate, 5-Phosphoribosylamine, L-Asparagine, N-(L-Arginino)succinate, Succinate, L-Glutamate, N-Acetylaspartylglutamylglutamate, Pyruvate, 2-Oxoglutarate, L-Alanine, D-Glucosamine 6-phosphate, Fumarate, Citrate, Succinate semialdehyde, N6-(1,2-Dicarboxyethyl)-AMP, D-Aspartate, N-Acetyl-L-aspartate, N-Carbamoyl-L-aspartate, Ammonia, Carbamoyl phosphate, beta-Citryl-L-glutamate, and 4-Aminobutanoate are intermediates or products of this pathway, further supporting its involvement.\n",
      "\n",
      "Line 20, target pathway N-Glycan biosynthesis, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is Glycosphingolipid biosynthesis - lacto and neolacto series. This conclusion is supported by the alterations observed in several genes involved in this pathway, including ST6GAL1, B4GALT1, ST6GAL2, and MGAT2. These genes are known to be involved in the synthesis of glycosphingolipids, which are a type of lipid molecule that play a crucial role in cell-cell recognition and signaling. The alterations in compounds such as Dolichol, Dolichyl phosphate D-mannose, and Polyprenol, which are intermediates in glycosphingolipid biosynthesis, further support this conclusion. The identification of TUSC3, a gene known to regulate glycosphingolipid metabolism, further strengthens this finding. Therefore, the alterations observed in this experiment suggest a disruption in the Glycosphingolipid biosynthesis - lacto and neolacto series pathway.\n",
      "\n",
      "Line 40, target pathway beta-Alanine metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Urea cycle. The Urea cycle is a process that mainly takes place in the liver and kidneys, where waste nitrogen is converted into urea and excreted from the body. The genes AOC2, AOC3, and ACOX1 are involved in the synthesis of malonate, a compound that is a key intermediate in the ureagenesis pathway. ALDH3B2 is involved in the metabolism of 3-Ureidopropionate, a compound produced during the Urea cycle. HIBCH is involved in the metabolism of L-Histidine, which is an amino acid that is a precursor for urea synthesis. The alteration of these genes and compounds suggests that the Urea cycle is dysregulated, which could lead to the accumulation of ammonia in the body, a condition known as Hyperammonemia.\n",
      "\n",
      "Line 58, target pathway Ascorbate and aldarate metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Ascorbate and Alcohol Metabolism. This conclusion is supported by the alterations observed in genes and compounds associated with this pathway. The genes UGT1A3, UGT2B10, GUSB, KL, UGT2B4, UGT2B15, UGT1A4, UGT1A7, UGT2B11 are all involved in the metabolism of ascorbate, a vitamin C derivative. The presence of compounds such as (4R,5S)-4,5,6-Trihydroxy-2,3-dioxohexanoate, Ascorbate, Monodehydroascorbate, L-Arabinose, L-Ribulose 5-phosphate, 2-Dehydro-3-deoxy-D-glucarate, D-Tagaturonate, L-Galactonate, 2,5-Dioxopentanoate, L-Galactose, Threonate, 2-Dehydro-3-deoxy-L-arabinonate, D-Glucuronolactone, D-Arabinonate, L-Xylonate, Dehydro-D-arabinono-1,4-lactone, L-Ascorbate 6-phosphate, D-Glucarate, Dehydroascorbate, GDP-mannose, beta-L-Galactose 1-phosphate, L-Gulose 1-phosphate, D-Arabinose, Pyruvate, further supports this conclusion as these compounds are intermediates or end products of this pathway. The observed alterations in these genes and compounds suggest a disruption in the Ascorbate and Alcohol Metabolism pathway.\n",
      "\n",
      "Line 64, target pathway Glycosaminoglycan biosynthesis - heparan sulfate / heparin, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the N-Glycan Biosynthesis pathway. This conclusion is supported by the alterations observed in multiple genes and compounds directly involved in this pathway. Genes such as EXTL1, NDST2, NDST4, B3GALT6, HS6ST2, EXTL2, HS6ST1, GLCE, EXT2, HS3ST3B1, B4GALT7, HS3ST3A1, and NDST3 are all known to play crucial roles in the N-Glycan Biosynthesis pathway. The presence of compounds like Acetate, Adenosine 3',5'-bisphosphate, [Protein]-L-serine, and 3'-Phosphoadenylyl sulfate further strengthens this finding, as they are intermediates or products in this pathway. Therefore, the alterations in these genes and compounds suggest that the N-Glycan Biosynthesis pathway is significantly impacted in the condition under investigation.\n",
      "\n",
      "Line 72, target pathway Pyrimidine metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Purine Metabolism. This conclusion is supported by the alterations observed in several genes and compounds directly involved in this pathway. The genes RRM2, NT5C3B, NT5C1A, UPP2, and NT5DC4 are all involved in purine nucleotide metabolism, specifically in the synthesis, degradation, and salvage of nucleotides. The compounds identified, such as Methylmalonate, UDP, Deoxycytidine, CMP, 2',3'-Cyclic UMP, Urea, and Thymine, are all intermediates or products in the Purine Metabolism pathway. Furthermore, the presence of compounds like 2'-Deoxy-5-hydroxymethylcytidine-5'-triphosphate, Pseudouridine 5'-phosphate, and 2'-Deoxy-5-hydroxymethylcytidine-5'-diphosphate, which are involved in RNA modification, suggest that the alterations in this pathway may have implications for RNA metabolism as well. Therefore, the Purine Metabolism pathway is strongly implicated in the observed condition.\n",
      "\n",
      "Line 89, target pathway Oxidative phosphorylation, model fine-tune-2:\n",
      "The altered pathway identified is the Citrate Cycle (TCA cycle), also known as the Krebs cycle. This conclusion is supported by the altered genes and compounds. Genes associated with electron transport chain complexes I, III, IV, and V, such as COX7C, COX17, NDUFAB1, COX8A, UQCRFS1, ATP6V1H, and COX15, are involved in the electron transport chain, which is a key component of the TCA cycle. Genes like SDHD, SDHB, and MT-ND6 are directly involved in the TCA cycle, indicating its involvement. The alterations in compounds such as Succinate, Fumarate, and Malate, intermediates in the TCA cycle, further support this conclusion. Additionally, the presence of NAD+ and NADH, which are key cofactors in the TCA cycle, further substantiates this finding.\n",
      "\n",
      "Line 97, target pathway Lysine degradation, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Tryptophan metabolism pathway. This conclusion is supported by the alterations observed in several genes and compounds associated with this pathway. The genes ALDH3A2, SETDB2, ALDH1B1, DLD, ACAT2, SETDB1, SMYD1, ALDH2, EZH1, SUV39H1, SETMAR, KMT5C, SETD7, AASS, SMYD3, KMT5B, SETD1A, HADHA, SUV39H2, PRDM7, and TMLHE are all involved in various steps of tryptophan metabolism. Similarly, the compounds 5-Acetamidopentanoate, Carnitine, L-2-Aminoadipate, (2R,5S)-2,5-Diaminohexanoate, Acetyl-CoA, (S)-3-Hydroxybutanoyl-CoA, Acetoacetate, (S)-2,3,4,5-Tetrahydropyridine-2-carboxylate, D-Lysine, are intermediates or products of this pathway. Therefore, the alterations in these genes and compounds suggest a disruption in the Tryptophan metabolism pathway.\n",
      "\n",
      "Line 109, target pathway Porphyrin metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Biosynthesis of Bacteriochlorophyll. This is evidenced by the alteration of several genes and compounds directly involved in this process. UGT1A3, EPRS1, HEPHL1, ALAS1, and HEPH are involved in heme biosynthesis, a precursor of bacteriochlorophyll. HMBS and HMBS, involved in folate metabolism, are also essential in the synthesis of bacteriochlorophyll. The altered compounds, including various forms of chlorophyll and porphyrins, are intermediates in this process, further supporting the identification of the Bacteriochlorophyll biosynthesis pathway.\n",
      "\n",
      "Line 129, target pathway Lipoic acid metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Lysine Degradation pathway. This conclusion is supported by the altered genes BCKDHB, BCKDHA, DLD, DHTKD1, and the compounds Lipoylprotein, 3-Methylbutanoyl-CoA, 2-Methylpropanoyl-CoA, Enzyme N6-(dihydrolipoyl)lysine, (S)-3-Methyl-2-oxopentanoic acid, Octanoyl-CoA, [Protein]-S8-aminomethyldihydrolipoyllysine, and Enzyme N6-(lipoyl)lysine. These genes and compounds are directly involved in the Lysine Degradation pathway. The altered genes encode enzymes that catalyze reactions in this pathway, and the altered compounds are intermediates or products of this pathway. Therefore, the alterations in these genes and compounds suggest that the Lysine Degradation pathway is significantly affected in the condition under study.\n",
      "\n",
      "Line 131, target pathway Pyrimidine metabolism, model fine-tune-2:\n",
      "The altered metabolic pathway identified in the experiment is the Purine Metabolism Pathway. This conclusion is supported by the alterations observed in the genes and compounds involved in this pathway. The affected genes, such as NT5C3B, NME1-NME2, NME2, CTPS1, NT5E, TK1, ENTPD6, UCKL1, NME3, DUT, ASMTL, ENTPD1, HDDC2, DCTPP1, UPP1, DHODH, TYMP, and NUDT2, are all known to be involved in the Purine Metabolism Pathway. The altered compounds, Cytosine, 2',3'-Cyclic CMP, and dUMP, are intermediates or products of this pathway. Therefore, the alterations in these genes and compounds indicate that the Purine Metabolism Pathway is significantly affected in the condition under study.\n",
      "\n",
      "Line 133, target pathway Amino sugar and nucleotide sugar metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Glycosaminoglycan biosynthesis - chondroitin sulfate/dermatan sulfate. The genes and compounds involved in this pathway are present in the experiment, including UDP-sulfoquinovose, Chitin, UDP-2-acetamido-4-amino-2,4,6-trideoxy-alpha-D-glucose, UDP-D-xylose, UDP-L-arabinose, UDP-alpha-D-galactose, UDP-N-acetyl-D-mannosamine, L-Fucose 1-phosphate, alpha-D-Arabinopyranose 1-phosphate, GDP-4-acetamido-4,6-dideoxy-alpha-D-mannose, D-Mannose 1-phosphate, D-Arabinose, GDP-D-arabinopyranose, N-Acetyl-D-glucosamine, N-Acetylneuraminate 9-phosphate, GDP-D-mannuronate, Chitosan, D-Galactosamine 6-phosphate, and D-Glucose. These are all key components of the Glycosaminoglycan biosynthesis pathway, which is responsible for the synthesis of chondroitin sulfate and dermatan sulfate, both of which are important components of the extracellular matrix. The alteration of this pathway suggests that it may be involved in the condition under investigation.\n",
      "\n",
      "Line 139, target pathway Mucin type O-glycan biosynthesis, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the N-Glycan Biosynthesis Pathway. The N-Glycan Biosynthesis Pathway is a process by which proteins are added glycan chains, specifically N-linked glycans, to their amino acid residues. This pathway is crucial for protein folding, protein-protein interactions, and protein targeting. The alterations in several genes (GCNT1, GALNT3, B3GNT6, ST3GAL1, ST6GALNAC4, GALNT13, B3GNT3, GALNT10, GALNTL5, POC1B-GALNT4, GALNT9, GALNTL6, GALNT2, GALNT4, C1GALT1C1, GALNT6, GALNT8, GALNT15, GALNT5) and compounds (UDP-N-acetyl-D-galactosamine, [Protein]-L-serine) involved in this pathway suggest that this pathway is affected. These genes and compounds are all directly involved in the synthesis and modification of N-linked glycans, making it highly likely that this pathway is altered in this experiment.\n",
      "\n",
      "Line 155, target pathway Neomycin, kanamycin and gentamicin biosynthesis, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the \"Galactose metabolism\" pathway. This conclusion is supported by the alterations observed in the genes HKDC1, GCK, and HK1. HKDC1 encodes glucose-6-phosphate dehydrogenase, which is a key enzyme in the pentose phosphate pathway, a branch of the galactose metabolism pathway. GCK, on the other hand, encodes glucokinase, a regulator of carbohydrate metabolism that is also involved in galactose metabolism. Lastly, HK1 is the first enzyme of the hexokinase pathway, a major pathway in galactose metabolism. The alterations in the compounds 4'-Oxonebramine, Kanamycin B, and 2'-Dehydrokanamycin A further support this conclusion, as these compounds are known to affect the galactose metabolism pathway. Therefore, the alterations in these genes and compounds suggest that the Galactose metabolism pathway is a likely candidate for the altered pathway.\n",
      "\n",
      "Line 167, target pathway 2-Oxocarboxylic acid metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Valine, Leucine, and Isoleucine biosynthesis pathway. This is supported by the changes observed in several genes and compounds directly involved in this pathway. Genes such as BCAT2, GOT2, IDH3B, IDH3G, GOT1L1, and GOT1 are key players in this pathway. The presence of intermediates like 2-Oxo-10-methylthiodecanoic acid, 2-Oxo-9-methylthiononanoic acid, and 2-Oxo-7-methylthioheptanoic acid is further evidence. These compounds are directly involved in the synthesis of valine, leucine, and isoleucine. The presence of dihomomethionine, a compound derived from these amino acids, further supports this conclusion. The alteration in this pathway could be a response to environmental stress, nutrient availability, or disease conditions.\n",
      "\n",
      "Line 173, target pathway Taurine and hypotaurine metabolism, model fine-tune-2:\n",
      "The altered pathway identified in the experiment is the Cysteine and Methionine Metabolism. This conclusion is supported by the changes observed in several genes and compounds directly involved in this pathway. The genes BAAT, GGT1, GGT5, GGT6, and GGT7 are all involved in the metabolism of glutamate, a key intermediate in this pathway. Similarly, the genes GAD1 and GAD2 are involved in the synthesis of GABA, a derivative of glutamate. The genes FMO2, FMO3, and FMO5 are involved in the methylation of amines, which is a crucial step in the metabolism of cysteine and methionine. The altered compounds, such as L-Glutamate, Hypotaurine, L-Cysteine, L-Alanine, and N-Phosphotaurocyamine, are all intermediates or products of this pathway. Therefore, the changes in these genes and compounds suggest that the Cysteine and Methionine Metabolism pathway is significantly altered in the condition under investigation.\n",
      "\n",
      "Line 186, target pathway Carbon metabolism, model fine-tune-2:\n",
      "The altered metabolic pathway is Propanoate metabolism. This is supported by the altered genes and compounds identified in the analysis. The genes PHGDH, ENO1, ME2, HK2, DLD, SUCLG2, PGD, IDH2, SDHB, GOT1, TKTL1, OGDHL, SHMT2, ALDH6A1, RPEL1, AGXT, HK3, ACSS1, PGAM2, ECHS1, PSAT1, are all involved in the metabolism of propanoate. The altered compounds, such as Formate, Trimethylamine, and 2-Oxoglutarate, are intermediates or products of this pathway. Therefore, the alterations in these genes and compounds suggest that the Propanoate metabolism pathway is involved in the condition under study.\n",
      "\n",
      "Line 189, target pathway Carbon metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Propanoate metabolism pathway. This is supported by the changes observed in multiple genes and compounds associated with this pathway. The genes PGK1, GOT2, PFKP, HK2, PRPS1, and TALDO1 are all involved in the conversion of pyruvate to propanoate, which is the initial step in this pathway. The compounds Propanoyl-CoA, 3-Hydroxypropionyl-CoA, Glyoxylate, 3-Hydroxypropanoate, Acetate, and 2-Oxoglutarate are intermediates in this pathway, further supporting its involvement. Additionally, the presence of compounds such as Citrate, Succinate semialdehyde, L-Serine, L-threo-3-Methylaspartate, and Formaldehyde suggest that other routes of propanoate metabolism are also affected. The changes in CO, Methane, and L-Cysteine further support this finding, as these compounds are products of propanoate metabolism. Therefore, the alteration in the Propanoate metabolism pathway is strongly indicated by the changes observed in this experiment.\n",
      "\n",
      "Line 190, target pathway Carbon metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Glycolysis/Gluconeogenesis pathway. This pathway was identified due to the alteration in the genes SUCLG1, SHMT2, ME1, CPS1, PGAM4, and PSPH, which are all involved in this pathway. SUCLG1 and PGAM4 are involved in the conversion of pyruvate to acetyl-CoA, a key step in glycolysis. SHMT2 and ME1 are involved in the conversion of serine to glycine and the formation of 5,10-Methylenetetrahydrofolate, respectively, both steps in the serine biosynthesis pathway, which is closely linked to the Glycolysis/Gluconeogenesis pathway. The alteration of CPS1, which is involved in the conversion of pyruvate to phosphoenolpyruvate, is also indicative of this pathway. The presence of 5,10-Methylenetetrahydrofolate, alpha-D-Glucose 6-phosphate, and Fumarate, all intermediates in this pathway, further supports this identification. Therefore, the alterations in these genes and compounds suggest that the Glycolysis/Gluconeogenesis pathway is significantly affected in this condition.\n",
      "\n",
      "Line 191, target pathway Carbon metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Glycine, serine and threonine metabolism. This is supported by the involvement of several genes and compounds directly associated with this pathway. HKDC1 and ALDOB are genes involved in the metabolism of glycine, serine, and threonine. The altered metabolites, such as D-Glyceraldehyde, D-Xylulose 5-phosphate, Acetoacetyl-CoA, Acetate, D-Glucono-1,5-lactone 6-phosphate, and D-arabino-Hex-3-ulose 6-phosphate, are intermediates in this pathway. Additionally, the presence of compounds like beta-D-Glucose 6-phosphate, L-Cysteine, and Phosphoenolpyruvate further supports this conclusion, as they are precursors or products in the Glycine, serine and threonine metabolism pathway. The alterations observed in this pathway suggest that it plays a crucial role in the condition under investigation.\n",
      "\n",
      "Line 194, target pathway Carbon metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Butanoate metabolism. This pathway was identified due to the alteration of several key genes and compounds. The genes ADPGK, ALDOB, and RPEL1 are involved in the glycolysis pathway, which converts glucose to pyruvate, a precursor for butanoate metabolism. GCK, a glucose-6-phosphate dehydrogenase, is also involved in this pathway. The presence of compounds such as Acetoacetyl-CoA, Pyruvate, Glycine, beta-D-Glucose 6-phosphate, 3-Phospho-D-glycerate, 2-Phosphoglycolate, and Butanoyl-CoA further support this conclusion, as they are intermediates in the butanoate metabolism pathway. The presence of 5-Methyl-5,6,7,8-tetrahydromethanopterin, a coenzyme in methanogenesis, and Methylamine, a precursor for butanoate, further strengthen this conclusion. The alteration of this pathway suggests a potential role in the metabolism of butanoate, a branched-chain fatty acid.\n",
      "\n",
      "Line 195, target pathway Carbon metabolism, model fine-tune-2:\n",
      "The altered metabolic pathway identified in this experiment is Propanoate metabolism. This conclusion is supported by the altered genes and compounds. The genes ALDH6A1, GLUD2, PDHB, OGDHL, PRPS1L1, PRPS1, CPS1, ACAT1, HAO2, PKLR, GOT1, ADPGK, PGAM1, IDH3A, IDH1, TKT, HAO1, PGP, RGN, and OGDH are all involved in this pathway. They participate in various stages of propanoate metabolism, including its production, degradation, and conversion to other compounds. The altered compounds, such as D-Ribose 5-phosphate, alpha-D-Glucose 6-phosphate, Acetyl phosphate, Coenzyme B, Glycolate, 3-Hydroxypropanoate, Formamide, and (R)-Methylmalonyl-CoA, are intermediates or products of this pathway. Therefore, the changes in these genes and compounds suggest that Propanoate metabolism is significantly affected in the condition under study.\n",
      "\n",
      "Line 196, target pathway Carbon metabolism, model fine-tune-2:\n",
      "The altered pathway identified in this experiment is the Glyoxylate and Dicarboxylate Metabolism. This conclusion is supported by the changes observed in genes and compounds associated with this pathway. ENO4, RPIA, and PGAM1 encode enzymes involved in glycolysis and gluconeogenesis, which are interconnected with the Glyoxylate and Dicarboxylate Metabolism pathway. ACOX3, a key enzyme in fatty acid metabolism, is also connected to this pathway. The presence of compounds such as (2R)-Ethylmalonyl-CoA, L-Aspartate, Succinyl-CoA, and Glycerone, which are intermediates in this pathway, further supports this conclusion. The alteration of the Glyoxylate and Dicarboxylate Metabolism pathway may indicate a shift in carbon metabolism, potentially as a response to stress or nutrient availability.\n",
      "\n",
      "Line 198, target pathway Carbon metabolism, model fine-tune-2:\n",
      "The altered pathway identified is the Butanoate metabolism. This conclusion is supported by several genes and compounds present in the experiment. HK3, IDH1, PDHA2, and GOT1 are all involved in this pathway, as they are enzymes in the citrate cycle, pyruvate metabolism, and amino acid metabolism, which are all interconnected with Butanoate metabolism. The presence of compounds such as 5-Methyl-5,6,7,8-tetrahydromethanopterin, Formylmethanofuran, and Methane, which are involved in methanogenesis, suggest a link to the butanoate metabolism pathway, as this pathway can produce acetyl-CoA, a precursor for methanogenesis. The presence of compounds like 2-Phospho-D-glycerate, Citrate, Pyruvate, and Butanoyl-CoA, which are intermediates in this pathway, further support the conclusion. The alteration in this pathway could potentially explain the observed changes in the experiment.\n"
     ]
    }
   ],
   "source": [
    "modelmap = {'mistral-7b':'open-mistral-7b',\n",
    "            'fine-tune-1':'ft:open-mistral-7b:5aebfd1c:20240627:e63dc09f',\n",
    "            'fine-tune-2':'ft:open-mistral-7b:5aebfd1c:20240628:de14ae4a',\n",
    "            'mistral-small':'mistral-small-latest'}\n",
    "\n",
    "for i in df.loc[(df['accuracy'] == 1) & (df['LLM'] == 'fine-tune-2')].index:\n",
    "    model = df.loc[i, 'LLM']\n",
    "    modelcode = modelmap[model]\n",
    "    answer = entries[i//4][\"output\"][modelcode]\n",
    "    print(f'\\nLine {i//4}, target pathway {entries[i//4][\"target\"]}, model {model}:')\n",
    "    print(answer)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Metabo venv",
   "language": "python",
   "name": "venv"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}