eda.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "3c4cca2e",
   "metadata": {},
   "source": [
    "### Download data files from the CATH database and set up the environment \n",
    "The data we want to use for this project is the latest release of the CATH classification data, namely version s40_v4_3_nonredundant.\n",
    "- 'domain_list_nonredundant_s40_v4_3.list' contains the list of cath ids for each protein domain.\n",
    "- 'sequences_nonredundant_s40_v4_3.list' contains the sequences of protein domains.\n",
    "- 'cath_domain_list.list' has the information about each protein domain incl. class, architecture, superfamily labels and more.\n",
    "- 'pdb_files.tgz' has the pdb files representing the 3D secondary structure of each protein domain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "583a0b7e-bfd2-4f4f-a0bd-dd31dcc9e154",
   "metadata": {},
   "outputs": [],
   "source": [
    "# curl http://download.cathdb.info/cath/releases/all-releases/v4_3_0/non-redundant-data-sets/cath-dataset-nonredundant-S40-v4_3_0.list --output domain_list_nonredundant_s40_v4_3.list\n",
    "#!curl http://download.cathdb.info/cath/releases/all-releases/v4_3_0/non-redundant-data-sets/cath-dataset-nonredundant-S40-v4_3_0.fa --output sequences_nonredundant_s40_v4_3.fa\n",
    "#!curl http://download.cathdb.info/cath/releases/all-releases/v4_3_0/cath-classification-data/cath-domain-list-v4_3_0.txt --output cath_domain_list.list\n",
    "#!curl http://download.cathdb.info/cath/releases/all-releases/v4_3_0/non-redundant-data-sets/cath-dataset-nonredundant-S40-v4_3_0.pdb.tgz --output pdb_files.tgz\n",
    "#!tar -xvzf pdb_files.tgz\n",
    "#!pip install transformers[torch] evaluate datasets requests pandas scikit-learn peft bitsandbytes matplotlib sentencepiece biopython py3dmol accelerate deepspeed\n",
    "#!apt install git-lfs\n",
    "#!pip install wandb\n",
    "#!wandb login"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cc96298a",
   "metadata": {},
   "source": [
    "### Prepare and explore the relevant data with its labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 235,
   "id": "097bcf8e-8d4f-4282-8936-d56a48552129",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "domain_list_nonredundant = pd.read_csv(\n",
    "    \"./domain_list_nonredundant_s40_v4_3.list\", header=None, delimiter=r\"\\s+\"\n",
    ")\n",
    "domain_list_nonredundant.columns = [\"domain_id\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 236,
   "id": "f500e64f-5c6b-4641-99f4-4eccae8731dc",
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_fasta_to_df(file_path):\n",
    "    \"\"\"\n",
    "    Reads a FASTA file and returns a DataFrame with columns for 'Full ID', 'ID', and 'Sequence'.\n",
    "\n",
    "    Parameters:\n",
    "    file_path (str): Path to the FASTA file.\n",
    "\n",
    "    Returns:\n",
    "    pd.DataFrame: DataFrame with the FASTA sequences.\n",
    "    \"\"\"\n",
    "    with open(file_path, \"r\") as file:\n",
    "        lines = file.readlines()\n",
    "\n",
    "    data = []\n",
    "    current_full_id = None\n",
    "    current_id = None\n",
    "    current_seq = []\n",
    "\n",
    "    for line in lines:\n",
    "        line = line.strip()\n",
    "        if line.startswith(\">\"):\n",
    "            if current_full_id is not None:\n",
    "                data.append([current_full_id, current_id, \"\".join(current_seq)])\n",
    "            current_full_id = line[1:]  # Remove the '>'\n",
    "            current_id = current_full_id.split(\"|\")[-1].split(\"/\")[\n",
    "                0\n",
    "            ]  # Extract the ID part\n",
    "            current_seq = []\n",
    "        else:\n",
    "            current_seq.append(line)\n",
    "\n",
    "    # Append the last sequence\n",
    "    if current_full_id is not None:\n",
    "        data.append([current_full_id, current_id, \"\".join(current_seq)])\n",
    "\n",
    "    df = pd.DataFrame(data, columns=[\"Full ID\", \"domain_id\", \"sequences\"])\n",
    "    return df\n",
    "\n",
    "\n",
    "# Example usage\n",
    "file_path = \"./sequences_nonredundant_s40_v4_3.fa\"\n",
    "sequences = read_fasta_to_df(file_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "id": "09012873-5b4b-4068-b48c-0668e198a35b",
   "metadata": {},
   "outputs": [],
   "source": [
    "domain_ids = domain_list_nonredundant[\n",
    "    \"domain_id\"\n",
    "].tolist()  # list of relevant domain ids we want to keep"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 238,
   "id": "8bc8d0b8-6009-4d4b-adb4-5c16959b6d92",
   "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>domain_id</th>\n",
       "      <th>cls</th>\n",
       "      <th>arch</th>\n",
       "      <th>tplgy</th>\n",
       "      <th>superfamily</th>\n",
       "      <th>resolution</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1oaiA00</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1go5A00</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>999.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3frhA01</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>1.20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3friA01</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>1.80</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3b89A01</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>2.60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>500233</th>\n",
       "      <td>4aybQ00</td>\n",
       "      <td>6</td>\n",
       "      <td>20</td>\n",
       "      <td>450</td>\n",
       "      <td>10</td>\n",
       "      <td>3.20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>500234</th>\n",
       "      <td>3hkzY00</td>\n",
       "      <td>6</td>\n",
       "      <td>20</td>\n",
       "      <td>450</td>\n",
       "      <td>10</td>\n",
       "      <td>3.40</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>500235</th>\n",
       "      <td>3hkzZ00</td>\n",
       "      <td>6</td>\n",
       "      <td>20</td>\n",
       "      <td>450</td>\n",
       "      <td>10</td>\n",
       "      <td>3.40</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>500236</th>\n",
       "      <td>3zbeA00</td>\n",
       "      <td>6</td>\n",
       "      <td>20</td>\n",
       "      <td>450</td>\n",
       "      <td>20</td>\n",
       "      <td>999.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>500237</th>\n",
       "      <td>3duzA04</td>\n",
       "      <td>6</td>\n",
       "      <td>20</td>\n",
       "      <td>460</td>\n",
       "      <td>10</td>\n",
       "      <td>2.95</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>500238 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       domain_id  cls  arch  tplgy  superfamily  resolution\n",
       "0        1oaiA00    1    10      8           10        1.00\n",
       "1        1go5A00    1    10      8           10      999.00\n",
       "2        3frhA01    1    10      8           10        1.20\n",
       "3        3friA01    1    10      8           10        1.80\n",
       "4        3b89A01    1    10      8           10        2.60\n",
       "...          ...  ...   ...    ...          ...         ...\n",
       "500233   4aybQ00    6    20    450           10        3.20\n",
       "500234   3hkzY00    6    20    450           10        3.40\n",
       "500235   3hkzZ00    6    20    450           10        3.40\n",
       "500236   3zbeA00    6    20    450           20      999.00\n",
       "500237   3duzA04    6    20    460           10        2.95\n",
       "\n",
       "[500238 rows x 6 columns]"
      ]
     },
     "execution_count": 238,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.read_csv(\n",
    "    \"./cath_domain_list.list\",\n",
    "    sep=\"\\s+\",\n",
    "    header=None,\n",
    "    usecols=[0, 1, 2, 3, 4, 11],\n",
    "    comment=\"#\",\n",
    ")\n",
    "data.columns = [\"domain_id\", \"cls\", \"arch\", \"tplgy\", \"superfamily\", \"resolution\"]\n",
    "\n",
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 239,
   "id": "05d2a104-9774-4602-a731-bd2cbbbe8e19",
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data[data[\"domain_id\"].isin(domain_ids)]\n",
    "data = data.reset_index(drop=True)\n",
    "data = data.dropna()  # drop nan rows"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 240,
   "id": "6eb945d1-0bd5-4765-a09e-c22c475a5537",
   "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>cls</th>\n",
       "      <th>arch</th>\n",
       "      <th>counts</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>4520</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>20</td>\n",
       "      <td>2224</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>30</td>\n",
       "      <td>975</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2</td>\n",
       "      <td>40</td>\n",
       "      <td>1386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2</td>\n",
       "      <td>60</td>\n",
       "      <td>2967</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>3</td>\n",
       "      <td>10</td>\n",
       "      <td>1315</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>3</td>\n",
       "      <td>20</td>\n",
       "      <td>1103</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>3</td>\n",
       "      <td>30</td>\n",
       "      <td>4815</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>3</td>\n",
       "      <td>40</td>\n",
       "      <td>7073</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>3</td>\n",
       "      <td>90</td>\n",
       "      <td>1529</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   cls  arch  counts\n",
       "0    1    10    4520\n",
       "1    1    20    2224\n",
       "2    2    30     975\n",
       "3    2    40    1386\n",
       "4    2    60    2967\n",
       "5    3    10    1315\n",
       "6    3    20    1103\n",
       "7    3    30    4815\n",
       "8    3    40    7073\n",
       "9    3    90    1529"
      ]
     },
     "execution_count": 240,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Group by class and architecture and select groups with more than 700 rows\n",
    "data = data.groupby([\"cls\", \"arch\"]).filter(lambda x: len(x) > 700)\n",
    "\n",
    "data.groupby([\"cls\", \"arch\"]).size().reset_index(name=\"counts\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "id": "db47ddf4-4063-4163-a2ae-07d1fc81087d",
   "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>domain_id</th>\n",
       "      <th>cls</th>\n",
       "      <th>arch</th>\n",
       "      <th>tplgy</th>\n",
       "      <th>superfamily</th>\n",
       "      <th>resolution</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1oaiA00</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>1.00</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3b89A01</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>2.60</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4heoA00</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>1.65</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1oksA00</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>1.80</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4wp2A00</td>\n",
       "      <td>1</td>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>10</td>\n",
       "      <td>1.70</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30768</th>\n",
       "      <td>3fefA00</td>\n",
       "      <td>3</td>\n",
       "      <td>90</td>\n",
       "      <td>1820</td>\n",
       "      <td>10</td>\n",
       "      <td>2.20</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30769</th>\n",
       "      <td>1ej6B00</td>\n",
       "      <td>3</td>\n",
       "      <td>90</td>\n",
       "      <td>1830</td>\n",
       "      <td>10</td>\n",
       "      <td>3.60</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30770</th>\n",
       "      <td>1m1cA00</td>\n",
       "      <td>3</td>\n",
       "      <td>90</td>\n",
       "      <td>1840</td>\n",
       "      <td>10</td>\n",
       "      <td>3.50</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30771</th>\n",
       "      <td>1mukA02</td>\n",
       "      <td>3</td>\n",
       "      <td>90</td>\n",
       "      <td>1850</td>\n",
       "      <td>10</td>\n",
       "      <td>2.50</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30772</th>\n",
       "      <td>2epgA00</td>\n",
       "      <td>3</td>\n",
       "      <td>90</td>\n",
       "      <td>1860</td>\n",
       "      <td>10</td>\n",
       "      <td>2.10</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>27907 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      domain_id  cls  arch  tplgy  superfamily  resolution  label\n",
       "0       1oaiA00    1    10      8           10        1.00      0\n",
       "1       3b89A01    1    10      8           10        2.60      0\n",
       "2       4heoA00    1    10      8           10        1.65      0\n",
       "3       1oksA00    1    10      8           10        1.80      0\n",
       "4       4wp2A00    1    10      8           10        1.70      0\n",
       "...         ...  ...   ...    ...          ...         ...    ...\n",
       "30768   3fefA00    3    90   1820           10        2.20      9\n",
       "30769   1ej6B00    3    90   1830           10        3.60      9\n",
       "30770   1m1cA00    3    90   1840           10        3.50      9\n",
       "30771   1mukA02    3    90   1850           10        2.50      9\n",
       "30772   2epgA00    3    90   1860           10        2.10      9\n",
       "\n",
       "[27907 rows x 7 columns]"
      ]
     },
     "execution_count": 241,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "architecture_names = {\n",
    "    (1, 10): \"Mainly Alpha: Orthogonal Bundle\",\n",
    "    (1, 20): \"Mainly Alpha: Up-down Bundle\",\n",
    "    (2, 30): \"Mainly Beta: Roll\",\n",
    "    (2, 40): \"Mainly Beta: Beta Barrel\",\n",
    "    (2, 60): \"Mainly Beta: Sandwich\",\n",
    "    (3, 10): \"Alpha Beta: Roll\",\n",
    "    (3, 20): \"Alpha Beta: Alpha-Beta Barrel\",\n",
    "    (3, 30): \"Alpha Beta: 2-Layer Sandwich\",\n",
    "    (3, 40): \"Alpha Beta: 3-Layer(aba) Sandwich\",\n",
    "    (3, 90): \"Alpha Beta: Alpha-Beta Complex\",\n",
    "}  # (class, architecture) are the existing labels\n",
    "\n",
    "sorted_keys = sorted(architecture_names.keys())\n",
    "label_index_mapping = {key: index for index, key in enumerate(sorted_keys)}\n",
    "\n",
    "data[\"label\"] = data.apply(\n",
    "    lambda row: label_index_mapping[(row[\"cls\"], row[\"arch\"])], axis=1\n",
    ")\n",
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 242,
   "id": "d6326f79-f0c9-41c4-9489-cee6c0c52c68",
   "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>domain_id</th>\n",
       "      <th>superfamily</th>\n",
       "      <th>label</th>\n",
       "      <th>sequences</th>\n",
       "      <th>resolution</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1oaiA00</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>PTLSPEQQEMLQAFSTQSGMNLEWSQKCLQDNNWDYTRSAQAFTHL...</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3b89A01</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>SLNINDALTSILASKKYRALCPDTVRRILTEEWGRHKSPKQTVEAA...</td>\n",
       "      <td>2.60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4heoA00</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>XVADDASKDVVRTXIRTHIKDRELRSELXDYLNRAETDEEVQEVAN...</td>\n",
       "      <td>1.65</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1oksA00</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>XASRSVIRSIIKSSRLEEDRKRYLXTLLDDIKGANDLAKFHQXLVK...</td>\n",
       "      <td>1.80</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4wp2A00</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "      <td>SNAATKAQLIAEVSRRTGMNVEYSQMXLTGAANWNLELALQSFEQQ...</td>\n",
       "      <td>1.70</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27902</th>\n",
       "      <td>3fefA00</td>\n",
       "      <td>10</td>\n",
       "      <td>9</td>\n",
       "      <td>XSLDQIKIAYIGGGSQGWARSLXSDLSIDERXSGTVALYDLDFEAA...</td>\n",
       "      <td>2.20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27903</th>\n",
       "      <td>1ej6B00</td>\n",
       "      <td>10</td>\n",
       "      <td>9</td>\n",
       "      <td>NKKTAQLLHADTPRLVTWDAGLCTSFKIVPIVPAQVPQDVLAYTFF...</td>\n",
       "      <td>3.60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27904</th>\n",
       "      <td>1m1cA00</td>\n",
       "      <td>10</td>\n",
       "      <td>9</td>\n",
       "      <td>MLRFVTKNSQDKSSDLFSICSDRGTFVAHNRVRTDFKFDNLVFNRV...</td>\n",
       "      <td>3.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27905</th>\n",
       "      <td>1mukA02</td>\n",
       "      <td>10</td>\n",
       "      <td>9</td>\n",
       "      <td>PTETVLQEYTQSPEIKVPIPQKDWTGPIGEIRILKDTTSSIARYLY...</td>\n",
       "      <td>2.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27906</th>\n",
       "      <td>2epgA00</td>\n",
       "      <td>10</td>\n",
       "      <td>9</td>\n",
       "      <td>XARTLFLEEGPXFFEKIAPYTYRIPRQGKXRVDAVFFASKEILKDL...</td>\n",
       "      <td>2.10</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>27907 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      domain_id  superfamily  label  \\\n",
       "0       1oaiA00           10      0   \n",
       "1       3b89A01           10      0   \n",
       "2       4heoA00           10      0   \n",
       "3       1oksA00           10      0   \n",
       "4       4wp2A00           10      0   \n",
       "...         ...          ...    ...   \n",
       "27902   3fefA00           10      9   \n",
       "27903   1ej6B00           10      9   \n",
       "27904   1m1cA00           10      9   \n",
       "27905   1mukA02           10      9   \n",
       "27906   2epgA00           10      9   \n",
       "\n",
       "                                               sequences  resolution  \n",
       "0      PTLSPEQQEMLQAFSTQSGMNLEWSQKCLQDNNWDYTRSAQAFTHL...        1.00  \n",
       "1      SLNINDALTSILASKKYRALCPDTVRRILTEEWGRHKSPKQTVEAA...        2.60  \n",
       "2      XVADDASKDVVRTXIRTHIKDRELRSELXDYLNRAETDEEVQEVAN...        1.65  \n",
       "3      XASRSVIRSIIKSSRLEEDRKRYLXTLLDDIKGANDLAKFHQXLVK...        1.80  \n",
       "4      SNAATKAQLIAEVSRRTGMNVEYSQMXLTGAANWNLELALQSFEQQ...        1.70  \n",
       "...                                                  ...         ...  \n",
       "27902  XSLDQIKIAYIGGGSQGWARSLXSDLSIDERXSGTVALYDLDFEAA...        2.20  \n",
       "27903  NKKTAQLLHADTPRLVTWDAGLCTSFKIVPIVPAQVPQDVLAYTFF...        3.60  \n",
       "27904  MLRFVTKNSQDKSSDLFSICSDRGTFVAHNRVRTDFKFDNLVFNRV...        3.50  \n",
       "27905  PTETVLQEYTQSPEIKVPIPQKDWTGPIGEIRILKDTTSSIARYLY...        2.50  \n",
       "27906  XARTLFLEEGPXFFEKIAPYTYRIPRQGKXRVDAVFFASKEILKDL...        2.10  \n",
       "\n",
       "[27907 rows x 5 columns]"
      ]
     },
     "execution_count": 242,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.merge(\n",
    "    data,\n",
    "    sequences,\n",
    "    on=\"domain_id\",\n",
    ")\n",
    "data = data[[\"domain_id\", \"superfamily\", \"label\", \"sequences\", \"resolution\"]]\n",
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 243,
   "id": "0128b586-d61d-4a69-9f88-e6ab03169e5a",
   "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>label</th>\n",
       "      <th>counts</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>4520</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>2224</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>975</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>1386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>2967</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>5</td>\n",
       "      <td>1315</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>6</td>\n",
       "      <td>1103</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>7</td>\n",
       "      <td>4815</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>8</td>\n",
       "      <td>7073</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>9</td>\n",
       "      <td>1529</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   label  counts\n",
       "0      0    4520\n",
       "1      1    2224\n",
       "2      2     975\n",
       "3      3    1386\n",
       "4      4    2967\n",
       "5      5    1315\n",
       "6      6    1103\n",
       "7      7    4815\n",
       "8      8    7073\n",
       "9      9    1529"
      ]
     },
     "execution_count": 243,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.groupby([\"label\"]).size().reset_index(name=\"counts\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 244,
   "id": "1554c2ec-742f-428c-94ff-df39ef205be2",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_2217/1407899608.py:2: DeprecationWarning: DataFrameGroupBy.apply operated on the grouping columns. This behavior is deprecated, and in a future version of pandas the grouping columns will be excluded from the operation. Either pass `include_groups=False` to exclude the groupings or explicitly select the grouping columns after groupby to silence this warning.\n",
      "  data = data.groupby(['label']).apply(lambda x: x.nlargest(800, 'resolution')).reset_index(drop=True)\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>label</th>\n",
       "      <th>counts</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>5</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>6</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>7</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>8</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>9</td>\n",
       "      <td>800</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   label  counts\n",
       "0      0     800\n",
       "1      1     800\n",
       "2      2     800\n",
       "3      3     800\n",
       "4      4     800\n",
       "5      5     800\n",
       "6      6     800\n",
       "7      7     800\n",
       "8      8     800\n",
       "9      9     800"
      ]
     },
     "execution_count": 244,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# keep 975 examples per label to create a balanced dataset after keeping the ones with highest resolution (high quality experiments as possible)\n",
    "data = (\n",
    "    data.groupby([\"label\"])\n",
    "    .apply(lambda x: x.nlargest(800, \"resolution\"))\n",
    "    .reset_index(drop=True)\n",
    ")\n",
    "\n",
    "data.groupby([\"label\"]).size().reset_index(name=\"counts\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 245,
   "id": "a2cd8b2d-0889-4acb-b8cf-89ed60e85e57",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Intersection of superfamilies: set()\n"
     ]
    }
   ],
   "source": [
    "# Stratified GroupKFold as we need to ensure that the distribution of classes in the training and validation sets are as identical as possible\n",
    "# We also need to ensure that the proteins belonging to one superfamily should be either in the training or in the validation set as we don't want to have some data leakage\n",
    "\n",
    "from sklearn.model_selection import StratifiedGroupKFold\n",
    "\n",
    "sgkf = StratifiedGroupKFold(n_splits=5, shuffle=True, random_state=658)\n",
    "\n",
    "df = data.copy()\n",
    "\n",
    "# Perform the split\n",
    "for train_idx, val_idx in sgkf.split(\n",
    "    df[\"sequences\"], df[\"label\"], groups=df[\"superfamily\"]\n",
    "):\n",
    "    train_df = df.iloc[train_idx][[\"sequences\", \"label\", \"superfamily\", \"domain_id\"]]\n",
    "    val_df = df.iloc[val_idx][[\"sequences\", \"label\", \"superfamily\", \"domain_id\"]]\n",
    "    break  # We only need the first split for train/validation\n",
    "\n",
    "\n",
    "# Verify no intersection\n",
    "train_superfamilies = set(train_df[\"superfamily\"])\n",
    "val_superfamilies = set(val_df[\"superfamily\"])\n",
    "print(\n",
    "    \"\\nIntersection of superfamilies:\",\n",
    "    train_superfamilies.intersection(val_superfamilies),\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 246,
   "id": "e740db19-564d-4e65-b3fc-f5b1a655fdce",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_sequences = train_df[\"sequences\"].tolist()\n",
    "test_sequences = val_df[\"sequences\"].tolist()\n",
    "# train_sequences_3d = train_df[\"seq3d\"].tolist()\n",
    "# test_sequences_3d = val_df[\"seq3d\"].tolist()\n",
    "train_labels = train_df[\"label\"].tolist()\n",
    "test_labels = val_df[\"label\"].tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 247,
   "id": "6bdf584a-b845-4042-8ac2-3d19593086b1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: >"
      ]
     },
     "execution_count": 247,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pd.Series(train_labels).value_counts(sort=True, ascending=False).plot(\n",
    "    kind=\"bar\", backend=\"matplotlib\", figsize=(10, 5)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 248,
   "id": "1b78e156-4988-4741-92b6-5650fd8dc3c8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: >"
      ]
     },
     "execution_count": 248,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pd.Series(test_labels).value_counts(sort=True, ascending=False).plot(\n",
    "    kind=\"bar\", backend=\"matplotlib\", figsize=(10, 5)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 249,
   "id": "475e8631-5960-41a3-aadf-ebb9a5456395",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df.to_csv(\"train.csv\", index=False)\n",
    "val_df.to_csv(\"val.csv\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "15728e59",
   "metadata": {},
   "source": [
    "### Data preparation of results from foldseek"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "id": "42f0f13d-5f35-49a0-9626-2608737abf76",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['onstart.sh', 'ports.log', 'cath_datasets', 'Untitled.ipynb', '.ipynb_checkpoints', 'domain_list_nonredundant_s40_v4_3.list', 'cath_domain_list.list', 'sequences_nonredundant_s40_v4_3.fa', 'train.csv', 'val.csv', 'foldseek', 'queryDB_ss_h', 'queryDB_ss_h.index', 'queryDB_ss_h.dbtype', 'queryDB_ss.fasta']\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "from Bio import SeqIO\n",
    "import os\n",
    "\n",
    "# Change to the desired directory\n",
    "# os.chdir(\"/root\")\n",
    "# Verify the change\n",
    "print(os.listdir(\"./\"))\n",
    "\n",
    "with open(\"./queryDB_ss.fasta\") as fasta_file:  # Will close handle cleanly\n",
    "    identifiers = []\n",
    "    seqs = []\n",
    "    for seq_record in SeqIO.parse(fasta_file, \"fasta\"):  # (generator)\n",
    "        identifiers.append(seq_record.id)\n",
    "        seqs.append(str(seq_record.seq).lower())\n",
    "\n",
    "s1 = pd.Series(identifiers, name=\"ID\")\n",
    "s2 = pd.Series(seqs, name=\"length\")\n",
    "# Gathering Series into a pandas DataFrame and rename index as ID column\n",
    "df = pd.DataFrame(dict(domain_id=s1, seq3d=s2)).set_index([\"domain_id\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "id": "bb22fc94-c79c-4d01-8cef-b57cf3256f62",
   "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>seq3d</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>domain_id</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>12asA00</th>\n",
       "      <td>dvvvllvllvlllvllvvlccvlqvaeeddddqkdwpfflqadcwl...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>132lA00</th>\n",
       "      <td>dadllrvlvllvlppacdvhdhslllsqqlvvpvqaqadkaddpqf...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>153lA00</th>\n",
       "      <td>dqdpqhdlvpddfqaadvvpqvvvpdpdtglvslqvllvvlvvlvv...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>155cA00</th>\n",
       "      <td>dvqacvvqvvcvvvvaqlewqqdvhggdrhhhvndynlfqqqlaqq...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16pkA02</th>\n",
       "      <td>dlvqllpqafdqeeeeeeaqavvvcvvlcvqcllphqeyeyaylnl...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6qulU00</th>\n",
       "      <td>ddpvvlvqffdawdddpvqvvccvpqvktktwtappddfvspqvsc...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6r62A00</th>\n",
       "      <td>daqllvvcvvvpaffeeaeeqdfdqpvllvvllppgqayeyeaaph...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6rxnA00</th>\n",
       "      <td>dwkkaqqppgdiddcvvvvnddpvvddqcdadpppghgpvrigtd</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7mdhA01</th>\n",
       "      <td>dpeaeeeeecclppvvlvpvvcvllavvvhpvhqyeyqyayapvcv...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7mdhA02</th>\n",
       "      <td>dclllqqlllcvlcvllvhdsvqkfqwfwkfdnallidtdqqridg...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>31885 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                       seq3d\n",
       "domain_id                                                   \n",
       "12asA00    dvvvllvllvlllvllvvlccvlqvaeeddddqkdwpfflqadcwl...\n",
       "132lA00    dadllrvlvllvlppacdvhdhslllsqqlvvpvqaqadkaddpqf...\n",
       "153lA00    dqdpqhdlvpddfqaadvvpqvvvpdpdtglvslqvllvvlvvlvv...\n",
       "155cA00    dvqacvvqvvcvvvvaqlewqqdvhggdrhhhvndynlfqqqlaqq...\n",
       "16pkA02    dlvqllpqafdqeeeeeeaqavvvcvvlcvqcllphqeyeyaylnl...\n",
       "...                                                      ...\n",
       "6qulU00    ddpvvlvqffdawdddpvqvvccvpqvktktwtappddfvspqvsc...\n",
       "6r62A00    daqllvvcvvvpaffeeaeeqdfdqpvllvvllppgqayeyeaaph...\n",
       "6rxnA00        dwkkaqqppgdiddcvvvvnddpvvddqcdadpppghgpvrigtd\n",
       "7mdhA01    dpeaeeeeecclppvvlvpvvcvllavvvhpvhqyeyqyayapvcv...\n",
       "7mdhA02    dclllqqlllcvlcvllvhdsvqkfqwfwkfdnallidtdqqridg...\n",
       "\n",
       "[31885 rows x 1 columns]"
      ]
     },
     "execution_count": 196,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "id": "39243189-3ac4-4af3-a501-d8b46587f954",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_csv(\n",
    "    \"./foldseek_seq3d.csv\"\n",
    ")  # contains all the 3D sequences from the queryDB after running foldseek"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 254,
   "id": "b2ca3d05-8f2e-4de9-87e1-73c77beea86b",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df = pd.read_csv(\"./train.csv\")\n",
    "val_df = pd.read_csv(\"./val.csv\")\n",
    "df_foldseek = pd.read_csv(\"./foldseek_seq3d.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 255,
   "id": "049fe611-cb7f-428a-923c-8e034fad201e",
   "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>sequences</th>\n",
       "      <th>label</th>\n",
       "      <th>superfamily</th>\n",
       "      <th>domain_id</th>\n",
       "      <th>seq3d</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>MARIAGINIPDHKHAVIALTSIYGVGKTRSKAILAAAGIAEDVKIS...</td>\n",
       "      <td>0</td>\n",
       "      <td>50</td>\n",
       "      <td>1vs5M01</td>\n",
       "      <td>dclvddddqfdafqlvsvcvfqwddnvrsvvlvvvqvadrpdtlvn...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>MSLVVQEQGSFQHILRLLNTNVDGNIKIVYALTTIKGVGRRYSNLV...</td>\n",
       "      <td>0</td>\n",
       "      <td>50</td>\n",
       "      <td>3u5cS01</td>\n",
       "      <td>dddddddpqfdawdddpngifgwpdlqlcrvvvppprdsvnsvvll...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>MSFVIEKESDFKYIHRILNTNIDGKRITPIALTGIRGIGRRFAYII...</td>\n",
       "      <td>0</td>\n",
       "      <td>50</td>\n",
       "      <td>2xzmM01</td>\n",
       "      <td>dppdddppplfdqwddadpplifhkdwqlcrccvppphhdpvsvvl...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>EKSRVAILKANLRKSPVAKDVDLEFLAKMTNGFSGADLTEICQRAC...</td>\n",
       "      <td>0</td>\n",
       "      <td>60</td>\n",
       "      <td>1ypwA06</td>\n",
       "      <td>dqvqqlvqlvvlpdcvpddsdddcfvrdpdgnvvsnqlscqvvcqv...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>DLWVSGNLISPDLDTVLYLFSDQIDRKRWWGIENDTFGTYERXKEL...</td>\n",
       "      <td>0</td>\n",
       "      <td>240</td>\n",
       "      <td>2ffeA02</td>\n",
       "      <td>dddfpnfdddfvlvlvlcvvlvfacvvvsgagnpfaqvvqvvvvvv...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6638</th>\n",
       "      <td>SPKEILNLTSELLQKCSSPAPGPGKEWEEYVQIRTLVEKIRKKQKG...</td>\n",
       "      <td>9</td>\n",
       "      <td>10</td>\n",
       "      <td>3smtA01</td>\n",
       "      <td>dlvvllvllvvllvllpdddvllvslvvsvvsqvvsqvvddfapdd...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6639</th>\n",
       "      <td>FDNNSHDRVKIKLGVSKSDRLYAXKAEVLARAGIPTSSVFALHFTE...</td>\n",
       "      <td>9</td>\n",
       "      <td>10</td>\n",
       "      <td>3smtA02</td>\n",
       "      <td>dvpdpwdkdkdkdaqdppdpplvlqvlcvvlvhhridiaiqtldpv...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6640</th>\n",
       "      <td>SMERRDGLSGLINMGNTCFMSSILQCLIHNPYFIRHSMSQIHSNNC...</td>\n",
       "      <td>9</td>\n",
       "      <td>10</td>\n",
       "      <td>4fk5A02</td>\n",
       "      <td>dvqqqqldffaappdlcqlprqvlvvqlqdplnlvvlvvcpcvvqf...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6641</th>\n",
       "      <td>MTNPSNSNLQALREELCTPGLDQGHLFEGWPETVDECNERQIALLT...</td>\n",
       "      <td>9</td>\n",
       "      <td>10</td>\n",
       "      <td>3ogzA01</td>\n",
       "      <td>dadqvrlvvvvvvccddlnnqvvlqvphdngpvrgddvsvvqsvvq...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6642</th>\n",
       "      <td>XSLNRHFTVSVFIVCKDKVLLHLHKKAKKXLPLGGHIEVNELPEEA...</td>\n",
       "      <td>9</td>\n",
       "      <td>10</td>\n",
       "      <td>3f6aA00</td>\n",
       "      <td>ddqqekefewedeqfwtkdfqqpvqraggfmdtadvvgdnvvrvqv...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>6643 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                              sequences  label  superfamily  \\\n",
       "0     MARIAGINIPDHKHAVIALTSIYGVGKTRSKAILAAAGIAEDVKIS...      0           50   \n",
       "1     MSLVVQEQGSFQHILRLLNTNVDGNIKIVYALTTIKGVGRRYSNLV...      0           50   \n",
       "2     MSFVIEKESDFKYIHRILNTNIDGKRITPIALTGIRGIGRRFAYII...      0           50   \n",
       "3     EKSRVAILKANLRKSPVAKDVDLEFLAKMTNGFSGADLTEICQRAC...      0           60   \n",
       "4     DLWVSGNLISPDLDTVLYLFSDQIDRKRWWGIENDTFGTYERXKEL...      0          240   \n",
       "...                                                 ...    ...          ...   \n",
       "6638  SPKEILNLTSELLQKCSSPAPGPGKEWEEYVQIRTLVEKIRKKQKG...      9           10   \n",
       "6639  FDNNSHDRVKIKLGVSKSDRLYAXKAEVLARAGIPTSSVFALHFTE...      9           10   \n",
       "6640  SMERRDGLSGLINMGNTCFMSSILQCLIHNPYFIRHSMSQIHSNNC...      9           10   \n",
       "6641  MTNPSNSNLQALREELCTPGLDQGHLFEGWPETVDECNERQIALLT...      9           10   \n",
       "6642  XSLNRHFTVSVFIVCKDKVLLHLHKKAKKXLPLGGHIEVNELPEEA...      9           10   \n",
       "\n",
       "     domain_id                                              seq3d  \n",
       "0      1vs5M01  dclvddddqfdafqlvsvcvfqwddnvrsvvlvvvqvadrpdtlvn...  \n",
       "1      3u5cS01  dddddddpqfdawdddpngifgwpdlqlcrvvvppprdsvnsvvll...  \n",
       "2      2xzmM01  dppdddppplfdqwddadpplifhkdwqlcrccvppphhdpvsvvl...  \n",
       "3      1ypwA06  dqvqqlvqlvvlpdcvpddsdddcfvrdpdgnvvsnqlscqvvcqv...  \n",
       "4      2ffeA02  dddfpnfdddfvlvlvlcvvlvfacvvvsgagnpfaqvvqvvvvvv...  \n",
       "...        ...                                                ...  \n",
       "6638   3smtA01  dlvvllvllvvllvllpdddvllvslvvsvvsqvvsqvvddfapdd...  \n",
       "6639   3smtA02  dvpdpwdkdkdkdaqdppdpplvlqvlcvvlvhhridiaiqtldpv...  \n",
       "6640   4fk5A02  dvqqqqldffaappdlcqlprqvlvvqlqdplnlvvlvvcpcvvqf...  \n",
       "6641   3ogzA01  dadqvrlvvvvvvccddlnnqvvlqvphdngpvrgddvsvvqsvvq...  \n",
       "6642   3f6aA00  ddqqekefewedeqfwtkdfqqpvqraggfmdtadvvgdnvvrvqv...  \n",
       "\n",
       "[6643 rows x 5 columns]"
      ]
     },
     "execution_count": 255,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 252,
   "id": "af40c4ae-e427-4338-b98c-7ba84a2b7795",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df = pd.merge(train_df, df_foldseek, on=\"domain_id\")\n",
    "val_df = pd.merge(val_df, df_foldseek, on=\"domain_id\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 253,
   "id": "e175a862-6c72-4af3-9c99-74d21c11027b",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df.to_csv(\"train.csv\", index=False)\n",
    "val_df.to_csv(\"val.csv\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d6bd412d-4019-4af3-a85b-95e9a5da644c",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}