processing_data.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Step1 Retrieve GO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def retrieve(go_save_dir, task, df):\n",
    "        for id in df['uniprot_id']:\n",
    "            os.system(f\"grep -w {id} datasets/idmapping_select.tab > {go_save_dir}/{task}/{id}.txt\")\n",
    "\n",
    "def integrate(go_save_dir, task, df):\n",
    "    goterms = []\n",
    "    for id in df['uniprot_id']:\n",
    "        f = open(f\"{go_save_dir}/{task}/{id}.txt\", \"r\")\n",
    "        content = f.read()\n",
    "        f.close()\n",
    "        go = content.strip().split(\"\\t\")[6]\n",
    "        goterms.append(go if go != \"\" else \"No goterm\")\n",
    "    df['GO'] = goterms\n",
    "    \n",
    "    if task in ['stability', 'fluorescence', 'remote_homology', 'secondary_structure']:\n",
    "        df = df[['uniprot_id', 'seq', 'GO', 'label']]\n",
    "    elif task == 'pdbbind':\n",
    "        df = df[['uniprot_id', 'seq', 'smiles', 'rdkit_smile', 'GO', 'label', 'dataset_type']]\n",
    "    else:\n",
    "        df = df[['molecule','uniprot_id','seq','GO','label']]\n",
    "    \n",
    "    return df\n",
    "\n",
    "def process_go(dataset_dir, data_pth):\n",
    "    go_save_dir = os.path.join(dataset_dir, 'go')\n",
    "    data_file = data_pth.split(\"/\")[-1]\n",
    "    task, dataset_type = data_file.split(\".\")[0], data_file.split(\".\")[1]\n",
    "    if not os.isdir(os.path.join(go_save_dir, task)):\n",
    "        os.makedirs(os.path.join(go_save_dir, task))\n",
    "        \n",
    "    abbrs = {'stability': 'stab', 'remote_homology': 'remo', 'fluorescence': 'fluo', 'secondary_structure': 'secstruc'}\n",
    "    df = pd.read_csv(data_pth)\n",
    "    if task in ['stability', 'remote_homology', 'fluorescence', 'secondary_structure']:\n",
    "        df.insert(0, 'uniprot_id', [f\"{abbrs[task]}_{dataset_type}_{i}\" for i in range(0, len(df))])\n",
    " \n",
    "    retrieve(go_save_dir, task, df)\n",
    "    new_df = integrate(go_save_dir, task, df)\n",
    "    \n",
    "    if not os.isdir(dataset_dir, f\"new/{task}\"):\n",
    "        os.makedirs(os.path.join(dataset_dir, f\"new/{task}\"))\n",
    "    \n",
    "    new_df.to_csv(os.path.join(dataset_dir, f\"new/{task}/{data_file}\"), index=False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "texts_embs_hub = {\n",
    "    'st': [\"datasets/raw/stability/stability.train.txt\",\n",
    "            \"datasets/raw/stability/stability.valid.txt\",\n",
    "            \"datasets/raw/stability/stability.test.txt\"],\n",
    "    'fl': [\"datasets/raw/fluorescence/fluorescence.train.txt\",\n",
    "            \"datasets/raw/fluorescence/fluorescence.valid.txt\",\n",
    "            \"datasets/raw/fluorescence/fluorescence.test.txt\"],\n",
    "    'rh': [\"datasets/raw/remote_homology/remote_homology.train.txt\",\n",
    "            \"datasets/raw/remote_homology/remote_homology.valid.txt\",\n",
    "            \"datasets/raw/remote_homology/remote_homology.test.txt\"],\n",
    "    'ss': [\"datasets/raw/secondary_structure/secondary_structure.train.txt\",\n",
    "            \"datasets/raw/secondary_structure/secondary_structure.valid.txt\",\n",
    "            \"datasets/raw/secondary_structure/secondary_structure.test.txt\"],\n",
    "    'pb': [\"datasets/raw/pdbbind/pdbbind.train.txt\",\n",
    "           \"datasets/raw/pdbbind/pdbbind.valid.txt\",\n",
    "           \"datasets/raw/pdbbind/pdbbind.test.txt\"],\n",
    "    'ki': [\"datasets/raw/kinase/kinase.train.txt\",\n",
    "           \"datasets/raw/kinase/kinase.test.txt\"], # no vaild\n",
    "}\n",
    "\n",
    "for task in ['st', 'fl', 'rh', 'pb']:\n",
    "    for data_pth in texts_embs_hub[task]:\n",
    "        process_go('datasets/', data_pth)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Step2 Processing label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import os\n",
    "import time\n",
    "import collections\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "texts_embs_hub = {\n",
    "    'st': [\"datasets/new/stability/stability.train.txt\",\n",
    "            \"datasets/new/stability/stability.valid.txt\",\n",
    "            \"datasets/new/stability/stability.test.txt\"],\n",
    "    'fl': [\"datasets/new/fluorescence/fluorescence.train.txt\",\n",
    "            \"datasets/new/fluorescence/fluorescence.valid.txt\",\n",
    "            \"datasets/new/fluorescence/fluorescence.test.txt\"],\n",
    "    'rh': [\"datasets/new/remote_homology/remote_homology.train.txt\",\n",
    "            \"datasets/new/remote_homology/remote_homology.valid.txt\",\n",
    "            \"datasets/new/remote_homology/remote_homology.test.txt\"],\n",
    "    'ss': [\"datasets/new/secondary_structure/secondary_structure.train.txt\",\n",
    "            \"datasets/new/secondary_structure/secondary_structure.valid.txt\",\n",
    "            \"datasets/new/secondary_structure/secondary_structure.test.txt\"],\n",
    "    'pb': [\"datasets/new/pdbbind/pdbbind.train.txt\",\n",
    "           \"datasets/new/pdbbind/pdbbind.valid.txt\",\n",
    "           \"datasets/new/pdbbind/pdbbind.test.txt\"],\n",
    "    'ki': [\"datasets/new/kinase/kinase.train.txt\",\n",
    "           \"datasets/new/kinase/kinase.test.txt\"], # no vaild\n",
    "}\n",
    "\n",
    "\n",
    "def get_texts(train_pth, valid_pth, test_pth=None, cat=False):\n",
    "    train_texts, valid_texts = pd.read_csv(train_pth), pd.read_csv(valid_pth)\n",
    "    if test_pth is not None:\n",
    "        test_texts = pd.read_csv(test_pth)\n",
    "    if cat:\n",
    "        if test_pth is None:\n",
    "            cat_texts = pd.concat([train_texts, valid_texts], axis=0)\n",
    "        else:\n",
    "            cat_texts = pd.concat([train_texts, valid_texts, test_texts], axis=0)\n",
    "        return cat_texts \n",
    "    else:\n",
    "        return train_texts, valid_texts, test_texts\n",
    "\n",
    "# 分析train, valid数据分布\n",
    "for task in ['st', 'fl', 'rh', 'pb']:\n",
    "    trn_texts, val_texts, tst_texts = get_texts(*texts_embs_hub[task], cat=False)\n",
    "    plt.figure()\n",
    "    fig, (ax1, ax2, ax3) = plt.subplots(3, 1)\n",
    "    ax1.hist(trn_texts['label'])\n",
    "    ax2.hist(val_texts['label'])\n",
    "    ax3.hist(val_texts['label'])\n",
    "    plt.suptitle(task.upper())\n",
    "#     plt.savefig(\"./results/{}_distrib.jpg\".format(task))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.Processing ST(Stability), FL(Fluorescence), RH(Remote Homology), PB(Pdbbind) label\n",
    "50 bins approximately, random sample, reindex continous label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stability unique labels: 20 \n",
      " {7: 12328, 8: 11278, 6: 9892, 9: 8062, 10: 6978, 11: 5951, 5: 5036, 12: 3472, 13: 2348, 4: 2091, 3: 799, 14: 316, 2: 269, 1: 81, 15: 32, 0: 16, 16: 13, 17: 10, 19: 3, 18: 2}\n",
      "Fluorescence unique labels: 20 \n",
      " {0: 13800, 16: 10410, 15: 6165, 17: 5819, 1: 4432, 2: 3503, 14: 3218, 13: 3143, 3: 999, 18: 718, 12: 621, 4: 380, 11: 303, 9: 107, 6: 99, 10: 92, 19: 79, 5: 73, 7: 32, 8: 32}\n",
      "Remote_homology unique labels: 20 \n",
      " {0: 3932, 1: 3184, 2: 2491, 4: 736, 5: 536, 3: 532, 7: 458, 6: 336, 13: 226, 19: 157, 10: 151, 12: 142, 17: 137, 14: 133, 8: 122, 18: 120, 15: 101, 11: 99, 9: 92, 16: 81}\n",
      "Pdbbind unique labels: 20 \n",
      " {9: 1955, 8: 1864, 7: 1854, 10: 1586, 6: 1548, 5: 1240, 4: 850, 11: 818, 3: 542, 12: 381, 2: 219, 13: 124, 14: 74, 1: 68, 15: 34, 0: 16, 17: 8, 16: 8, 18: 4, 19: 3}\n"
     ]
    }
   ],
   "source": [
    "\n",
    "st_texts = get_texts(*texts_embs_hub['st'], cat=True)\n",
    "fl_texts = get_texts(*texts_embs_hub['fl'], cat=True)\n",
    "rh_texts = get_texts(*texts_embs_hub['rh'], cat=True) \n",
    "pb_texts = get_texts(*texts_embs_hub['pb'], cat=True)\n",
    "\n",
    "def lbls_to_idxs_sorted(texts):\n",
    "    counter = collections.Counter(texts['label'])\n",
    "    sorted_counter = dict(sorted(counter.items(), key=lambda kv: kv[1], reverse=True))\n",
    "    lbl_to_idx = {lbl: idx for idx, lbl in enumerate(sorted_counter.keys())}\n",
    "    texts['label'] = texts['label'].map(lambda x: lbl_to_idx[x])\n",
    "    return texts, sorted_counter\n",
    "\n",
    "def label_processor(task, texts, bins=50, seed=1145114):\n",
    "    bin_labels = pd.cut(texts['label'], bins=bins, labels=False, ordered=True)\n",
    "    texts['label'] = bin_labels \n",
    "    texts, sorted_counter = lbls_to_idxs_sorted(texts)\n",
    "    print(\"{} unique labels: {}\".format(task.capitalize(), len(sorted_counter)),\n",
    "            \"\\n\", sorted_counter)\n",
    "    return texts\n",
    "\n",
    "bins, seed = 20, 1145114\n",
    "num_samples = 'all'\n",
    "for task, texts in list(zip(['stability', 'fluorescence', 'remote_homology', 'pdbbind'], [st_texts, fl_texts, rh_texts, pb_texts])):\n",
    "    if num_samples != 'all':\n",
    "        texts = texts.sample(n=num_samples, random_state=seed)\n",
    "        texts.reset_index(drop=True, inplace=True)\n",
    "    \n",
    "    texts = label_processor(task, texts, bins, seed)\n",
    "    \n",
    "    texts = texts[['uniprot_id', 'seq', 'GO', 'label']]\n",
    "    print(texts['label'].value_counts())\n",
    "    \n",
    "    os.makedirs(\"./processed_data/{}\".format(task), exist_ok=True)\n",
    "    texts.to_csv(\"./processed_data/{}/sequence_go_label_{}.txt\".format(task, num_samples), index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.Processing SS(Secondary Structure) label\n",
    "random sample, mds reduction, kmeans cluster as label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.manifold import MDS\n",
    "from sklearn.cluster import KMeans\n",
    "\n",
    "ss_texts = get_texts(*texts_embs_hub['ss'], cat=True)\n",
    "num_samples = 2500\n",
    "\n",
    "if num_samples != 'all':\n",
    "    ss_texts = ss_texts.sample(n=num_samples, random_state=seed)\n",
    "    ss_texts.reset_index(drop=True, inplace=True)\n",
    "    \n",
    "ss_texts['label_len'] = ss_texts['label'].map(lambda x: len(x))\n",
    "\n",
    "ss_max_lbl_len = ss_texts['label_len'].max()\n",
    "# pad (-1): seq_len -> max_seq_len\n",
    "ss_pad_idx_labels = ss_texts['label'].map(lambda x: [int(n) for n in x]+[-1]*(ss_max_lbl_len-len(x)))\n",
    "ss_pad_idx_labels_arr = np.stack(ss_pad_idx_labels, axis=0)\n",
    "\n",
    "# MDS Reduction\n",
    "print(\"MDS reducing dimensions...\")\n",
    "ts = time.time()\n",
    "embedding = MDS(n_components=2, n_init=4, max_iter=300, n_jobs=-1)\n",
    "mds_embs = embedding.fit_transform(ss_pad_idx_labels_arr)\n",
    "print(\"Finish MDS in {:.2f} secs\".format(time.time()-ts))\n",
    "# KMEANS cluster label as final label\n",
    "print(\"KMEANS clustering...\")\n",
    "ts = time.time()\n",
    "kmeans = KMeans(n_clusters=20, n_init=10, max_iter=1000, random_state=1145114).fit(mds_embs)\n",
    "print(\"Finish KMEANS in {} secs\".format(time.time()-ts))\n",
    "ss_cluster_label = kmeans.labels_\n",
    "\n",
    "# 保存\n",
    "labelized_ss_texts = ss_texts[['uniprot_id', 'seq', 'GO', 'label']]\n",
    "labelized_ss_texts['label'] = ss_cluster_label\n",
    "os.makedirs(\"./processed_data/secondary_structure\", exist_ok=True)\n",
    "labelized_ss_texts.to_csv(f\"./processed_data/secondary_structure/sequence_go_label_{num_samples}.txt\", index=False,\n",
    "                        header=['uniprot_id', 'seq', 'GO', 'label'])\n",
    "print(labelized_ss_texts[3].value_counts())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.Processing KI(Kinase) label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ki_texts = pd.concat([pd.read_csv(texts_embs_hub['ki'][0]),\n",
    "                      pd.read_csv(texts_embs_hub['ki'][1])], axis=0).reset_index(drop=True)\n",
    "\n",
    "renamed_ids_train = pd.read_csv(\"datasets/new/kinase/kinase.train.txt\")\n",
    "renamed_ids_test = pd.read_csv(\"datasets/new/kinase/kinase.test.txt\")\n",
    "renamed_ids = pd.concat([renamed_ids_train, renamed_ids_test], axis=0).reset_index(drop=True)\n",
    "\n",
    "num_samples = 2500\n",
    "if num_samples != 'all':\n",
    "    ki_texts = ki_texts.sample(n=num_samples, random_state=seed)\n",
    "    ki_texts.reset_index(drop=True, inplace=True)\n",
    "    \n",
    "# 0.0, 1.0 -> 0, 1\n",
    "ki_texts['label'] = ki_texts['label'].astype(int)\n",
    "\n",
    "ki_texts = ki_texts[['uniprot_id', 'seq', 'GO', 'label']]\n",
    "os.makedirs(\"./processed_data/kinase\", exist_ok=True)\n",
    "ki_texts.to_csv(f\"./processed_data/kinase/sequence_go_label_{num_samples}.txt\", index=False,\n",
    "                header=['uniprot_id', 'seq', 'GO', 'label'])\n",
    "print(ki_texts['label'].value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "mulfusion",
   "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": "3.7.16"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "080091f32b56b05cf068fd1e7f07db5afbabfd14d069b7ab5cb0975ca780965e"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}