250_pandas_final_1_bis_bc-0.745-Copy1.nbconvert.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "2310db84",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:39.114051Z",
     "iopub.status.busy": "2022-07-26T20:05:39.113877Z",
     "iopub.status.idle": "2022-07-26T20:05:39.117306Z",
     "shell.execute_reply": "2022-07-26T20:05:39.117010Z"
    },
    "id": "b5b7fa3d"
   },
   "outputs": [],
   "source": [
    "#print(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "f5feca77",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:39.119032Z",
     "iopub.status.busy": "2022-07-26T20:05:39.118904Z",
     "iopub.status.idle": "2022-07-26T20:05:39.135007Z",
     "shell.execute_reply": "2022-07-26T20:05:39.134709Z"
    },
    "id": "2-1zeh_R06W9"
   },
   "outputs": [],
   "source": [
    "#%%capture\n",
    "#!pip install biopython\n",
    "#!pip install tensorflow-addons\n",
    "#!pip install tensorflow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "9c4d02b9",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:39.136963Z",
     "iopub.status.busy": "2022-07-26T20:05:39.136829Z",
     "iopub.status.idle": "2022-07-26T20:05:45.195487Z",
     "shell.execute_reply": "2022-07-26T20:05:45.194180Z"
    },
    "id": "YjE_wrIt04xe"
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import Bio\n",
    "from Bio.Seq import Seq\n",
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "from sklearn.model_selection import train_test_split\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from tqdm import tqdm\n",
    "from sklearn.model_selection import train_test_split\n",
    "import tensorflow.keras.layers as L\n",
    "import keras.backend as K\n",
    "import tensorflow as tf\n",
    "from sklearn.model_selection import StratifiedKFold, KFold, GroupKFold\n",
    "from sklearn.cluster import KMeans\n",
    "from Bio import SeqIO\n",
    "from Bio.Seq import Seq\n",
    "from joblib import Parallel, delayed\n",
    "import os\n",
    "import gc\n",
    "import random\n",
    "from tensorflow.keras.callbacks import TerminateOnNaN\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "import tensorflow_addons\n",
    "import tensorflow_addons as tfa\n",
    "from tensorflow.keras.callbacks import Callback"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ffc4399b",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:45.199287Z",
     "iopub.status.busy": "2022-07-26T20:05:45.198978Z",
     "iopub.status.idle": "2022-07-26T20:05:45.206131Z",
     "shell.execute_reply": "2022-07-26T20:05:45.205642Z"
    },
    "id": "672b6894",
    "outputId": "01a61119-3e6f-4196-b155-15cccca218ac"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "128"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import multiprocessing\n",
    "multiprocessing.cpu_count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "2ae811e1",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:45.208603Z",
     "iopub.status.busy": "2022-07-26T20:05:45.208054Z",
     "iopub.status.idle": "2022-07-26T20:05:45.222729Z",
     "shell.execute_reply": "2022-07-26T20:05:45.222181Z"
    },
    "id": "dtrZsneQLu9y"
   },
   "outputs": [],
   "source": [
    "def set_seeds(seed):\n",
    "    os.environ['PYTHONHASHSEED'] = '0'\n",
    "    random.seed(seed)\n",
    "    tf.random.set_seed(seed)\n",
    "    np.random.seed(seed)\n",
    "\n",
    "def set_global_determinism(seed):\n",
    "    set_seeds(seed=seed)\n",
    "\n",
    "    os.environ['TF_DETERMINISTIC_OPS'] = '1'\n",
    "    #os.environ['TF_CUDNN_DETERMINISTIC'] = '1'\n",
    "    \n",
    "    #tf.config.threading.set_inter_op_parallelism_threads(1)\n",
    "    #tf.config.threading.set_intra_op_parallelism_threads(1)\n",
    "\n",
    "# Call the above function with seed value\n",
    "set_seeds(seed=42)\n",
    "set_global_determinism(seed=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "89a398ef",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "c700d7b3",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:45.226069Z",
     "iopub.status.busy": "2022-07-26T20:05:45.225848Z",
     "iopub.status.idle": "2022-07-26T20:05:45.238647Z",
     "shell.execute_reply": "2022-07-26T20:05:45.238170Z"
    }
   },
   "outputs": [],
   "source": [
    "num_threads = 2\n",
    "# Maximum number of threads to use for OpenMP parallel regions.\n",
    "os.environ[\"OMP_NUM_THREADS\"] = str(num_threads)\n",
    "# Without setting below 2 environment variables, it didn't work for me. Thanks to @cjw85 \n",
    "os.environ[\"TF_NUM_INTRAOP_THREADS\"] = str(num_threads)\n",
    "os.environ[\"TF_NUM_INTEROP_THREADS\"] = str(num_threads)\n",
    "\n",
    "tf.config.threading.set_inter_op_parallelism_threads(\n",
    "    num_threads\n",
    ")\n",
    "tf.config.threading.set_intra_op_parallelism_threads(\n",
    "    num_threads\n",
    ")\n",
    "tf.config.set_soft_device_placement(True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "a12668a4",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:46.045767Z",
     "iopub.status.busy": "2022-07-26T20:05:46.045625Z",
     "iopub.status.idle": "2022-07-26T20:05:46.049378Z",
     "shell.execute_reply": "2022-07-26T20:05:46.048854Z"
    }
   },
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import MinMaxScaler\n",
    "scaler = MinMaxScaler(feature_range=(0.001, 0.999))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "8f76d671",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 225
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:46.051712Z",
     "iopub.status.busy": "2022-07-26T20:05:46.051247Z",
     "iopub.status.idle": "2022-07-26T20:05:58.422498Z",
     "shell.execute_reply": "2022-07-26T20:05:58.421674Z"
    },
    "id": "77e308c4",
    "outputId": "edec5cc4-4922-422d-ada0-dba5cb230592"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6739258, 2)\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>seq</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>TGCATTTTTTTCACATCTCTTTGCCACGGGGTGAAGGATAGGATGG...</td>\n",
       "      <td>0.646765</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>TGCATTTTTTTCACATCTATGTTGCGTTAGAACGATATTGGAACAC...</td>\n",
       "      <td>0.353235</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>TGCATTTTTTTCACATCTGTGAAGAATATCAGCTTTCAATCGTATT...</td>\n",
       "      <td>0.470647</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>TGCATTTTTTTCACATCAATCCGAGATATCTGTTGATAAACTTACC...</td>\n",
       "      <td>0.529353</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>TGCATTTTTTTCACATCAAGTTATCTGGTGTACGTTTTCTCGTATA...</td>\n",
       "      <td>0.705471</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 seq    target\n",
       "0  TGCATTTTTTTCACATCTCTTTGCCACGGGGTGAAGGATAGGATGG...  0.646765\n",
       "1  TGCATTTTTTTCACATCTATGTTGCGTTAGAACGATATTGGAACAC...  0.353235\n",
       "2  TGCATTTTTTTCACATCTGTGAAGAATATCAGCTTTCAATCGTATT...  0.470647\n",
       "3  TGCATTTTTTTCACATCAATCCGAGATATCTGTTGATAAACTTACC...  0.529353\n",
       "4  TGCATTTTTTTCACATCAAGTTATCTGGTGTACGTTTTCTCGTATA...  0.705471"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#get some stats for the train set and transform y\n",
    "train = pd.read_csv('train_sequences.txt',header=None,sep='\\t')\n",
    "train.columns = ['seq','target']\n",
    "scaler.fit(train[['target']])\n",
    "\n",
    "print(train.shape)\n",
    "#train=train[train.seq.str.len()==110]\n",
    "max_seq_len = train.seq.str.len().max()\n",
    "target_mean = train.target.mean()\n",
    "target_std = train.target.std()\n",
    "\n",
    "train['target']=scaler.transform(train[['target']])\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "ddc28e1d",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 225
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:58.431156Z",
     "iopub.status.busy": "2022-07-26T20:05:58.430771Z",
     "iopub.status.idle": "2022-07-26T20:05:58.848037Z",
     "shell.execute_reply": "2022-07-26T20:05:58.847504Z"
    },
    "id": "L9stm5ZA6OVe",
    "outputId": "b730ddb5-6a8b-4f17-ef04-45924762e90b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10000, 2)\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>seq</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>TGCATTTTTTTCACATCCGTGACTTTGTCAGTCCCGTGTGCCGCAG...</td>\n",
       "      <td>0.869883</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>TGCATTTTTTTCACATCGTTATTGGGCGAGGTGAGGCATCCATTGT...</td>\n",
       "      <td>0.764176</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>TGCATTTTTTTCACATCTACCCTGCTTATTATCCTCTAGAATTAGA...</td>\n",
       "      <td>0.529353</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>TGCATTTTTTTCACATCATATTGTGTAGTCACATCGCAATAGGTTT...</td>\n",
       "      <td>0.129245</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>TGCATTTTTTTCACATCCTTATCTGTAGAGATAAGTTATCGTCGCT...</td>\n",
       "      <td>0.705471</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 seq    target\n",
       "0  TGCATTTTTTTCACATCCGTGACTTTGTCAGTCCCGTGTGCCGCAG...  0.869883\n",
       "1  TGCATTTTTTTCACATCGTTATTGGGCGAGGTGAGGCATCCATTGT...  0.764176\n",
       "2  TGCATTTTTTTCACATCTACCCTGCTTATTATCCTCTAGAATTAGA...  0.529353\n",
       "3  TGCATTTTTTTCACATCATATTGTGTAGTCACATCGCAATAGGTTT...  0.129245\n",
       "4  TGCATTTTTTTCACATCCTTATCTGTAGAGATAAGTTATCGTCGCT...  0.705471"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#set apart 10000 promoters for test set\n",
    "X_test = train.sample(n=10000, random_state=42)\n",
    "X_test=X_test.reset_index(drop=True)\n",
    "print(X_test.shape)\n",
    "X_test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "2e85de8d",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 243
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:58.850189Z",
     "iopub.status.busy": "2022-07-26T20:05:58.850061Z",
     "iopub.status.idle": "2022-07-26T20:05:59.446331Z",
     "shell.execute_reply": "2022-07-26T20:05:59.445717Z"
    },
    "id": "JcBTnTz26rNb",
    "outputId": "b1785075-3c1e-406f-ee97-771a49e78ca0"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6739258, 2)\n",
      "(6729258, 2)\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>seq</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>TGCATTTTTTTCACATCTTTCTTTACAGGGAAGTTTGGTTTTTGAA...</td>\n",
       "      <td>0.822882</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>TGCATTTTTTTCACATCAGGTCCAGCTAGATTAGAAATGTTGTTCT...</td>\n",
       "      <td>0.646765</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>TGCATTTTTTTCACATCTCTCTTATGTATAAAGACTCGATGATTAG...</td>\n",
       "      <td>0.588059</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>TGCATTTTTTTCACATCGTCTAGGCTTATTTTCTTGAGTCTATTTG...</td>\n",
       "      <td>0.776059</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>TGCATTTTTTTCACATCTGGTCAAGCTTCCGATGTCAAATTTTGTG...</td>\n",
       "      <td>0.646765</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 seq    target\n",
       "0  TGCATTTTTTTCACATCTTTCTTTACAGGGAAGTTTGGTTTTTGAA...  0.822882\n",
       "1  TGCATTTTTTTCACATCAGGTCCAGCTAGATTAGAAATGTTGTTCT...  0.646765\n",
       "2  TGCATTTTTTTCACATCTCTCTTATGTATAAAGACTCGATGATTAG...  0.588059\n",
       "3  TGCATTTTTTTCACATCGTCTAGGCTTATTTTCTTGAGTCTATTTG...  0.776059\n",
       "4  TGCATTTTTTTCACATCTGGTCAAGCTTCCGATGTCAAATTTTGTG...  0.646765"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#remove from train set the test set\n",
    "print(train.shape)\n",
    "train = train[~train.index.isin(X_test.index)]\n",
    "train=train.reset_index(drop=True)\n",
    "print(train.shape)\n",
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "d4770600",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 206
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:59.448711Z",
     "iopub.status.busy": "2022-07-26T20:05:59.448482Z",
     "iopub.status.idle": "2022-07-26T20:05:59.548270Z",
     "shell.execute_reply": "2022-07-26T20:05:59.547714Z"
    },
    "id": "5c8db578",
    "outputId": "953dabc7-de62-41fb-e8fe-73d0f5ba5a95"
   },
   "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>seq</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAAAAACTAGCA...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                 seq  target\n",
       "0  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAAAAACTAGCA...     NaN\n",
       "1  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...     NaN\n",
       "2  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...     NaN\n",
       "3  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...     NaN\n",
       "4  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...     NaN"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#prepare the sequences to predict\n",
    "X_predict = pd.read_csv('test_sequences.txt',header=None,sep='\\t')\n",
    "X_predict.columns = ['seq','target']\n",
    "X_predict.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "25ef1c1e",
   "metadata": {
    "id": "c5OeqCbO58gd"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "362e945b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:59.555953Z",
     "iopub.status.busy": "2022-07-26T20:05:59.555499Z",
     "iopub.status.idle": "2022-07-26T20:05:59.566552Z",
     "shell.execute_reply": "2022-07-26T20:05:59.566076Z"
    },
    "id": "eFVKnNxEHC7a"
   },
   "outputs": [],
   "source": [
    "#this assigment is not necessary but here for hystorical reasons....\n",
    "X_train = train\n",
    "y_train = train['target']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "8f165f12",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:59.569687Z",
     "iopub.status.busy": "2022-07-26T20:05:59.569471Z",
     "iopub.status.idle": "2022-07-26T20:05:59.581299Z",
     "shell.execute_reply": "2022-07-26T20:05:59.580837Z"
    },
    "id": "e850b5a3",
    "outputId": "cc233766-87a7-41ef-a165-2ff5801233cd"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((6729258, 2), (10000, 2))"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape,X_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "b2233469",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 283
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:05:59.583490Z",
     "iopub.status.busy": "2022-07-26T20:05:59.583084Z",
     "iopub.status.idle": "2022-07-26T20:06:00.433948Z",
     "shell.execute_reply": "2022-07-26T20:06:00.433385Z"
    },
    "id": "e32dddb6",
    "outputId": "1102b4a3-d151-4b12-8626-5c1ef19ea07f"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x2b56102e9960>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#check distribution of train and test set\n",
    "fig,ax=plt.subplots(figsize=(8,4))\n",
    "X_train['target'].plot(ax=ax,kind='hist',histtype='step',label='train',density=True,bins=50)\n",
    "X_test['target'].plot(ax=ax,kind='hist',histtype='step',label='test',density=True,bins=50)\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "f11659be",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:00.436423Z",
     "iopub.status.busy": "2022-07-26T20:06:00.436011Z",
     "iopub.status.idle": "2022-07-26T20:06:00.444233Z",
     "shell.execute_reply": "2022-07-26T20:06:00.443770Z"
    },
    "id": "da7de9b5"
   },
   "outputs": [],
   "source": [
    "#tis function adds some padding for the shorter sequences\n",
    "def pad_sequence(X):\n",
    "    pad_length = max_seq_len-len(X)\n",
    "    return X+''.join(['P']*pad_length)\n",
    "\n",
    "#this function makes the complement of a dna sequences\n",
    "def to_complement(X):\n",
    "    dict_complement = {\n",
    "        'A':'T',\n",
    "        'T':'A',\n",
    "        'G':'C',\n",
    "        'C':'G',\n",
    "        'P':'P',\n",
    "        'N':'N',\n",
    "    }\n",
    "    return ''.join([dict_complement[n] for n in X])\n",
    "\n",
    "#one hot encoding of the dna\n",
    "def return_ohe(n, i):\n",
    "    tmp = [0] * n\n",
    "    tmp[i] = 1\n",
    "    return tmp\n",
    "\n",
    "#this function encode train data\n",
    "#it is applaied in batches during training\n",
    "#rather then to the full dataset\n",
    "def get_input(train, random_select='yes'):\n",
    "    #padding to the longest sequence in train\n",
    "    train.loc[:, \"seq\"]=train[\"seq\"].apply(pad_sequence)\n",
    "    #add the reverse-complement strand\n",
    "    train.loc[:, \"rc_seq\"] = train[\"seq\"].apply(to_complement)\n",
    "    ## get node features, which is one hot encoded\n",
    "    #print(train.head())\n",
    "    mapping = {}\n",
    "    vocab = [\"A\", \"G\", \"C\", \"T\", \"N\", \"P\"]\n",
    "    for i, s in enumerate(vocab):\n",
    "        mapping[s] = return_ohe(len(vocab), i)\n",
    "    #print(mapping)\n",
    "    #encoding the forward\n",
    "    X_node = np.stack(train[\"seq\"].apply(lambda x : list(map(lambda y : mapping[y], list(x)))))\n",
    "    #encoding the reverse compleent\n",
    "    X_node_rc = np.stack(train[\"rc_seq\"].apply(lambda x : list(map(lambda y : mapping[y], list(x)))))\n",
    "\n",
    "    #concatenate, forward first\n",
    "    X_node_1 = np.concatenate([X_node, X_node_rc], axis = 2).astype(np.float32) \n",
    "    \n",
    "    #concatenate, reverse complement first\n",
    "    X_node_2 = np.concatenate([X_node_rc, X_node], axis = 2).astype(np.float32)\n",
    "      #print(X_node.shape)\n",
    "\n",
    "    #return randomly forward first or reverse complement first\n",
    "    #this was to experiment different way of training\n",
    "    #final submission uses the forward first\n",
    "    if random_select == 'yes':\n",
    "        select = np.random.choice([0, 1])\n",
    "    else:\n",
    "        select=random_select\n",
    "    \n",
    "    if select == 1:\n",
    "        #return  forward first\n",
    "        return X_node_1\n",
    "    else:\n",
    "        #return reverse complement first\n",
    "        return X_node_2\n",
    "\n",
    "\n",
    "    #return np.concatenate([X_node_1, X_node_2], axis = 1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0e3e423a",
   "metadata": {
    "id": "mBgfZbDJvs7H"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "211e4f17",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:00.446545Z",
     "iopub.status.busy": "2022-07-26T20:06:00.446149Z",
     "iopub.status.idle": "2022-07-26T20:06:00.462507Z",
     "shell.execute_reply": "2022-07-26T20:06:00.461968Z"
    },
    "id": "6b22a372"
   },
   "outputs": [],
   "source": [
    "#https://github.com/WenYanger/Keras_Metrics/blob/master/PearsonCorr.py\n",
    "from keras import backend as K\n",
    "def pearson_r(y_true, y_pred):\n",
    "    #print(y_true,y_pred)\n",
    "    x = y_true\n",
    "    y = y_pred\n",
    "    mx = K.mean(x, axis=0)\n",
    "    my = K.mean(y, axis=0)\n",
    "    xm, ym = x - mx, y - my\n",
    "    r_num = K.sum(xm * ym)\n",
    "    x_square_sum = K.sum(xm * xm)\n",
    "    y_square_sum = K.sum(ym * ym)\n",
    "    r_den = K.sqrt(x_square_sum * y_square_sum)\n",
    "    r = r_num / r_den\n",
    "    return K.mean(r)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b3c04360",
   "metadata": {
    "id": "b788ae9d"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "12ecd41c",
   "metadata": {
    "id": "8aaed127"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "754ea17b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:00.465480Z",
     "iopub.status.busy": "2022-07-26T20:06:00.464962Z",
     "iopub.status.idle": "2022-07-26T20:06:00.477035Z",
     "shell.execute_reply": "2022-07-26T20:06:00.476529Z"
    },
    "id": "515b732d"
   },
   "outputs": [],
   "source": [
    "#first time to use a data loader, saving some useful ref\n",
    "\n",
    "#https://datascience.stackexchange.com/questions/47623/how-feed-a-numpy-array-in-batches-in-keras\n",
    "#https://stanford.edu/~shervine/blog/keras-how-to-generate-data-on-the-fly\n",
    "#https://pyimagesearch.com/2018/12/24/how-to-use-keras-fit-and-fit_generator-a-hands-on-tutorial/\n",
    "#https://datascience.stackexchange.com/questions/47623/how-feed-a-numpy-array-in-batches-in-keras\n",
    "#https://medium.com/analytics-vidhya/write-your-own-custom-data-generator-for-tensorflow-keras-1252b64e41c3\n",
    "#https://www.tensorflow.org/api_docs/python/tf/keras/utils/Sequence\n",
    "#https://numpy.org/doc/stable/reference/generated/numpy.array_split.html#numpy.array_split\n",
    "\n",
    "#The dataset loader grabs batches of data from the panda df\n",
    "#It pads sequences and apply one-hot encoding to forward and reverse strand.\n",
    "\n",
    "#The dataset is initialised with the number of batches that it shuld use and\n",
    "#figures out the number of samples in each batch using array_split from numpy\n",
    "\n",
    "from tensorflow.keras.utils import Sequence\n",
    "class DataGenerator(Sequence):\n",
    "    def __init__(self, x_data, num_batches=4500, shuffle=True, target_col='target', random_select='yes'):\n",
    "\n",
    "        self.data = x_data\n",
    "        #store the index of the df in array\n",
    "        self.indexes = np.array(self.data.index.values)\n",
    "        self.num_batches = num_batches\n",
    "        \n",
    "        #array_split distribute the data into the batches\n",
    "        #if the num_batches is not divisible by the length\n",
    "        #of the df, the batches will have sligtly different\n",
    "        #number of items\n",
    "        self.batches = np.array_split(self.indexes, self.num_batches)\n",
    "        \n",
    "\n",
    "        #it is good practice to shuffle the data between epochs\n",
    "        #but you do not want to shuffle for prediction\n",
    "        self.shuffle = shuffle\n",
    "        self.target_col = target_col\n",
    "        self.random_select = random_select\n",
    "    \n",
    "    def __len__(self):\n",
    "        #must be defined as the number\n",
    "        #of batches\n",
    "        return self.num_batches\n",
    "    \n",
    "    def __getitem__(self, idx):\n",
    "        #must be defined to return one batch of data\n",
    "        batch_x = self.data.loc[self.batches[idx],:]\n",
    "\n",
    "        batch_y = batch_x[self.target_col]\n",
    "      \n",
    "        #get_input makes the feature extraction\n",
    "        x_trn = get_input(batch_x, random_select = self.random_select)\n",
    "        return x_trn, batch_y.values\n",
    "    \n",
    "    def on_epoch_end(self):\n",
    "        #shuffle all data every epoch end\n",
    "        if self.shuffle:\n",
    "            np.random.shuffle(self.indexes)\n",
    "            self.batches = np.array_split(self.indexes, self.num_batches)\n",
    "            \n",
    "    def on_epoch_start(self):\n",
    "        #shuffle all data every epoch start\n",
    "        if self.shuffle:\n",
    "            np.random.shuffle(self.indexes)\n",
    "            self.batches = np.array_split(self.indexes, self.num_batches)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "3c4ba1b3",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:00.479281Z",
     "iopub.status.busy": "2022-07-26T20:06:00.479064Z",
     "iopub.status.idle": "2022-07-26T20:06:00.545362Z",
     "shell.execute_reply": "2022-07-26T20:06:00.544535Z"
    },
    "id": "36f9b731"
   },
   "outputs": [],
   "source": [
    "num_batches=3700\n",
    "train_generator = DataGenerator(X_train, num_batches = num_batches, shuffle=True, random_select=1)\n",
    "test_generator = DataGenerator(X_test, num_batches = 5, random_select=1)\n",
    "predict_generator = DataGenerator(X_predict, num_batches = num_batches, random_select=1) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "c2ce20ff",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:00.547944Z",
     "iopub.status.busy": "2022-07-26T20:06:00.547707Z",
     "iopub.status.idle": "2022-07-26T20:06:00.552020Z",
     "shell.execute_reply": "2022-07-26T20:06:00.551582Z"
    },
    "id": "0599ae23",
    "outputId": "b906bafa-b939-4600-9393-01af0d9851bd"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1819"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train_generator.batches[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "562e6aa5",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 424
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:00.576309Z",
     "iopub.status.busy": "2022-07-26T20:06:00.575930Z",
     "iopub.status.idle": "2022-07-26T20:06:00.590862Z",
     "shell.execute_reply": "2022-07-26T20:06:00.590373Z"
    },
    "id": "0c12c84b",
    "outputId": "0e8c2014-6a7b-4b39-923e-8ce9304e5b88"
   },
   "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>seq</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>TGCATTTTTTTCACATCTTTCTTTACAGGGAAGTTTGGTTTTTGAA...</td>\n",
       "      <td>0.822882</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>TGCATTTTTTTCACATCAGGTCCAGCTAGATTAGAAATGTTGTTCT...</td>\n",
       "      <td>0.646765</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>TGCATTTTTTTCACATCTCTCTTATGTATAAAGACTCGATGATTAG...</td>\n",
       "      <td>0.588059</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>TGCATTTTTTTCACATCGTCTAGGCTTATTTTCTTGAGTCTATTTG...</td>\n",
       "      <td>0.776059</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>TGCATTTTTTTCACATCTGGTCAAGCTTCCGATGTCAAATTTTGTG...</td>\n",
       "      <td>0.646765</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1814</th>\n",
       "      <td>TGCATTTTTTTCACATCGTCAAGTGTGTTCTTCGCCCGGCGGGGAT...</td>\n",
       "      <td>0.881588</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1815</th>\n",
       "      <td>TGCATTTTTTTCACATCGTTCAAGTGTCGGGACTGTCAGATTTAGT...</td>\n",
       "      <td>0.881588</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1816</th>\n",
       "      <td>TGCATTTTTTTCACATCGCAACCTGTTTTTTGTGGGGTTCCCTCGC...</td>\n",
       "      <td>0.588059</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1817</th>\n",
       "      <td>TGCATTTTTTTCACATCGTTTTAAGAGGATTCGAGAAACCACAGAG...</td>\n",
       "      <td>0.759040</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1818</th>\n",
       "      <td>TGCATTTTTTTCACATCCGAGTTAGCAAGTTTTAGGCTCTTATCAG...</td>\n",
       "      <td>0.588059</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1819 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    seq    target\n",
       "0     TGCATTTTTTTCACATCTTTCTTTACAGGGAAGTTTGGTTTTTGAA...  0.822882\n",
       "1     TGCATTTTTTTCACATCAGGTCCAGCTAGATTAGAAATGTTGTTCT...  0.646765\n",
       "2     TGCATTTTTTTCACATCTCTCTTATGTATAAAGACTCGATGATTAG...  0.588059\n",
       "3     TGCATTTTTTTCACATCGTCTAGGCTTATTTTCTTGAGTCTATTTG...  0.776059\n",
       "4     TGCATTTTTTTCACATCTGGTCAAGCTTCCGATGTCAAATTTTGTG...  0.646765\n",
       "...                                                 ...       ...\n",
       "1814  TGCATTTTTTTCACATCGTCAAGTGTGTTCTTCGCCCGGCGGGGAT...  0.881588\n",
       "1815  TGCATTTTTTTCACATCGTTCAAGTGTCGGGACTGTCAGATTTAGT...  0.881588\n",
       "1816  TGCATTTTTTTCACATCGCAACCTGTTTTTTGTGGGGTTCCCTCGC...  0.588059\n",
       "1817  TGCATTTTTTTCACATCGTTTTAAGAGGATTCGAGAAACCACAGAG...  0.759040\n",
       "1818  TGCATTTTTTTCACATCCGAGTTAGCAAGTTTTAGGCTCTTATCAG...  0.588059\n",
       "\n",
       "[1819 rows x 2 columns]"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.loc[train_generator.batches[0]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "f6ded2b3",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:00.592955Z",
     "iopub.status.busy": "2022-07-26T20:06:00.592731Z",
     "iopub.status.idle": "2022-07-26T20:06:00.598398Z",
     "shell.execute_reply": "2022-07-26T20:06:00.597906Z"
    },
    "id": "ZhNS7Cx-cPt4",
    "outputId": "b4938b33-3f4e-489d-d249-dcd50827e2a0"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([   0,    1,    2, ..., 1997, 1998, 1999])"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_generator.batches[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "e54dc8c6",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:01.072997Z",
     "iopub.status.busy": "2022-07-26T20:06:01.072484Z",
     "iopub.status.idle": "2022-07-26T20:06:01.511101Z",
     "shell.execute_reply": "2022-07-26T20:06:01.510562Z"
    },
    "id": "afa8555b",
    "outputId": "901f740b-d670-4e2e-bd79-b5ab2fd72c76"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1819, 142, 12), (1819,))"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#visualize shape of train gen\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "t = train_generator[0]\n",
    "t[0].shape,t[1].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "da15b8a9",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:01.513633Z",
     "iopub.status.busy": "2022-07-26T20:06:01.513157Z",
     "iopub.status.idle": "2022-07-26T20:06:01.517709Z",
     "shell.execute_reply": "2022-07-26T20:06:01.517157Z"
    },
    "id": "8a9e69df"
   },
   "outputs": [],
   "source": [
    "def rmse(y_actual, y_pred):\n",
    "    mse = tf.keras.losses.mean_squared_error(y_actual, y_pred)\n",
    "    return K.sqrt(mse)\n",
    "\n",
    "\n",
    "#a shortcut for a Bidirectional GRU layer\n",
    "def gru_layer(hidden_dim=256, dropout=0.1):\n",
    "    return layers.Bidirectional(layers.GRU(hidden_dim, \n",
    "                                 dropout=dropout, \n",
    "                                 return_sequences=True, \n",
    "                                 kernel_initializer = 'orthogonal',\n",
    "                                 bias_initializer='orthogonal'\n",
    "                                ))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "62ff4391",
   "metadata": {
    "id": "YlX70LmoxB_l"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "e621f0fc",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:01.519962Z",
     "iopub.status.busy": "2022-07-26T20:06:01.519460Z",
     "iopub.status.idle": "2022-07-26T20:06:01.535899Z",
     "shell.execute_reply": "2022-07-26T20:06:01.535371Z"
    },
    "id": "54599fc7"
   },
   "outputs": [],
   "source": [
    "#The actual model\n",
    "#two Bidirectional GRUs followed by CONVs and fully connected \n",
    "\n",
    "def build_model(\n",
    "    input_shape,\n",
    "    gru_hidden_dim=64,\n",
    "    mlp_units=[512, 32],\n",
    "    do_conv=True,    \n",
    "    kernel_size1=6,\n",
    "    kernel_size2=3,\n",
    "    kernel_size3=3,\n",
    "    pool_size1=3,\n",
    "    pool_size2=3,\n",
    "    filters1=256,\n",
    "    filters2=128,\n",
    "    filters3=64):\n",
    "    \n",
    "    inputs = keras.Input(shape=input_shape)\n",
    "    \n",
    "    \n",
    "    x = inputs\n",
    "    \n",
    "    x = gru_layer(hidden_dim=gru_hidden_dim,dropout=0)(x)\n",
    "    x = gru_layer(hidden_dim=gru_hidden_dim,dropout=0)(x)\n",
    "    \n",
    "    if do_conv:\n",
    "        x = layers.Conv1D(filters=filters1, kernel_size=kernel_size1, activation='LeakyReLU')(x)\n",
    "        x = layers.MaxPooling1D(pool_size=pool_size1)(x)\n",
    "        x = layers.Conv1D(filters=filters2, kernel_size=kernel_size2, activation='LeakyReLU')(x)\n",
    "        x = layers.MaxPooling1D(pool_size=pool_size2)(x)\n",
    "        x = layers.Conv1D(filters=filters3, kernel_size=kernel_size3, activation='LeakyReLU')(x)\n",
    "    else:\n",
    "        x = layers.Conv1D(filters=filters1, kernel_size=kernel_size1, activation='LeakyReLU')(x)\n",
    "    \n",
    "    x = layers.Flatten()(x)\n",
    "    \n",
    "    for dim in mlp_units:\n",
    "        x = layers.Dense(dim)(x)\n",
    "        x = layers.BatchNormalization()(x)\n",
    "        x = layers.LeakyReLU()(x)\n",
    "    outputs = layers.Dense(1, activation='sigmoid')(x)\n",
    "    return keras.Model(inputs, outputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cf475b89",
   "metadata": {
    "id": "928ad2cb"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "294d7ec0",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:01.539461Z",
     "iopub.status.busy": "2022-07-26T20:06:01.538787Z",
     "iopub.status.idle": "2022-07-26T20:06:01.550817Z",
     "shell.execute_reply": "2022-07-26T20:06:01.550220Z"
    },
    "id": "2048a8cc"
   },
   "outputs": [],
   "source": [
    "def coeff_det(y_true, y_pred):\n",
    "    SS_res =  K.sum(K.square( y_true-y_pred ))\n",
    "    SS_tot = K.sum(K.square( y_true - K.mean(y_true) ) )\n",
    "    return ( 1 - SS_res/(SS_tot + K.epsilon()) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "fedee018",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:01.553547Z",
     "iopub.status.busy": "2022-07-26T20:06:01.553424Z",
     "iopub.status.idle": "2022-07-26T20:06:01.560553Z",
     "shell.execute_reply": "2022-07-26T20:06:01.560058Z"
    }
   },
   "outputs": [],
   "source": [
    "#11 PearsonR² 0.919 Spearman 0.962 ScorePearsonR² 0.745 ScoreSpearman 0.802\n",
    "config={\n",
    "    'gru_hidden_dim':256,\n",
    "    'mlp_units':[512, 32],\n",
    "    'do_conv': True,    \n",
    "    'kernel_size1':6,\n",
    "    'kernel_size2':3,\n",
    "    'kernel_size3':3,\n",
    "    'pool_size1':3,\n",
    "    'pool_size2':2,\n",
    "    'filters1':256,\n",
    "    'filters2':128,\n",
    "    'filters3':64    \n",
    "    \n",
    " }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "3ed54eec",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:01.563661Z",
     "iopub.status.busy": "2022-07-26T20:06:01.563190Z",
     "iopub.status.idle": "2022-07-26T20:06:04.253770Z",
     "shell.execute_reply": "2022-07-26T20:06:04.253049Z"
    },
    "id": "1b1f51a2",
    "outputId": "55bf4b22-b448-4ea2-af66-b1a10e0cef25"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-26 21:06:02.128339: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  SSE4.1 SSE4.2 AVX AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2022-07-26 21:06:02.821959: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1525] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 43648 MB memory:  -> device: 0, name: NVIDIA A40, pci bus id: 0000:81:00.0, compute capability: 8.6\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_1 (InputLayer)        [(None, 142, 12)]         0         \n",
      "                                                                 \n",
      " bidirectional (Bidirectiona  (None, 142, 512)         414720    \n",
      " l)                                                              \n",
      "                                                                 \n",
      " bidirectional_1 (Bidirectio  (None, 142, 512)         1182720   \n",
      " nal)                                                            \n",
      "                                                                 \n",
      " conv1d (Conv1D)             (None, 137, 256)          786688    \n",
      "                                                                 \n",
      " max_pooling1d (MaxPooling1D  (None, 45, 256)          0         \n",
      " )                                                               \n",
      "                                                                 \n",
      " conv1d_1 (Conv1D)           (None, 43, 128)           98432     \n",
      "                                                                 \n",
      " max_pooling1d_1 (MaxPooling  (None, 21, 128)          0         \n",
      " 1D)                                                             \n",
      "                                                                 \n",
      " conv1d_2 (Conv1D)           (None, 19, 64)            24640     \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 1216)              0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 512)               623104    \n",
      "                                                                 \n",
      " batch_normalization (BatchN  (None, 512)              2048      \n",
      " ormalization)                                                   \n",
      "                                                                 \n",
      " leaky_re_lu_3 (LeakyReLU)   (None, 512)               0         \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 32)                16416     \n",
      "                                                                 \n",
      " batch_normalization_1 (Batc  (None, 32)               128       \n",
      " hNormalization)                                                 \n",
      "                                                                 \n",
      " leaky_re_lu_4 (LeakyReLU)   (None, 32)                0         \n",
      "                                                                 \n",
      " dense_2 (Dense)             (None, 1)                 33        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 3,148,929\n",
      "Trainable params: 3,147,841\n",
      "Non-trainable params: 1,088\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "for n in train_generator:\n",
    "    input_shape = n[0].shape[1:]\n",
    "    break\n",
    "\n",
    "model_name = 'sub_model'\n",
    "\n",
    "model = build_model(\n",
    "    input_shape=input_shape,\n",
    "    **config   \n",
    ")\n",
    "mae = tf.keras.losses.MeanAbsoluteError()\n",
    "\n",
    "#binary crossentropy is used with the sigmoid activation on the output layer\n",
    "bc = tf.keras.losses.BinaryCrossentropy()\n",
    "\n",
    "\n",
    "model.compile(\n",
    "    loss=bc, \n",
    "    optimizer=keras.optimizers.Adam(epsilon=1e-6, global_clipnorm=1.0),\n",
    "    metrics=[pearson_r, coeff_det],\n",
    ")\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "3b149f75",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:04.260416Z",
     "iopub.status.busy": "2022-07-26T20:06:04.260043Z",
     "iopub.status.idle": "2022-07-26T20:06:04.329542Z",
     "shell.execute_reply": "2022-07-26T20:06:04.328849Z"
    },
    "id": "d972655d"
   },
   "outputs": [],
   "source": [
    "\n",
    "#set the final number of batches\n",
    "num_batches=6700\n",
    "train_generator = DataGenerator(X_train, num_batches = num_batches, shuffle=True, random_select=1)\n",
    "test_generator = DataGenerator(X_test, num_batches = 5, random_select=1)\n",
    "predict_generator = DataGenerator(X_predict, num_batches = num_batches, random_select=1) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9afbd3b3",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "265b15c8",
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "execution": {
     "iopub.execute_input": "2022-07-26T20:06:04.332963Z",
     "iopub.status.busy": "2022-07-26T20:06:04.332718Z",
     "iopub.status.idle": "2022-07-27T04:12:06.752023Z",
     "shell.execute_reply": "2022-07-27T04:12:06.751443Z"
    },
    "id": "fb77caaa",
    "outputId": "c0627b2f-d841-4691-d2a4-f9af60165e69"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-07-26 21:06:10.836246: I tensorflow/stream_executor/cuda/cuda_dnn.cc:366] Loaded cuDNN version 8201\n",
      "2022-07-26 21:06:15.378784: I tensorflow/stream_executor/cuda/cuda_blas.cc:1774] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6700/6700 [==============================] - 2629s 391ms/step - loss: 0.6240 - pearson_r: 0.6794 - coeff_det: 0.4536 - val_loss: 0.6200 - val_pearson_r: 0.7153 - val_coeff_det: 0.5093\n",
      "6700/6700 [==============================] - 2662s 397ms/step - loss: 0.6207 - pearson_r: 0.7259 - coeff_det: 0.5252 - val_loss: 0.6183 - val_pearson_r: 0.7391 - val_coeff_det: 0.5451\n",
      "6700/6700 [==============================] - 2659s 397ms/step - loss: 0.6198 - pearson_r: 0.7398 - coeff_det: 0.5456 - val_loss: 0.6202 - val_pearson_r: 0.7452 - val_coeff_det: 0.5056\n",
      "6700/6700 [==============================] - 2649s 395ms/step - loss: 0.6194 - pearson_r: 0.7451 - coeff_det: 0.5535 - val_loss: 0.6175 - val_pearson_r: 0.7509 - val_coeff_det: 0.5632\n",
      "6700/6700 [==============================] - 2651s 395ms/step - loss: 0.6192 - pearson_r: 0.7488 - coeff_det: 0.5591 - val_loss: 0.6172 - val_pearson_r: 0.7552 - val_coeff_det: 0.5695\n",
      "6700/6700 [==============================] - 2657s 396ms/step - loss: 0.6190 - pearson_r: 0.7518 - coeff_det: 0.5635 - val_loss: 0.6170 - val_pearson_r: 0.7594 - val_coeff_det: 0.5748\n",
      "6700/6700 [==============================] - 2642s 394ms/step - loss: 0.6188 - pearson_r: 0.7544 - coeff_det: 0.5676 - val_loss: 0.6177 - val_pearson_r: 0.7611 - val_coeff_det: 0.5583\n",
      "6700/6700 [==============================] - 2651s 395ms/step - loss: 0.6186 - pearson_r: 0.7570 - coeff_det: 0.5714 - val_loss: 0.6176 - val_pearson_r: 0.7636 - val_coeff_det: 0.5616\n",
      "6700/6700 [==============================] - 2642s 394ms/step - loss: 0.6184 - pearson_r: 0.7592 - coeff_det: 0.5748 - val_loss: 0.6165 - val_pearson_r: 0.7657 - val_coeff_det: 0.5837\n",
      "6700/6700 [==============================] - 2648s 395ms/step - loss: 0.6183 - pearson_r: 0.7614 - coeff_det: 0.5782 - val_loss: 0.6171 - val_pearson_r: 0.7718 - val_coeff_det: 0.5705\n",
      "6700/6700 [==============================] - 2655s 396ms/step - loss: 0.6181 - pearson_r: 0.7634 - coeff_det: 0.5812 - val_loss: 0.6164 - val_pearson_r: 0.7698 - val_coeff_det: 0.5876\n"
     ]
    }
   ],
   "source": [
    "#this code train one epoch at time and save the weights\n",
    "#the EarlyStopping and TerminateOnNaN are not necessary\n",
    "#but still here for hystorical reasons\n",
    "verbosity = 1\n",
    "for epoch in range(1, 12):\n",
    "    \n",
    "    callbacks=[\n",
    "        \n",
    "                #tf.keras.callbacks.ReduceLROnPlateau(),\n",
    "                #tf.keras.callbacks.ReduceLROnPlateau(patience = 3, factor = 0.25, min_lr = 1e-06,\n",
    "                #                                     verbose = 1, monitor='val_pearson_r',mode='max'),\n",
    "                tf.keras.callbacks.EarlyStopping(patience = 3, restore_best_weights = True,  \n",
    "                                                 monitor='val_pearson_r', mode='max'),\n",
    "                tf.keras.callbacks.ModelCheckpoint(model_name+'.{epoch}.h5'.format(epoch=epoch), save_best_only=True,\n",
    "                                                   mode='max', monitor='val_pearson_r'),\n",
    "                tf.keras.callbacks.TerminateOnNaN()]\n",
    "    \n",
    "    if epoch >1:\n",
    "        model = tf.keras.models.load_model(model_name+'.{epoch}.h5'.format(epoch=epoch-1), custom_objects={'rmse': rmse,'pearson_r':pearson_r,'coeff_det':coeff_det})\n",
    "\n",
    "    model.fit(train_generator, validation_data=test_generator, epochs=1, verbose=verbosity, callbacks=callbacks)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "e55899b6",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-27T04:12:06.755634Z",
     "iopub.status.busy": "2022-07-27T04:12:06.755275Z",
     "iopub.status.idle": "2022-07-27T04:13:59.792686Z",
     "shell.execute_reply": "2022-07-27T04:13:59.791641Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "make prediction using:  sub_model.11.h5\n",
      "shape: (71103, 1) (71103, 2)\n",
      "                                                 seq     preds\n",
      "0  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAAAAACTAGCA...  0.708024\n",
      "1  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...  0.709611\n",
      "2  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...  0.699942\n",
      "3  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...  0.692968\n",
      "4  TGCATTTTTTTCACATCAAAAAAAAAAAAAAAAAAAGAAAACGAGC...  0.742253\n",
      "done\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#this code loads a model and create predictions\n",
    "#it also transfor the predictions into the \n",
    "#format for submission.\n",
    "\n",
    "import json\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from collections import OrderedDict\n",
    "\n",
    "def make_prediction(model_id, model_name, is_class = False):\n",
    "    print('make prediction using: ', model_name+'.{model_id}.h5'.format(model_id=model_id))\n",
    "    model = tf.keras.models.load_model(\n",
    "        model_name+'.{model_id}.h5'.format(model_id=model_id),custom_objects={\n",
    "    'rmse': rmse,'pearson_r':pearson_r,'coeff_det':coeff_det})\n",
    "    predict_generator_1 = DataGenerator(X_predict, num_batches = 10000, random_select=1, shuffle=False) \n",
    "    out_1 = model.predict(predict_generator_1)\n",
    "    print('shape:', out_1.shape, X_predict.shape)\n",
    "    if is_class:\n",
    "        #trained a classification once, then dropped the idea\n",
    "        X_predict['preds_1'] = np.argmax(out_1,axis=1).squeeze()\n",
    "    else:\n",
    "        X_predict['preds_1'] = out_1.max(axis=1).squeeze()\n",
    "    X_predict['preds_1'].plot(kind='hist',bins=50)\n",
    "    X_predict['preds']=X_predict['preds_1']\n",
    "    \n",
    "    X_predict[['seq','preds']].to_csv(\n",
    "        model_name+'.{model_id}.csv'.format(model_id=model_id),\n",
    "        header=False,index=False,sep='\\t') \n",
    "    \n",
    "    with open('sample_submission.json', 'r') as f:\n",
    "        ground=json.load(f)\n",
    "    indices=np.array([int(indice) for indice in list(ground.keys())])\n",
    "\n",
    "    df_pred = pd.read_csv(model_name+'.{model_id}.csv'.format(model_id=model_id), sep='\\t',header=None)\n",
    "    df_pred.columns = ['seq','preds']\n",
    "\n",
    "    Y_pred=df_pred.preds.values\n",
    "    print(df_pred.head())\n",
    "    PRED_DATA = OrderedDict()\n",
    "    for i in indices:\n",
    "        PRED_DATA[str(i)]= float(Y_pred[i])\n",
    "\n",
    "    def dump(pred_dict, pred_file):\n",
    "        with open(pred_file, 'w') as f:\n",
    "            json.dump(pred_dict, f)\n",
    "    dump(PRED_DATA,model_name+'.{model_id}.csv.json'.format(model_id=model_id)) \n",
    "    print('done')\n",
    "    \n",
    "#select the epoch and predict\n",
    "make_prediction(11, model_name=model_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "225023cc",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-27T04:13:59.801224Z",
     "iopub.status.busy": "2022-07-27T04:13:59.801096Z",
     "iopub.status.idle": "2022-07-27T04:13:59.811060Z",
     "shell.execute_reply": "2022-07-27T04:13:59.810599Z"
    }
   },
   "outputs": [],
   "source": [
    "#outs from leaderbord \n",
    "#11 PearsonR² 0.919 Spearman 0.962 ScorePearsonR² 0.745 ScoreSpearman 0.802 *\n",
    "#10 PearsonR² 0.911 Spearman 0.960 ScorePearsonR² 0.729 ScoreSpearman 0.792"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "88f6fa20",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-27T04:13:59.814718Z",
     "iopub.status.busy": "2022-07-27T04:13:59.814366Z",
     "iopub.status.idle": "2022-07-27T04:13:59.826137Z",
     "shell.execute_reply": "2022-07-27T04:13:59.825172Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tqdm==4.64.0\n",
      "matplotlib==3.5.1\n",
      "tensorflow==2.7.0\n",
      "keras==2.7.0\n",
      "biopython==1.79\n",
      "numpy==1.22.3\n",
      "pandas==1.4.2\n",
      "joblib==1.1.0\n",
      "scikit-learn==1.0.2\n",
      "tensorflow-addons==0.16.1\n"
     ]
    }
   ],
   "source": [
    "#using this code to save the version of the used packages\n",
    "#https://stackoverflow.com/questions/40428931/package-for-listing-version-of-packages-used-in-a-jupyter-notebook\n",
    "import pkg_resources\n",
    "import types\n",
    "def get_imports():\n",
    "    for name, val in globals().items():\n",
    "        #print(name)\n",
    "        if isinstance(val, types.ModuleType):\n",
    "            # Split ensures you get root package, \n",
    "            # not just imported function\n",
    "            name = val.__name__.split(\".\")[0]\n",
    "\n",
    "        elif isinstance(val, type):\n",
    "            name = val.__module__.split(\".\")[0]\n",
    "\n",
    "        # Some packages are weird and have different\n",
    "        # imported names vs. system/pip names. Unfortunately,\n",
    "        # there is no systematic way to get pip names from\n",
    "        # a package's imported name. You'll have to add\n",
    "        # exceptions to this list manually!\n",
    "        poorly_named_packages = {\n",
    "            \"PIL\":     \"Pillow\",\n",
    "            \"sklearn\": \"scikit-learn\",\n",
    "            \"Bio\":     \"biopython\",\n",
    "            \"tensorflow_addons\":\"tensorflow-addons\"\n",
    "        }\n",
    "        if name in poorly_named_packages.keys():\n",
    "            name = poorly_named_packages[name]\n",
    "\n",
    "        yield name\n",
    "imports = list(set(get_imports()))\n",
    "\n",
    "# The only way I found to get the version of the root package\n",
    "# from only the name of the package is to cross-check the names \n",
    "# of installed packages vs. imported packages\n",
    "requirements = []\n",
    "for m in pkg_resources.working_set:\n",
    "    if m.project_name in imports and m.project_name!=\"pip\":\n",
    "        requirements.append((m.project_name, m.version))\n",
    "\n",
    "for r in requirements:\n",
    "    print(\"{}=={}\".format(*r))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "6464178d",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-07-27T04:13:59.841130Z",
     "iopub.status.busy": "2022-07-27T04:13:59.840867Z",
     "iopub.status.idle": "2022-07-27T04:13:59.851698Z",
     "shell.execute_reply": "2022-07-27T04:13:59.851209Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/homes/mtinti/miniconda3/envs/proteinBert2/bin/python\n",
      "3.10.4 | packaged by conda-forge | (main, Mar 24 2022, 17:38:57) [GCC 10.3.0]\n",
      "sys.version_info(major=3, minor=10, micro=4, releaselevel='final', serial=0)\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "print(sys.executable)\n",
    "print(sys.version)\n",
    "print(sys.version_info)"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "machine_shape": "hm",
   "name": "pandas_final_1_bis(4).ipynb",
   "provenance": []
  },
  "gpuClass": "standard",
  "kernelspec": {
   "display_name": "conda-env-miniconda3-proteinBert2-py",
   "language": "python",
   "name": "proteinbert2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.4"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}