{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Secondary Structure Elements Word2Vec Encoder Demo\n",
    "\n",
    "This demo creates a dataset by extracting secondary structure elements \"H\", then encode an overlapping Ngram feature vector\n",
    "\n",
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyspark import SparkConf, SparkContext, SQLContext\n",
    "from mmtfPyspark.ml import ProteinSequenceEncoder\n",
    "from mmtfPyspark.mappers import StructureToPolymerChains\n",
    "from mmtfPyspark.filters import ContainsLProteinChain\n",
    "from mmtfPyspark.datasets import secondaryStructureElementExtractor\n",
    "from mmtfPyspark.webfilters import Pisces\n",
    "from mmtfPyspark.io import mmtfReader"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Configure Spark Context"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "conf = SparkConf() \\\n",
    "            .setMaster(\"local[*]\") \\\n",
    "            .setAppName(\"SecondaryStructureElementsWord2VecEncoderDemo\")\n",
    "\n",
    "sc = SparkContext(conf = conf)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " ## Read MMTF Hadoop sequence file and \n",
    " \n",
    " Create a non-redundant set(<=20% seq. identity) of L-protein chains"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "path = \"../../resources/mmtf_reduced_sample/\"\n",
    "fraction = 0.05\n",
    "seed = 123\n",
    "\n",
    "pdb = mmtfReader \\\n",
    "        .read_sequence_file(path, sc) \\\n",
    "        .flatMap(StructureToPolymerChains(False, True)) \\\n",
    "        .filter(ContainsLProteinChain()) \\\n",
    "        .sample(False, fraction, seed)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Extract Element \"H\" from Secondary Structure"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "original data   : 3297\n",
      "+-----------------------------+-----+\n",
      "|sequence                     |label|\n",
      "+-----------------------------+-----+\n",
      "|ACAGV                        |H    |\n",
      "|GIGLHLAVRLA                  |H    |\n",
      "|RLWEAARA                     |H    |\n",
      "|KSVAAARE                     |H    |\n",
      "|EDAVASVLDVNVVGTVRMLQAFLPDMKRR|H    |\n",
      "|VYCASKFALEGLCESLAVLLLPF      |H    |\n",
      "|IHTFHRFYQYLALSKQVFREA        |H    |\n",
      "|EEVAEVFLTALR                 |H    |\n",
      "|LPLLRMRL                     |H    |\n",
      "|NYVTAMHREVF                  |H    |\n",
      "+-----------------------------+-----+\n",
      "only showing top 10 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "label = \"H\"\n",
    "data = secondaryStructureElementExtractor.get_dataset(pdb, label).cache()\n",
    "print(f\"original data   : {data.count()}\")\n",
    "data.show(10, False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Word2Vec encoded feature Vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+--------------------+-----+--------------------+--------------------+\n",
      "|            sequence|label|               ngram|            features|\n",
      "+--------------------+-----+--------------------+--------------------+\n",
      "|               ACAGV|    H|    [AC, CA, AG, GV]|[0.46121373027563...|\n",
      "|         GIGLHLAVRLA|    H|[GI, IG, GL, LH, ...|[-0.2606903441250...|\n",
      "|            RLWEAARA|    H|[RL, LW, WE, EA, ...|[0.16112836982522...|\n",
      "|            KSVAAARE|    H|[KS, SV, VA, AA, ...|[1.15827076775687...|\n",
      "|EDAVASVLDVNVVGTVR...|    H|[ED, DA, AV, VA, ...|[0.37046241248026...|\n",
      "+--------------------+-----+--------------------+--------------------+\n",
      "only showing top 5 rows\n",
      "\n"
     ]
    }
   ],
   "source": [
    "segmentLength = 11\n",
    "n = 2\n",
    "windowSize = (segmentLength-1)/2\n",
    "vectorSize = 50\n",
    "\n",
    "encoder = ProteinSequenceEncoder(data)\n",
    "# overlapping_ngram_word2vec_encode uses keyword attributes\n",
    "data = encoder.overlapping_ngram_word2vec_encode(n=n, windowSize=windowSize, vectorSize=vectorSize)\n",
    "\n",
    "data.show(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Terminate Spark Context"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "sc.stop()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}