{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Building the holo CCR2 MSM\n",
    "\n",
    "TICA, clustering, and model building on all the holo Anton data.\n",
    "\n",
    "- Bryn C. Taylor\n",
    "- Christopher T. Lee\n",
    "- Rommie E. Amaro"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "import pyemma\n",
    "pyemma.__version__\n",
    "import os\n",
    "%pylab inline\n",
    "matplotlib.rcParams.update({'font.size': 12})\n",
    "import pyemma.coordinates as coor\n",
    "import pyemma.msm as msm\n",
    "import pyemma.plots as mplt\n",
    "import numpy as np\n",
    "import pickle \n",
    "import mdtraj as md"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load coordinates and select features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "topfile = '../holo_trajs/holo_proteinonly.pdb'\n",
    "from glob import glob\n",
    "traj_list = glob('../holo_trajs/*holo*')   # these are the simulations used in the indivudual Holo and Holo analyses notebooks\n",
    "traj_list.sort()\n",
    "traj_list"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Choose residues to evaluate distances between"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "net = [(9,168), (191,107), (274,107), (274,191), (18,261),(89, 260), (89, 228), (260, 228), (18, 67), (67, 89), \n",
    "       (67, 260), (67, 261), (19, 228), (215,50), (214,103), (214,50), (213,274), (274,50), (274,32), (274,103), (274,36), \n",
    "       (36,213)]\n",
    "net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# choose features to look at\n",
    "feat = coor.featurizer(topfile)\n",
    "feat.add_residue_mindist(residue_pairs=net)\n",
    "\n",
    "# describe the features\n",
    "feat.describe()\n",
    "print(feat.dimension())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# set input as features in trajectory \n",
    "inp = coor.source(traj_list, feat)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## TICA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dt = .24\n",
    "lag = np.int(400 / dt) # determine this from the timescales v. lag time plot\n",
    "tica_obj_400 = coor.tica(inp, lag=lag, kinetic_map=True, var_cutoff=0.90, reversible=True)\n",
    "\n",
    "print('TICA Dimension with a lag of 400ns:', tica_obj_400.dimension())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "tica_features_lag400 = tica_obj_400.feature_TIC_correlation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.savetxt('/msm_holo/tica_features_correlation_ticalag400.txt', tica_features_lag400)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "for i in range(2):\n",
    "    if tica_obj_400.eigenvectors[0, i] > 0:\n",
    "        tica_obj_400.eigenvectors[:, i] *= -1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Y = tica_obj_400.get_output() \n",
    "total_sim_Y = np.concatenate(Y, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## save all \n",
    "with open('/msm_holo/Y_ticalag400.txt', 'wb') as fp:\n",
    "    pickle.dump(Y, fp)\n",
    "with open('/msm_holo/total_sim_Y_ticalag400.txt', 'wb') as fp:\n",
    "    pickle.dump(total_sim_Y, fp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## open all\n",
    "with open('/msm_holo/Y_ticalag400.txt', 'rb') as fp:\n",
    "    Y = pickle.load(fp)\n",
    "with open('/msm_holo/total_sim_Y_ticalag400.txt', 'rb') as fp:\n",
    "    total_sim_Y = pickle.load(fp)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## K-means clustering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clustering_ticalag400 = coor.cluster_kmeans(Y, max_iter=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "dtrajs = clustering_ticalag400.dtrajs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## save all\n",
    "\n",
    "##clustercenters\n",
    "with open('/msm_holo/clustercenters_ticalag400_clustdefault_net2.txt', 'wb') as fp:\n",
    "    pickle.dump(clustering_ticalag400.clustercenters, fp)\n",
    "\n",
    "##dtrajs\n",
    "with open('/msm_holo/dtrajs_ticalag400_clustdefault.txt', 'wb') as fp:\n",
    "    pickle.dump(dtrajs, fp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# open all \n",
    "\n",
    "# clustercenters  \n",
    "with open('/msm_holo/clustercenters_ticalag400_clustdefault_net2.txt', 'rb') as fp:\n",
    "    clustering_holo_clustercenters = pickle.load(fp)\n",
    "\n",
    "# dtrajs  \n",
    "with open('/msm_holo/dtrajs_ticalag400_clustdefault.txt', 'rb') as fp:\n",
    "    dtrajs_holo = pickle.load(fp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.figure(figsize=(12,9), dpi=120)\n",
    "plt.title('Free Energy and Clusters of CCR2 Holo \\nProjected in tICA Coordinate Space', fontsize=16)\n",
    "\n",
    "plt.subplot(321)\n",
    "plt.xlabel('tIC 0', fontsize=14)\n",
    "plt.ylabel('tIC 1', fontsize=14)\n",
    "\n",
    "_, ax = mplt.plot_free_energy(total_sim_Y[:, 0], total_sim_Y[:, 1])\n",
    "ax.scatter(clustering_holo_clustercenters[:,0], clustering_holo_clustercenters[:,1], color='white', s=10)\n",
    "\n",
    "\n",
    "plt.subplot(322)\n",
    "#plt.title('Free Energy of CCR2 Holo \\nProjected in the First Two tICA Coordinates', fontsize=16)\n",
    "plt.xlabel('tIC 1', fontsize=14)\n",
    "plt.ylabel('tIC 2', fontsize=14)\n",
    "\n",
    "_, ax = mplt.plot_free_energy(total_sim_Y[:, 1], total_sim_Y[:, 2])\n",
    "ax.scatter(clustering_holo_clustercenters[:,1], clustering_holo_clustercenters[:,2], color='white', s=10)\n",
    "\n",
    "\n",
    "plt.subplot(323)\n",
    "#plt.title('Free Energy of CCR2 Holo \\nProjected in the First Two tICA Coordinates', fontsize=16)\n",
    "plt.xlabel('tIC 0', fontsize=14)\n",
    "plt.ylabel('tIC 2', fontsize=14)\n",
    "\n",
    "_, ax = mplt.plot_free_energy(total_sim_Y[:, 0], total_sim_Y[:, 2])\n",
    "ax.scatter(clustering_holo_clustercenters[:,0], clustering_holo_clustercenters[:,2], color='white', s=10)\n",
    "\n",
    "\n",
    "plt.subplot(324)\n",
    "#plt.title('Free Energy of CCR2 Holo \\nProjected in the First Two tICA Coordinates', fontsize=16)\n",
    "plt.xlabel('tIC 0', fontsize=14)\n",
    "plt.ylabel('tIC 3', fontsize=14)\n",
    "\n",
    "_, ax = mplt.plot_free_energy(total_sim_Y[:, 0], total_sim_Y[:, 3])\n",
    "ax.scatter(clustering_holo_clustercenters[:,0], clustering_holo_clustercenters[:,3], color='white', s=10)\n",
    "\n",
    "plt.subplot(325)\n",
    "#plt.title('Free Energy of CCR2 Holo \\nProjected in the First Two tICA Coordinates', fontsize=16)\n",
    "plt.xlabel('tIC 1', fontsize=14)\n",
    "plt.ylabel('tIC 3', fontsize=14)\n",
    "\n",
    "_, ax = mplt.plot_free_energy(total_sim_Y[:, 1], total_sim_Y[:, 3])\n",
    "ax.scatter(clustering_holo_clustercenters[:,1], clustering_holo_clustercenters[:,3], color='white', s=10)\n",
    "\n",
    "plt.subplot(326)\n",
    "#plt.title('Free Energy of CCR2 Holo \\nProjected in the First Two tICA Coordinates', fontsize=16)\n",
    "plt.xlabel('tIC 2', fontsize=14)\n",
    "plt.ylabel('tIC 3', fontsize=14)\n",
    "\n",
    "_, ax = mplt.plot_free_energy(total_sim_Y[:, 2], total_sim_Y[:, 3])\n",
    "ax.scatter(clustering_holo_clustercenters[:,2], clustering_holo_clustercenters[:,3], color='white', s=10)\n",
    "\n",
    "plt.tight_layout()\n",
    "#plt.savefig('/msm_holo/images/01_clustering_ticalag400_FE.png')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.figure(figsize=(12,9), dpi=300)\n",
    "plt.title('Free Energy and Clusters of CCR2 Holo \\nProjected in tICA Coordinate Space', fontsize=16)\n",
    "\n",
    "plt.xlabel('tIC 0', fontsize=14)\n",
    "plt.ylabel('tIC 1', fontsize=14)\n",
    "\n",
    "_, ax = mplt.plot_free_energy(total_sim_Y[:, 0], total_sim_Y[:, 1])\n",
    "ax.scatter(clustering_holo_clustercenters[:,0], clustering_holo_clustercenters[:,1], color='white', s=10)\n",
    "\n",
    "#plt.savefig('/msm_holo/images/01_clustering_ticalag400_FE_network_tic01.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Implied Timescale Plots"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "its_holo_bayes = msm.its(dtrajs_holo, errors='bayes', lags=500)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.figure(figsize=(12,9), dpi=300)\n",
    "dt = 0.24\n",
    "mplt.plot_implied_timescales(its_holo_bayes, units=r'$ns$', nits=10, linewidth=2, dt=dt)\n",
    "plt.title('Implied Timescale Plot with Errors: Holo CCR2')\n",
    "#plt.savefig('/msm_holo/images/02_its_bayes_ns.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Markov Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dt =.24\n",
    "msm_lag = 200 # choose this from ITS plots\n",
    "M_holo_200 = msm.estimate_markov_model(dtrajs_holo, msm_lag, reversible=True)\n",
    "print('fraction of states used = ', M_holo_200.active_state_fraction)\n",
    "print('fraction of counts used = ', M_holo_200.active_count_fraction)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dt =.24\n",
    "msm_lag = 200 # choose this from ITS plots\n",
    "M_holo_200_bayes = msm.bayesian_markov_model(dtrajs_holo, msm_lag, reversible=True)\n",
    "print('fraction of states used = ', M_holo_200_bayes.active_state_fraction)\n",
    "print('fraction of counts used = ', M_holo_200_bayes.active_count_fraction)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#save as\n",
    "\n",
    "with open('/msm_holo/M_holo_emm_ticalag400ns_msmlag200steps.txt', 'wb') as fp:\n",
    "    pickle.dump(M_holo_200, fp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#open up \n",
    "     \n",
    "with open ('/msm_holo/M_holo_emm_ticalag400ns_msmlag200steps.txt', 'rb') as fp:\n",
    "    M_holo_200 = pickle.load(fp)  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CK Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "cktest7 = M_holo_200_bayes.cktest(7, err_est=True, mlags=4)\n",
    "mplt.plot_cktest(cktest7, figsize=(30,30), layout=(2,2), padding_top=0.1, y01=False, padding_between=0.3, units='steps')\n",
    "plt.savefig('/msm_holo/images/05_cktest_7.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## HMM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "nhidden=7\n",
    "\n",
    "estimator = MaximumLikelihoodHMSM(lag=M_holo_200.lagtime, \n",
    "                                  nstates=nhidden, \n",
    "                                  msm_init=M_holo_200, \n",
    "                                  reversible=M_holo_200.reversible,\n",
    "                                  dt_traj=M_holo_200.dt_traj,\n",
    "                                  mincount_connectivity=0\n",
    "                                 )\n",
    "estimator.estimate(M_holo_200.discrete_trajectories_full)\n",
    "hmm = estimator.model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "rates = hmm.P/48*1e6 # 1/ms\n",
    "print(\"Scaled Rates in 1/ms\")\n",
    "print(np.array2string(rates, precision=2))\n",
    "print(\"Probabilities\")\n",
    "print(np.array2string(hmm.P, precision=2))\n",
    "mplt.plot_markov_model(hmm, minflux=1e-25, arrow_labels=rates, arrow_label_format='%10.8f')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Plot FE + Macrostates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "proj_Y = total_sim_Y[:,[0,1]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "metastable_hmm_traj = [hmm.metastable_assignments[traj[:]] for traj in M_holo_200.discrete_trajectories_full]\n",
    "metastable_assign = np.concatenate(metastable_hmm_traj, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "counts = np.zeros(nhidden)\n",
    "macroCenters = np.zeros((nhidden,2))\n",
    "for i in np.arange(0,len(proj_Y)):\n",
    "    state = metastable_assign[i]\n",
    "    counts[state] += 1\n",
    "    macroCenters[state] += proj_Y[i]\n",
    "macroCenters = macroCenters / counts[:,None]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "fig, ax = mplt.plot_free_energy(total_sim_Y[:, 0], total_sim_Y[:, 1], \n",
    "                                kT=0.592172822, \n",
    "                                cbar_label='Free Energy (kcal/mol)',\n",
    "                                avoid_zero_count=True\n",
    "                               )\n",
    "fig.set_size_inches((12,9))\n",
    "ax.set_title('''CCR2 Holo Macrostates''', fontsize=16)\n",
    "ax.set_ylabel('TIC 0', fontsize=14)\n",
    "ax.set_xlabel('TIC 1', fontsize=14)\n",
    "\n",
    "ax.scatter(macroCenters[:,0], macroCenters[:,1], marker='*', s=500, c='w')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "print(np.array2string(macroCenters))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Timescales"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "M_holo_200.timescales(10)/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "hmm.timescales()/1000"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Make FE plot with all clusters in HMM states"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# clustercenters\n",
    "    \n",
    "with open('/msm_holo/clustercenters_ticalag400_clustdefault_net2.txt', 'rb') as fp:\n",
    "    clustering_holo_clustercenters = pickle.load(fp, encoding='latin1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "hmm.metastable_sets\n",
    "\n",
    "state0 = hmm.metastable_sets[0]\n",
    "state1 = hmm.metastable_sets[1]\n",
    "state2 = hmm.metastable_sets[2]\n",
    "state3 = hmm.metastable_sets[3]\n",
    "state4 = hmm.metastable_sets[4]\n",
    "state5 = hmm.metastable_sets[5]\n",
    "state6 = hmm.metastable_sets[6]\n",
    "len(state6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "fig, ax, misc = mplt.plot_free_energy(total_sim_Y[:, 0], total_sim_Y[:, 1], \n",
    "                                kT=0.592172822, \n",
    "                                cbar_label='Free Energy (kcal/mol)',\n",
    "                                avoid_zero_count=True,\n",
    "                                legacy=False\n",
    "                               )\n",
    "\n",
    "fig.set_size_inches((12,9))\n",
    "fig.set_dpi(300)\n",
    "# ax.set_title('''Holo''', fontsize=16)\n",
    "ax.set_ylabel('Holo TIC 1', fontsize=36)\n",
    "ax.set_xlabel('Holo TIC 0', fontsize=36)\n",
    "misc['cbar'].ax.tick_params(labelsize=36)\n",
    "misc['cbar'].set_label('Free Energy (kcal/mol)',fontsize=36)\n",
    "plt.tick_params(axis='both', which='major', labelsize=36)\n",
    "alpha=.8\n",
    "\n",
    "for k in range(22):\n",
    "    ax.plot(clustering_holo_clustercenters[state0[k],0], clustering_holo_clustercenters[state0[k],1], linewidth=20, color='white', marker='o', markersize=10, alpha=alpha)\n",
    "for k in range(252):\n",
    "    ax.plot(clustering_holo_clustercenters[state1[k],0], clustering_holo_clustercenters[state1[k],1], linewidth=20, color='red', marker='o', markersize=10, alpha=alpha)\n",
    "for k in range(54):\n",
    "    ax.plot(clustering_holo_clustercenters[state2[k],0], clustering_holo_clustercenters[state2[k],1], linewidth=20, color='yellow', marker='o', markersize=10, alpha=alpha)    \n",
    "for k in range(74):\n",
    "    ax.plot(clustering_holo_clustercenters[state3[k],0], clustering_holo_clustercenters[state3[k],1], linewidth=20, color='cyan', marker='o', markersize=10, alpha=alpha)   \n",
    "for k in range(239):\n",
    "    ax.plot(clustering_holo_clustercenters[state4[k],0], clustering_holo_clustercenters[state4[k],1], linewidth=20, color='grey', marker='o', markersize=10, alpha=alpha)    \n",
    "for k in range(114):\n",
    "    ax.plot(clustering_holo_clustercenters[state5[k],0], clustering_holo_clustercenters[state5[k],1], linewidth=20, color='purple', marker='o', markersize=10, alpha=alpha)    \n",
    "for k in range(35):\n",
    "    ax.plot(clustering_holo_clustercenters[state6[k],0], clustering_holo_clustercenters[state6[k],1], linewidth=20, color='black', marker='o', markersize=10, alpha=alpha)\n",
    "    \n",
    "    \n",
    "ax.scatter(macroCenters[0,0], macroCenters[0,1], marker='o', s=1600, c='white', linewidths=2, edgecolors='black', zorder=99)\n",
    "ax.scatter(macroCenters[1,0], macroCenters[1,1], marker='o', s=1600, c='red', linewidths=2, edgecolors='black', zorder=99)\n",
    "ax.scatter(macroCenters[2,0], macroCenters[2,1], marker='o', s=1600, c='yellow', linewidths=2, edgecolors='black', zorder=99)\n",
    "ax.scatter(macroCenters[3,0], macroCenters[3,1], marker='o', s=1600, c='cyan', linewidths=2, edgecolors='black', zorder=99)\n",
    "ax.scatter(macroCenters[4,0], macroCenters[4,1], marker='o', s=1600, c='grey', linewidths=2, edgecolors='black', zorder=99)\n",
    "ax.scatter(macroCenters[5,0], macroCenters[5,1], marker='o', s=1600, c='purple', linewidths=2, edgecolors='black', zorder=99)\n",
    "ax.scatter(macroCenters[6,0], macroCenters[6,1], marker='o', s=1600, c='black', linewidths=2, edgecolors='white', zorder=99)\n",
    "\n",
    "\n",
    "fig.savefig('/apo_holo_msm_images/holo_hmm.eps', dpi=300, format='eps')"
   ]
  }
 ],
 "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}