diff --git a/algorithms/SGDClassifier-LDA-htcai.ipynb b/algorithms/SGDClassifier-LDA-htcai.ipynb
new file mode 100644
index 0000000..ce5d47d
--- /dev/null
+++ b/algorithms/SGDClassifier-LDA-htcai.ipynb
@@ -0,0 +1,839 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "# Create a logistic regression model to predict TP53 mutation from gene expression data in TCGA"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [],
+ "source": [
+ "import os\n",
+ "import urllib\n",
+ "import random\n",
+ "import warnings\n",
+ "import resource\n",
+ "\n",
+ "import pandas as pd\n",
+ "import numpy as np\n",
+ "import matplotlib.pyplot as plt\n",
+ "import seaborn as sns\n",
+ "from sklearn import preprocessing, grid_search\n",
+ "from sklearn.linear_model import SGDClassifier\n",
+ "from sklearn.cross_validation import train_test_split\n",
+ "from sklearn.metrics import roc_auc_score, roc_curve\n",
+ "from sklearn.pipeline import make_pipeline\n",
+ "from sklearn.preprocessing import StandardScaler\n",
+ "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis\n",
+ "from sklearn.decomposition import PCA\n",
+ "from statsmodels.robust.scale import mad"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [],
+ "source": [
+ "%matplotlib inline\n",
+ "plt.style.use('seaborn-notebook')"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Specify model configuration"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 3,
+ "metadata": {
+ "collapsed": true
+ },
+ "outputs": [],
+ "source": [
+ "# We're going to be building a 'TP53' classifier \n",
+ "GENE = '7157' # TP53"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 4,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [],
+ "source": [
+ "# Parameter Sweep for Hyperparameters\n",
+ "n_feature_pca = 300\n",
+ "n_feature_kept = 10\n",
+ "param_fixed = {\n",
+ " 'loss': 'log',\n",
+ " 'penalty': 'elasticnet',\n",
+ "}\n",
+ "param_grid = {\n",
+ " 'alpha': [10 ** x for x in range(-5, 2)],\n",
+ " 'l1_ratio': [0, 0.05, 0.1, 0.2, 0.5, 0.8, 0.9, 0.95, 1],\n",
+ "}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "*Here is some [documentation](http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.SGDClassifier.html) regarding the classifier and hyperparameters*\n",
+ "\n",
+ "*Here is some [information](https://ghr.nlm.nih.gov/gene/TP53) about TP53*"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Load Data"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 5,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "CPU times: user 2min 48s, sys: 4.95 s, total: 2min 53s\n",
+ "Wall time: 2min 55s\n"
+ ]
+ }
+ ],
+ "source": [
+ "%%time\n",
+ "path = os.path.join('..', 'download', 'expression-matrix.tsv.bz2')\n",
+ "X = pd.read_table(path, index_col=0)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 6,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "CPU times: user 1min 33s, sys: 3.99 s, total: 1min 37s\n",
+ "Wall time: 1min 38s\n"
+ ]
+ }
+ ],
+ "source": [
+ "%%time\n",
+ "path = os.path.join('..', 'download', 'mutation-matrix.tsv.bz2')\n",
+ "Y = pd.read_table(path, index_col=0)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 7,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [],
+ "source": [
+ "y = Y[GENE]"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 8,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "data": {
+ "text/plain": [
+ "sample_id\n",
+ "TCGA-02-0047-01 0\n",
+ "TCGA-02-0055-01 1\n",
+ "TCGA-02-2483-01 1\n",
+ "TCGA-02-2485-01 1\n",
+ "TCGA-02-2486-01 0\n",
+ "TCGA-04-1348-01 1\n",
+ "Name: 7157, dtype: int64"
+ ]
+ },
+ "execution_count": 8,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "# The Series now holds TP53 Mutation Status for each Sample\n",
+ "y.head(6)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 9,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "data": {
+ "text/plain": [
+ "0 0.645907\n",
+ "1 0.354093\n",
+ "Name: 7157, dtype: float64"
+ ]
+ },
+ "execution_count": 9,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "# Here are the percentage of tumors with NF1\n",
+ "y.value_counts(True)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Set aside 10% of the data for testing"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 10,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "data": {
+ "text/plain": [
+ "'Size: 20,530 features, 6,575 training samples, 731 testing samples'"
+ ]
+ },
+ "execution_count": 10,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "# Typically, this can only be done where the number of mutations is large enough\n",
+ "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=0)\n",
+ "'Size: {:,} features, {:,} training samples, {:,} testing samples'.format(len(X.columns), len(X_train), len(X_test))"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Reduce the dimensionality via Principal Component Analysis and Linear Discriminant Analysis"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 11,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [],
+ "source": [
+ "pca = PCA(n_components=n_feature_pca)\n",
+ "lda = LinearDiscriminantAnalysis(n_components=n_feature_kept)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Define pipeline and Cross validation model fitting"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 12,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [],
+ "source": [
+ "# Include loss='log' in param_grid doesn't work with pipeline somehow\n",
+ "clf = SGDClassifier(random_state=0, class_weight='balanced',\n",
+ " loss=param_fixed['loss'], penalty=param_fixed['penalty'])\n",
+ "\n",
+ "# joblib is used to cross-validate in parallel by setting `n_jobs=-1` in GridSearchCV\n",
+ "# Supress joblib warning. See https://github.com/scikit-learn/scikit-learn/issues/6370\n",
+ "warnings.filterwarnings('ignore', message='Changing the shape of non-C contiguous array')\n",
+ "clf_grid = grid_search.GridSearchCV(estimator=clf, param_grid=param_grid, n_jobs=-1, scoring='roc_auc')\n",
+ "pipeline = make_pipeline(\n",
+ " StandardScaler(), # Feature scaling\n",
+ " pca, # Dimensionality reduction via PCA\n",
+ " lda, # Dimensionality reduciton via LDA\n",
+ " clf_grid)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 13,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "The max memory usage is 4.957 GB\n",
+ "CPU times: user 10min 13s, sys: 15.8 s, total: 10min 29s\n",
+ "Wall time: 5min 32s\n"
+ ]
+ }
+ ],
+ "source": [
+ "%%time\n",
+ "# Fit the model (the computationally intensive part)\n",
+ "pipeline.fit(X=X_train, y=y_train)\n",
+ "best_clf = clf_grid.best_estimator_\n",
+ "print(\"The max memory usage is {:.3f} GB\".format(resource.getrusage(resource.RUSAGE_SELF).ru_maxrss/(1024**3)))"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 14,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "data": {
+ "text/plain": [
+ "{'alpha': 1e-05, 'l1_ratio': 0}"
+ ]
+ },
+ "execution_count": 14,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "clf_grid.best_params_"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 15,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "data": {
+ "text/plain": [
+ "SGDClassifier(alpha=1e-05, average=False, class_weight='balanced',\n",
+ " epsilon=0.1, eta0=0.0, fit_intercept=True, l1_ratio=0,\n",
+ " learning_rate='optimal', loss='log', n_iter=5, n_jobs=1,\n",
+ " penalty='elasticnet', power_t=0.5, random_state=0, shuffle=True,\n",
+ " verbose=0, warm_start=False)"
+ ]
+ },
+ "execution_count": 15,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "best_clf"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Visualize hyperparameters performance"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 16,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [],
+ "source": [
+ "def grid_scores_to_df(grid_scores):\n",
+ " \"\"\"\n",
+ " Convert a sklearn.grid_search.GridSearchCV.grid_scores_ attribute to \n",
+ " a tidy pandas DataFrame where each row is a hyperparameter-fold combinatination.\n",
+ " \"\"\"\n",
+ " rows = list()\n",
+ " for grid_score in grid_scores:\n",
+ " for fold, score in enumerate(grid_score.cv_validation_scores):\n",
+ " row = grid_score.parameters.copy()\n",
+ " row['fold'] = fold\n",
+ " row['score'] = score\n",
+ " rows.append(row)\n",
+ " df = pd.DataFrame(rows)\n",
+ " return df"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {},
+ "source": [
+ "## Process Mutation Matrix"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 17,
+ "metadata": {
+ "collapsed": false
+ },
+ "outputs": [
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "
\n",
+ " \n",
+ " \n",
+ " | \n",
+ " alpha | \n",
+ " fold | \n",
+ " l1_ratio | \n",
+ " score | \n",
+ "
\n",
+ " \n",
+ " \n",
+ " \n",
+ " | 0 | \n",
+ " 0.00001 | \n",
+ " 0 | \n",
+ " 0.0 | \n",
+ " 0.945005 | \n",
+ "
\n",
+ " \n",
+ " | 1 | \n",
+ " 0.00001 | \n",
+ " 1 | \n",
+ " 0.0 | \n",
+ " 0.947590 | \n",
+ "
\n",
+ " \n",
+ "
\n",
+ "
\n",
+ "
\n",
+ " \n",
+ " \n",
+ " | \n",
+ " sample_id | \n",
+ " testing | \n",
+ " status | \n",
+ " decision_function | \n",
+ " probability | \n",
+ " probability_str | \n",
+ "
\n",
+ " \n",
+ " | sample_id | \n",
+ " | \n",
+ " | \n",
+ " | \n",
+ " | \n",
+ " | \n",
+ " | \n",
+ "
\n",
+ " \n",
+ " \n",
+ " \n",
+ " | TCGA-EI-6513-01 | \n",
+ " TCGA-EI-6513-01 | \n",
+ " 1 | \n",
+ " 0 | \n",
+ " 21.435377 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ " | TCGA-CK-5912-01 | \n",
+ " TCGA-CK-5912-01 | \n",
+ " 0 | \n",
+ " 0 | \n",
+ " 19.463339 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ " | TCGA-L5-A4OH-01 | \n",
+ " TCGA-L5-A4OH-01 | \n",
+ " 0 | \n",
+ " 0 | \n",
+ " 19.094222 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ " | TCGA-CQ-A4CI-01 | \n",
+ " TCGA-CQ-A4CI-01 | \n",
+ " 0 | \n",
+ " 0 | \n",
+ " 19.046985 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ " | TCGA-G2-A2EF-01 | \n",
+ " TCGA-G2-A2EF-01 | \n",
+ " 1 | \n",
+ " 0 | \n",
+ " 18.318575 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ " | TCGA-BA-5149-01 | \n",
+ " TCGA-BA-5149-01 | \n",
+ " 0 | \n",
+ " 0 | \n",
+ " 17.999266 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ " | TCGA-L5-A8NR-01 | \n",
+ " TCGA-L5-A8NR-01 | \n",
+ " 0 | \n",
+ " 0 | \n",
+ " 17.902813 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ " | TCGA-46-3765-01 | \n",
+ " TCGA-46-3765-01 | \n",
+ " 0 | \n",
+ " 0 | \n",
+ " 17.706273 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ " | TCGA-CV-6954-01 | \n",
+ " TCGA-CV-6954-01 | \n",
+ " 0 | \n",
+ " 0 | \n",
+ " 16.745752 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ " | TCGA-IK-7675-01 | \n",
+ " TCGA-IK-7675-01 | \n",
+ " 1 | \n",
+ " 0 | \n",
+ " 16.720051 | \n",
+ " 1.0 | \n",
+ " 100.0% | \n",
+ "
\n",
+ " \n",
+ "
\n",
+ "