diff --git a/examples/class1_allele_specific_models.ipynb b/examples/class1_allele_specific_models.ipynb
deleted file mode 100644
index e6857e7c23eada48232dcbcf595a0d294ec279a2..0000000000000000000000000000000000000000
--- a/examples/class1_allele_specific_models.ipynb
+++ /dev/null
@@ -1,1976 +0,0 @@
-{
- "cells": [
- {
- "cell_type": "code",
- "execution_count": 22,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "name": "stdout",
- "output_type": "stream",
- "text": [
- "MHCflurry version: 1.0.0\n"
- ]
- }
- ],
- "source": [
- "import pandas\n",
- "import numpy\n",
- "import seaborn\n",
- "import logging\n",
- "from matplotlib import pyplot\n",
- "\n",
- "import mhcflurry\n",
- "\n",
- "print(\"MHCflurry version: %s\" % (mhcflurry.__version__))"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "# Download data and models"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 2,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "name": "stdout",
- "output_type": "stream",
- "text": [
- "Fetching 0/6 downloads from release 1.0.0\r\n",
- "DOWNLOAD NAME ALREADY DOWNLOADED? WILL DOWNLOAD NOW? URL \r\n",
- "models_class1 YES NO http://github.com/hammerlab/mhcflurry/releases/download/pre-1.0/models_class1.tar.bz2 \r\n",
- "models_class1_experiments1 NO NO http://github.com/hammerlab/mhcflurry/releases/download/pre-1.0/models_class1_experiments1.tar.bz2 \r\n",
- "cross_validation_class1 NO NO http://github.com/hammerlab/mhcflurry/releases/download/pre-1.0/cross_validation_class1.tar.bz2 \r\n",
- "data_iedb NO NO https://github.com/hammerlab/mhcflurry/releases/download/pre-1.0/data_iedb.tar.bz2 \r\n",
- "data_kim2014 NO NO http://github.com/hammerlab/mhcflurry/releases/download/0.9.1/data_kim2014.tar.bz2 \r\n",
- "data_curated YES NO https://github.com/hammerlab/mhcflurry/releases/download/pre-1.0/data_curated.tar.bz2 \r\n"
- ]
- }
- ],
- "source": [
- "!mhcflurry-downloads fetch"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "# Making predictions with `Class1AffinityPredictor`"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 3,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "name": "stdout",
- "output_type": "stream",
- "text": [
- "Help on class Class1AffinityPredictor in module mhcflurry.class1_affinity_predictor:\n",
- "\n",
- "class Class1AffinityPredictor(builtins.object)\n",
- " | High-level interface for peptide/MHC I binding affinity prediction.\n",
- " | \n",
- " | This is the class most users will want to use.\n",
- " | \n",
- " | This class delegates to one or more `Class1NeuralNetwork` instances.\n",
- " | It supports prediction across multiple alleles using ensembles of single-\n",
- " | or pan-allele predictors.\n",
- " | \n",
- " | Methods defined here:\n",
- " | \n",
- " | __init__(self, allele_to_allele_specific_models=None, class1_pan_allele_models=None, allele_to_pseudosequence=None, manifest_df=None, allele_to_percent_rank_transform=None)\n",
- " | Parameters\n",
- " | ----------\n",
- " | allele_to_allele_specific_models : dict of string -> list of Class1NeuralNetwork\n",
- " | Ensemble of single-allele models to use for each allele. \n",
- " | \n",
- " | class1_pan_allele_models : list of Class1NeuralNetwork\n",
- " | Ensemble of pan-allele models.\n",
- " | \n",
- " | allele_to_pseudosequence : dict of string -> string\n",
- " | Required only if class1_pan_allele_models is specified.\n",
- " | \n",
- " | manifest_df : pandas.DataFrame, optional\n",
- " | Must have columns: model_name, allele, config_json, model.\n",
- " | Only required if you want to update an existing serialization of a\n",
- " | Class1AffinityPredictor. Otherwise this dataframe will be generated\n",
- " | automatically based on the supplied models.\n",
- " | \n",
- " | allele_to_percent_rank_transform : dict of string -> PercentRankTransform, optional\n",
- " | PercentRankTransform instances to use for each allele\n",
- " | \n",
- " | calibrate_percentile_ranks(self, peptides=None, num_peptides_per_length=100000, alleles=None, bins=None, quiet=False)\n",
- " | Compute the cumulative distribution of ic50 values for a set of alleles\n",
- " | over a large universe of random peptides, to enable computing quantiles in\n",
- " | this distribution later.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | peptides : sequence of string, optional\n",
- " | Peptides to use\n",
- " | num_peptides_per_length : int, optional\n",
- " | If peptides argument is not specified, then num_peptides_per_length\n",
- " | peptides are randomly sampled from a uniform distribution for each\n",
- " | supported length\n",
- " | alleles : sequence of string, optional\n",
- " | Alleles to perform calibration for. If not specified all supported\n",
- " | alleles will be calibrated.\n",
- " | bins : object\n",
- " | Anything that can be passed to numpy.histogram's \"bins\" argument\n",
- " | can be used here, i.e. either an integer or a sequence giving bin\n",
- " | edges. This is in ic50 space.\n",
- " | quiet : boolean\n",
- " | If False (default), status updates will be printed to stdout.\n",
- " | \n",
- " | fit_allele_specific_predictors(self, n_models, architecture_hyperparameters, allele, peptides, affinities, models_dir_for_save=None, verbose=1, progress_preamble='')\n",
- " | Fit one or more allele specific predictors for a single allele using a\n",
- " | single neural network architecture.\n",
- " | \n",
- " | The new predictors are saved in the Class1AffinityPredictor instance\n",
- " | and will be used on subsequent calls to `predict`.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | n_models : int\n",
- " | Number of neural networks to fit\n",
- " | \n",
- " | architecture_hyperparameters : dict \n",
- " | \n",
- " | allele : string\n",
- " | \n",
- " | peptides : EncodableSequences or list of string\n",
- " | \n",
- " | affinities : list of float\n",
- " | nM affinities\n",
- " | \n",
- " | models_dir_for_save : string, optional\n",
- " | If specified, the Class1AffinityPredictor is (incrementally) written\n",
- " | to the given models dir after each neural network is fit.\n",
- " | \n",
- " | verbose : int\n",
- " | Keras verbosity\n",
- " | \n",
- " | progress_preamble : string\n",
- " | Optional string of information to include in each progress update\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | list of Class1NeuralNetwork\n",
- " | \n",
- " | fit_class1_pan_allele_models(self, n_models, architecture_hyperparameters, alleles, peptides, affinities, models_dir_for_save=None, verbose=1, progress_preamble='')\n",
- " | Fit one or more pan-allele predictors using a single neural network\n",
- " | architecture.\n",
- " | \n",
- " | The new predictors are saved in the Class1AffinityPredictor instance\n",
- " | and will be used on subsequent calls to `predict`.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | n_models : int\n",
- " | Number of neural networks to fit\n",
- " | \n",
- " | architecture_hyperparameters : dict\n",
- " | \n",
- " | alleles : list of string\n",
- " | Allele names (not pseudosequences) corresponding to each peptide \n",
- " | \n",
- " | peptides : EncodableSequences or list of string\n",
- " | \n",
- " | affinities : list of float\n",
- " | nM affinities\n",
- " | \n",
- " | models_dir_for_save : string, optional\n",
- " | If specified, the Class1AffinityPredictor is (incrementally) written\n",
- " | to the given models dir after each neural network is fit.\n",
- " | \n",
- " | verbose : int\n",
- " | Keras verbosity\n",
- " | \n",
- " | progress_preamble : string\n",
- " | Optional string of information to include in each progress update\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | list of Class1NeuralNetwork\n",
- " | \n",
- " | percentile_ranks(self, affinities, allele=None, alleles=None, throw=True)\n",
- " | Return percentile ranks for the given ic50 affinities and alleles.\n",
- " | \n",
- " | The 'allele' and 'alleles' argument are as in the predict() method.\n",
- " | Specify one of these.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | affinities : sequence of float\n",
- " | nM affinities\n",
- " | allele : string\n",
- " | alleles : sequence of string\n",
- " | throw : boolean\n",
- " | If True, a ValueError will be raised in the case of unsupported\n",
- " | alleles. If False, a warning will be logged and NaN will be returned\n",
- " | for those percentile ranks.\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | numpy.array of float\n",
- " | \n",
- " | predict(self, peptides, alleles=None, allele=None, throw=True)\n",
- " | Predict nM binding affinities.\n",
- " | \n",
- " | If multiple predictors are available for an allele, the predictions are\n",
- " | the geometric means of the individual model predictions.\n",
- " | \n",
- " | One of 'allele' or 'alleles' must be specified. If 'allele' is specified\n",
- " | all predictions will be for the given allele. If 'alleles' is specified\n",
- " | it must be the same length as 'peptides' and give the allele\n",
- " | corresponding to each peptide.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | peptides : EncodableSequences or list of string\n",
- " | alleles : list of string\n",
- " | allele : string\n",
- " | throw : boolean\n",
- " | If True, a ValueError will be raised in the case of unsupported\n",
- " | alleles or peptide lengths. If False, a warning will be logged and\n",
- " | the predictions for the unsupported alleles or peptides will be NaN.\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | numpy.array of predictions\n",
- " | \n",
- " | predict_to_dataframe(self, peptides, alleles=None, allele=None, throw=True, include_individual_model_predictions=False, include_percentile_ranks=True)\n",
- " | Predict nM binding affinities. Gives more detailed output than `predict`\n",
- " | method, including 5-95% prediction intervals.\n",
- " | \n",
- " | If multiple predictors are available for an allele, the predictions are\n",
- " | the geometric means of the individual model predictions.\n",
- " | \n",
- " | One of 'allele' or 'alleles' must be specified. If 'allele' is specified\n",
- " | all predictions will be for the given allele. If 'alleles' is specified\n",
- " | it must be the same length as 'peptides' and give the allele\n",
- " | corresponding to each peptide. \n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | peptides : EncodableSequences or list of string\n",
- " | alleles : list of string\n",
- " | allele : string\n",
- " | throw : boolean\n",
- " | If True, a ValueError will be raised in the case of unsupported\n",
- " | alleles or peptide lengths. If False, a warning will be logged and\n",
- " | the predictions for the unsupported alleles or peptides will be NaN.\n",
- " | include_individual_model_predictions : boolean\n",
- " | If True, the predictions of each individual model are included as\n",
- " | columns in the result dataframe.\n",
- " | include_percentile_ranks : boolean, default True\n",
- " | If True, a \"prediction_percentile\" column will be included giving the\n",
- " | percentile ranks. If no percentile rank information is available,\n",
- " | this will be ignored with a warning.\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | pandas.DataFrame of predictions\n",
- " | \n",
- " | save(self, models_dir, model_names_to_write=None)\n",
- " | Serialize the predictor to a directory on disk.\n",
- " | \n",
- " | The serialization format consists of a file called \"manifest.csv\" with\n",
- " | the configurations of each Class1NeuralNetwork, along with per-network\n",
- " | files giving the model weights. If there are pan-allele predictors in\n",
- " | the ensemble, the allele pseudosequences are also stored in the\n",
- " | directory.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | models_dir : string\n",
- " | Path to directory\n",
- " | \n",
- " | model_names_to_write : list of string, optional\n",
- " | Only write the weights for the specified models. Useful for\n",
- " | incremental updates during training.\n",
- " | \n",
- " | ----------------------------------------------------------------------\n",
- " | Class methods defined here:\n",
- " | \n",
- " | merge(predictors) from builtins.type\n",
- " | Merge the ensembles of two or more Class1AffinityPredictor instances.\n",
- " | \n",
- " | Note: the resulting merged predictor will NOT have calibrated percentile\n",
- " | ranks. Call calibrate_percentile_ranks() on it if these are needed.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | predictors : sequence of Class1AffinityPredictor\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | Class1AffinityPredictor\n",
- " | \n",
- " | ----------------------------------------------------------------------\n",
- " | Static methods defined here:\n",
- " | \n",
- " | load(models_dir=None, max_models=None)\n",
- " | Deserialize a predictor from a directory on disk.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | models_dir : string\n",
- " | Path to directory\n",
- " | \n",
- " | max_models : int, optional\n",
- " | Maximum number of Class1NeuralNetwork instances to load\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | Class1AffinityPredictor\n",
- " | \n",
- " | load_weights(filename)\n",
- " | Restore model weights from the given filename, which should have been\n",
- " | created with `save_weights`.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | filename : string\n",
- " | Should end in \".npz\".\n",
- " | \n",
- " | \n",
- " | Returns\n",
- " | ----------\n",
- " | \n",
- " | list of array\n",
- " | \n",
- " | model_name(allele, num)\n",
- " | Generate a model name\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | allele : string\n",
- " | num : int\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | string\n",
- " | \n",
- " | save_weights(weights_list, filename)\n",
- " | Save the model weights to the given filename using numpy's \".npz\"\n",
- " | format.\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | weights_list : list of array\n",
- " | \n",
- " | filename : string\n",
- " | Should end in \".npz\".\n",
- " | \n",
- " | weights_path(models_dir, model_name)\n",
- " | Generate the path to the weights file for a model\n",
- " | \n",
- " | Parameters\n",
- " | ----------\n",
- " | models_dir : string\n",
- " | model_name : string\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | string\n",
- " | \n",
- " | ----------------------------------------------------------------------\n",
- " | Data descriptors defined here:\n",
- " | \n",
- " | __dict__\n",
- " | dictionary for instance variables (if defined)\n",
- " | \n",
- " | __weakref__\n",
- " | list of weak references to the object (if defined)\n",
- " | \n",
- " | neural_networks\n",
- " | List of the neural networks in the ensemble.\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | list of Class1NeuralNetwork\n",
- " | \n",
- " | num_networks\n",
- " | \n",
- " | supported_alleles\n",
- " | Alleles for which predictions can be made.\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | list of string\n",
- " | \n",
- " | supported_peptide_lengths\n",
- " | (minimum, maximum) lengths of peptides supported by *all models*,\n",
- " | inclusive.\n",
- " | \n",
- " | Returns\n",
- " | -------\n",
- " | (int, int) tuple\n",
- "\n"
- ]
- }
- ],
- "source": [
- "help(mhcflurry.Class1AffinityPredictor)"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 4,
- "metadata": {
- "collapsed": false
- },
- "outputs": [],
- "source": [
- "downloaded_predictor = mhcflurry.Class1AffinityPredictor.load()"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 5,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "name": "stderr",
- "output_type": "stream",
- "text": [
- "Using TensorFlow backend.\n"
- ]
- },
- {
- "data": {
- "text/plain": [
- "array([ 4899.04784343, 5685.25682682])"
- ]
- },
- "execution_count": 5,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "downloaded_predictor.predict(allele=\"HLA-A0201\", peptides=[\"SIINFEKL\", \"SIINFEQL\"])"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 6,
- "metadata": {
- "collapsed": false
- },
- "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>allele</th>\n",
- " <th>peptide</th>\n",
- " <th>prediction</th>\n",
- " <th>prediction_low</th>\n",
- " <th>prediction_high</th>\n",
- " <th>prediction_percentile</th>\n",
- " </tr>\n",
- " </thead>\n",
- " <tbody>\n",
- " <tr>\n",
- " <th>0</th>\n",
- " <td>HLA-A0201</td>\n",
- " <td>SIINFEKL</td>\n",
- " <td>4899.047843</td>\n",
- " <td>2767.763654</td>\n",
- " <td>7269.683643</td>\n",
- " <td>6.509787</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>1</th>\n",
- " <td>HLA-A0201</td>\n",
- " <td>SIINFEQL</td>\n",
- " <td>5685.256827</td>\n",
- " <td>3815.923563</td>\n",
- " <td>7476.714466</td>\n",
- " <td>7.436687</td>\n",
- " </tr>\n",
- " </tbody>\n",
- "</table>\n",
- "</div>"
- ],
- "text/plain": [
- " allele peptide prediction prediction_low prediction_high \\\n",
- "0 HLA-A0201 SIINFEKL 4899.047843 2767.763654 7269.683643 \n",
- "1 HLA-A0201 SIINFEQL 5685.256827 3815.923563 7476.714466 \n",
- "\n",
- " prediction_percentile \n",
- "0 6.509787 \n",
- "1 7.436687 "
- ]
- },
- "execution_count": 6,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "downloaded_predictor.predict_to_dataframe(allele=\"HLA-A0201\", peptides=[\"SIINFEKL\", \"SIINFEQL\"])"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 7,
- "metadata": {
- "collapsed": false
- },
- "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>allele</th>\n",
- " <th>peptide</th>\n",
- " <th>prediction</th>\n",
- " <th>prediction_low</th>\n",
- " <th>prediction_high</th>\n",
- " <th>prediction_percentile</th>\n",
- " </tr>\n",
- " </thead>\n",
- " <tbody>\n",
- " <tr>\n",
- " <th>0</th>\n",
- " <td>HLA-A0201</td>\n",
- " <td>SIINFEKL</td>\n",
- " <td>4899.047942</td>\n",
- " <td>2767.763654</td>\n",
- " <td>7269.683643</td>\n",
- " <td>6.509787</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>1</th>\n",
- " <td>HLA-B*57:01</td>\n",
- " <td>SIINFEQL</td>\n",
- " <td>26704.220115</td>\n",
- " <td>23198.059394</td>\n",
- " <td>30635.114610</td>\n",
- " <td>47.611925</td>\n",
- " </tr>\n",
- " </tbody>\n",
- "</table>\n",
- "</div>"
- ],
- "text/plain": [
- " allele peptide prediction prediction_low prediction_high \\\n",
- "0 HLA-A0201 SIINFEKL 4899.047942 2767.763654 7269.683643 \n",
- "1 HLA-B*57:01 SIINFEQL 26704.220115 23198.059394 30635.114610 \n",
- "\n",
- " prediction_percentile \n",
- "0 6.509787 \n",
- "1 47.611925 "
- ]
- },
- "execution_count": 7,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "downloaded_predictor.predict_to_dataframe(alleles=[\"HLA-A0201\", \"HLA-B*57:01\"], peptides=[\"SIINFEKL\", \"SIINFEQL\"])"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 8,
- "metadata": {
- "collapsed": false
- },
- "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>allele</th>\n",
- " <th>peptide</th>\n",
- " <th>prediction</th>\n",
- " <th>prediction_low</th>\n",
- " <th>prediction_high</th>\n",
- " <th>model_single_0</th>\n",
- " <th>model_single_1</th>\n",
- " <th>model_single_2</th>\n",
- " <th>model_single_3</th>\n",
- " <th>model_single_4</th>\n",
- " <th>model_single_5</th>\n",
- " <th>model_single_6</th>\n",
- " <th>model_single_7</th>\n",
- " <th>prediction_percentile</th>\n",
- " </tr>\n",
- " </thead>\n",
- " <tbody>\n",
- " <tr>\n",
- " <th>0</th>\n",
- " <td>HLA-A0201</td>\n",
- " <td>SIINFEKL</td>\n",
- " <td>4899.047843</td>\n",
- " <td>2767.763654</td>\n",
- " <td>7269.683643</td>\n",
- " <td>5245.313773</td>\n",
- " <td>4131.368053</td>\n",
- " <td>4599.034976</td>\n",
- " <td>7350.344042</td>\n",
- " <td>2230.774145</td>\n",
- " <td>6754.462616</td>\n",
- " <td>4220.768251</td>\n",
- " <td>7122.226730</td>\n",
- " <td>6.509787</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>1</th>\n",
- " <td>HLA-A0201</td>\n",
- " <td>SIINFEQL</td>\n",
- " <td>5685.256827</td>\n",
- " <td>3815.923563</td>\n",
- " <td>7476.714466</td>\n",
- " <td>5711.583293</td>\n",
- " <td>5718.509616</td>\n",
- " <td>5459.967973</td>\n",
- " <td>7709.914896</td>\n",
- " <td>3394.800793</td>\n",
- " <td>7062.179615</td>\n",
- " <td>4741.495660</td>\n",
- " <td>6983.180148</td>\n",
- " <td>7.436687</td>\n",
- " </tr>\n",
- " </tbody>\n",
- "</table>\n",
- "</div>"
- ],
- "text/plain": [
- " allele peptide prediction prediction_low prediction_high \\\n",
- "0 HLA-A0201 SIINFEKL 4899.047843 2767.763654 7269.683643 \n",
- "1 HLA-A0201 SIINFEQL 5685.256827 3815.923563 7476.714466 \n",
- "\n",
- " model_single_0 model_single_1 model_single_2 model_single_3 \\\n",
- "0 5245.313773 4131.368053 4599.034976 7350.344042 \n",
- "1 5711.583293 5718.509616 5459.967973 7709.914896 \n",
- "\n",
- " model_single_4 model_single_5 model_single_6 model_single_7 \\\n",
- "0 2230.774145 6754.462616 4220.768251 7122.226730 \n",
- "1 3394.800793 7062.179615 4741.495660 6983.180148 \n",
- "\n",
- " prediction_percentile \n",
- "0 6.509787 \n",
- "1 7.436687 "
- ]
- },
- "execution_count": 8,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "downloaded_predictor.predict_to_dataframe(\n",
- " allele=\"HLA-A0201\",\n",
- " peptides=[\"SIINFEKL\", \"SIINFEQL\"],\n",
- " include_individual_model_predictions=True)"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 9,
- "metadata": {
- "collapsed": false
- },
- "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>allele</th>\n",
- " <th>peptide</th>\n",
- " <th>prediction</th>\n",
- " <th>prediction_low</th>\n",
- " <th>prediction_high</th>\n",
- " <th>prediction_percentile</th>\n",
- " </tr>\n",
- " </thead>\n",
- " <tbody>\n",
- " <tr>\n",
- " <th>0</th>\n",
- " <td>HLA-A0201</td>\n",
- " <td>SIINFEKL</td>\n",
- " <td>4899.047843</td>\n",
- " <td>2767.763654</td>\n",
- " <td>7269.683643</td>\n",
- " <td>6.509787</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>1</th>\n",
- " <td>HLA-A0201</td>\n",
- " <td>SIINFEQL</td>\n",
- " <td>5685.256827</td>\n",
- " <td>3815.923563</td>\n",
- " <td>7476.714466</td>\n",
- " <td>7.436687</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>2</th>\n",
- " <td>HLA-A0201</td>\n",
- " <td>TAAAALANGGGGGGGG</td>\n",
- " <td>NaN</td>\n",
- " <td>NaN</td>\n",
- " <td>NaN</td>\n",
- " <td>100.000000</td>\n",
- " </tr>\n",
- " </tbody>\n",
- "</table>\n",
- "</div>"
- ],
- "text/plain": [
- " allele peptide prediction prediction_low prediction_high \\\n",
- "0 HLA-A0201 SIINFEKL 4899.047843 2767.763654 7269.683643 \n",
- "1 HLA-A0201 SIINFEQL 5685.256827 3815.923563 7476.714466 \n",
- "2 HLA-A0201 TAAAALANGGGGGGGG NaN NaN NaN \n",
- "\n",
- " prediction_percentile \n",
- "0 6.509787 \n",
- "1 7.436687 \n",
- "2 100.000000 "
- ]
- },
- "execution_count": 9,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "downloaded_predictor.predict_to_dataframe(\n",
- " allele=\"HLA-A0201\",\n",
- " peptides=[\"SIINFEKL\", \"SIINFEQL\", \"TAAAALANGGGGGGGG\"],\n",
- " throw=False) # Without throw=False, you'll get a ValueError for invalid peptides or alleles"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "# Instantiating a `Class1AffinityPredictor` from a saved model on disk"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 10,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "data": {
- "text/plain": [
- "'/Users/tim/Library/Application Support/mhcflurry/4/1.0.0/models_class1/models'"
- ]
- },
- "execution_count": 10,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "models_dir = mhcflurry.downloads.get_path(\"models_class1\", \"models\")\n",
- "models_dir"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 11,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "data": {
- "text/plain": [
- "array([ 28227.29890915, 26568.72745054, 39043.95304442])"
- ]
- },
- "execution_count": 11,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "# This will be the same predictor we instantiated above. We're just being explicit about what models to load.\n",
- "downloaded_predictor = mhcflurry.Class1AffinityPredictor.load(models_dir)\n",
- "downloaded_predictor.predict([\"SIINFEKL\", \"SIQNPEKP\", \"SYNFPEPI\"], allele=\"HLA-A0301\")"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "# Fit a model: first load some data"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 12,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "data": {
- "text/plain": [
- "'/Users/tim/Library/Application Support/mhcflurry/4/1.0.0/data_curated/curated_training_data.csv.bz2'"
- ]
- },
- "execution_count": 12,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "# This is the data the downloaded models were trained on\n",
- "data_path = mhcflurry.downloads.get_path(\"data_curated\", \"curated_training_data.csv.bz2\")\n",
- "data_path"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 13,
- "metadata": {
- "collapsed": false
- },
- "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>allele</th>\n",
- " <th>peptide</th>\n",
- " <th>measurement_value</th>\n",
- " <th>measurement_type</th>\n",
- " <th>measurement_source</th>\n",
- " <th>original_allele</th>\n",
- " </tr>\n",
- " </thead>\n",
- " <tbody>\n",
- " <tr>\n",
- " <th>0</th>\n",
- " <td>BoLA-1*21:01</td>\n",
- " <td>AENDTLVVSV</td>\n",
- " <td>7817.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/competitive/fluorescence</td>\n",
- " <td>BoLA-1*02101</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>1</th>\n",
- " <td>BoLA-1*21:01</td>\n",
- " <td>NQFNGGCLLV</td>\n",
- " <td>1086.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-1*02101</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>2</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>AAHCIHAEW</td>\n",
- " <td>21.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>3</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>AAKHMSNTY</td>\n",
- " <td>1299.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>4</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>DSYAYMRNGW</td>\n",
- " <td>2.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>5</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>HTTNTQNNDW</td>\n",
- " <td>40.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>6</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>KVYANIAPTY</td>\n",
- " <td>10000.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/competitive/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>7</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>KVYNPPRTNY</td>\n",
- " <td>393.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>8</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>LAAKHMSNT</td>\n",
- " <td>1380.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>9</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>LLVAMVPEW</td>\n",
- " <td>2.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>10</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>MTAHITVPY</td>\n",
- " <td>51.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>11</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>RAAHCIHAEW</td>\n",
- " <td>1.7</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>12</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>SSVGVTHGY</td>\n",
- " <td>6.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>13</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>TTNTQNNDW</td>\n",
- " <td>9.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>14</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>TVYNGTSKY</td>\n",
- " <td>5.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>15</th>\n",
- " <td>BoLA-2*08:01</td>\n",
- " <td>VVRHEGNLTW</td>\n",
- " <td>11.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*00801</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>16</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>AAHGMGKVGK</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>17</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>ASFNYGAIK</td>\n",
- " <td>67.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>18</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>ASHGMGKVGK</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>19</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>ASSHGMGKVGK</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>20</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>CIYQITHGK</td>\n",
- " <td>1035.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/direct/fluorescence</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>21</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>GVNRYDQYK</td>\n",
- " <td>8693.0</td>\n",
- " <td>quantitative</td>\n",
- " <td>Barlow - purified MHC/competitive/fluorescence</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>22</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>KAAHGMGKVGK</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>23</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>KAHGMGKVGK</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>24</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>KASHGMGKVGK</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>25</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>KEHGMGKVGK</td>\n",
- " <td>50000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>26</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>KKHGMGKVGK</td>\n",
- " <td>50000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>27</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>KKKHGMGKVGK</td>\n",
- " <td>50000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>28</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>KKSHGMGKVGK</td>\n",
- " <td>50000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>29</th>\n",
- " <td>BoLA-2*12:01</td>\n",
- " <td>KLHGMGKVGK</td>\n",
- " <td>50000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Morrison - cellular MHC/T cell inhibition</td>\n",
- " <td>BoLA-2*01201</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>...</th>\n",
- " <td>...</td>\n",
- " <td>...</td>\n",
- " <td>...</td>\n",
- " <td>...</td>\n",
- " <td>...</td>\n",
- " <td>...</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239917</th>\n",
- " <td>SLA-2*04:01</td>\n",
- " <td>MTAHITVPY</td>\n",
- " <td>50.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Golde - purified MHC</td>\n",
- " <td>SLA-2*0401</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239918</th>\n",
- " <td>SLA-2*04:01</td>\n",
- " <td>NMTAHITVPY</td>\n",
- " <td>50000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Golde - purified MHC</td>\n",
- " <td>SLA-2*0401</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239919</th>\n",
- " <td>SLA-2*04:01</td>\n",
- " <td>NTYLSGIAQY</td>\n",
- " <td>50.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Golde - purified MHC</td>\n",
- " <td>SLA-2*0401</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239920</th>\n",
- " <td>SLA-2*04:01</td>\n",
- " <td>QSSVGVTHGY</td>\n",
- " <td>50000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Golde - purified MHC</td>\n",
- " <td>SLA-2*0401</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239921</th>\n",
- " <td>SLA-2*04:01</td>\n",
- " <td>SSVGVTHGY</td>\n",
- " <td>50.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Golde - purified MHC</td>\n",
- " <td>SLA-2*0401</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239922</th>\n",
- " <td>SLA-2*04:01</td>\n",
- " <td>TVYNGTSKY</td>\n",
- " <td>50000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Golde - purified MHC</td>\n",
- " <td>SLA-2*0401</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239923</th>\n",
- " <td>SLA-2*04:01</td>\n",
- " <td>YLSGIAQYY</td>\n",
- " <td>50.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Golde - purified MHC</td>\n",
- " <td>SLA-2*0401</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239924</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KANTQFTAV</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC/direct/fluorescence</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239925</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KENTQFTAV</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239926</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KFNTQFTAV</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239927</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KINTQFTAV</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239928</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KLNTQFTAV</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239929</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMATQFTAV</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC/direct/fluorescence</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239930</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMETQFTAV</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239931</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMFTQFTAV</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239932</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMNTAFTAV</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC/direct/fluorescence</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239933</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMNTQFTAA</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC/direct/fluorescence</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239934</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMNTQFTAF</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239935</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMNTQFTAI</td>\n",
- " <td>500.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239936</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMNTQFTAL</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239937</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMNTQFTAV</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - x-ray crystallography</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239938</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMNTQFTAV</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239939</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMNTQFTAV</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC/direct/fluorescence</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239940</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMQTQFTAV</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239941</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KMRTQFTAV</td>\n",
- " <td>500.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239942</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KRNTQFTAV</td>\n",
- " <td>500.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239943</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KTNTQFTAV</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239944</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>KVNTQFTAV</td>\n",
- " <td>5000.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239945</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>MQFSSLTV</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " <tr>\n",
- " <th>239946</th>\n",
- " <td>SLA-3*02:02</td>\n",
- " <td>RRNYFTAEV</td>\n",
- " <td>100.0</td>\n",
- " <td>qualitative</td>\n",
- " <td>Xia - purified MHC</td>\n",
- " <td>SLA-3*0202</td>\n",
- " </tr>\n",
- " </tbody>\n",
- "</table>\n",
- "<p>239947 rows × 6 columns</p>\n",
- "</div>"
- ],
- "text/plain": [
- " allele peptide measurement_value measurement_type \\\n",
- "0 BoLA-1*21:01 AENDTLVVSV 7817.0 quantitative \n",
- "1 BoLA-1*21:01 NQFNGGCLLV 1086.0 quantitative \n",
- "2 BoLA-2*08:01 AAHCIHAEW 21.0 quantitative \n",
- "3 BoLA-2*08:01 AAKHMSNTY 1299.0 quantitative \n",
- "4 BoLA-2*08:01 DSYAYMRNGW 2.0 quantitative \n",
- "5 BoLA-2*08:01 HTTNTQNNDW 40.0 quantitative \n",
- "6 BoLA-2*08:01 KVYANIAPTY 10000.0 quantitative \n",
- "7 BoLA-2*08:01 KVYNPPRTNY 393.0 quantitative \n",
- "8 BoLA-2*08:01 LAAKHMSNT 1380.0 quantitative \n",
- "9 BoLA-2*08:01 LLVAMVPEW 2.0 quantitative \n",
- "10 BoLA-2*08:01 MTAHITVPY 51.0 quantitative \n",
- "11 BoLA-2*08:01 RAAHCIHAEW 1.7 quantitative \n",
- "12 BoLA-2*08:01 SSVGVTHGY 6.0 quantitative \n",
- "13 BoLA-2*08:01 TTNTQNNDW 9.0 quantitative \n",
- "14 BoLA-2*08:01 TVYNGTSKY 5.0 quantitative \n",
- "15 BoLA-2*08:01 VVRHEGNLTW 11.0 quantitative \n",
- "16 BoLA-2*12:01 AAHGMGKVGK 100.0 qualitative \n",
- "17 BoLA-2*12:01 ASFNYGAIK 67.0 quantitative \n",
- "18 BoLA-2*12:01 ASHGMGKVGK 100.0 qualitative \n",
- "19 BoLA-2*12:01 ASSHGMGKVGK 100.0 qualitative \n",
- "20 BoLA-2*12:01 CIYQITHGK 1035.0 quantitative \n",
- "21 BoLA-2*12:01 GVNRYDQYK 8693.0 quantitative \n",
- "22 BoLA-2*12:01 KAAHGMGKVGK 100.0 qualitative \n",
- "23 BoLA-2*12:01 KAHGMGKVGK 100.0 qualitative \n",
- "24 BoLA-2*12:01 KASHGMGKVGK 100.0 qualitative \n",
- "25 BoLA-2*12:01 KEHGMGKVGK 50000.0 qualitative \n",
- "26 BoLA-2*12:01 KKHGMGKVGK 50000.0 qualitative \n",
- "27 BoLA-2*12:01 KKKHGMGKVGK 50000.0 qualitative \n",
- "28 BoLA-2*12:01 KKSHGMGKVGK 50000.0 qualitative \n",
- "29 BoLA-2*12:01 KLHGMGKVGK 50000.0 qualitative \n",
- "... ... ... ... ... \n",
- "239917 SLA-2*04:01 MTAHITVPY 50.0 qualitative \n",
- "239918 SLA-2*04:01 NMTAHITVPY 50000.0 qualitative \n",
- "239919 SLA-2*04:01 NTYLSGIAQY 50.0 qualitative \n",
- "239920 SLA-2*04:01 QSSVGVTHGY 50000.0 qualitative \n",
- "239921 SLA-2*04:01 SSVGVTHGY 50.0 qualitative \n",
- "239922 SLA-2*04:01 TVYNGTSKY 50000.0 qualitative \n",
- "239923 SLA-2*04:01 YLSGIAQYY 50.0 qualitative \n",
- "239924 SLA-3*02:02 KANTQFTAV 100.0 qualitative \n",
- "239925 SLA-3*02:02 KENTQFTAV 5000.0 qualitative \n",
- "239926 SLA-3*02:02 KFNTQFTAV 5000.0 qualitative \n",
- "239927 SLA-3*02:02 KINTQFTAV 5000.0 qualitative \n",
- "239928 SLA-3*02:02 KLNTQFTAV 5000.0 qualitative \n",
- "239929 SLA-3*02:02 KMATQFTAV 100.0 qualitative \n",
- "239930 SLA-3*02:02 KMETQFTAV 5000.0 qualitative \n",
- "239931 SLA-3*02:02 KMFTQFTAV 5000.0 qualitative \n",
- "239932 SLA-3*02:02 KMNTAFTAV 100.0 qualitative \n",
- "239933 SLA-3*02:02 KMNTQFTAA 100.0 qualitative \n",
- "239934 SLA-3*02:02 KMNTQFTAF 5000.0 qualitative \n",
- "239935 SLA-3*02:02 KMNTQFTAI 500.0 qualitative \n",
- "239936 SLA-3*02:02 KMNTQFTAL 5000.0 qualitative \n",
- "239937 SLA-3*02:02 KMNTQFTAV 100.0 qualitative \n",
- "239938 SLA-3*02:02 KMNTQFTAV 100.0 qualitative \n",
- "239939 SLA-3*02:02 KMNTQFTAV 100.0 qualitative \n",
- "239940 SLA-3*02:02 KMQTQFTAV 100.0 qualitative \n",
- "239941 SLA-3*02:02 KMRTQFTAV 500.0 qualitative \n",
- "239942 SLA-3*02:02 KRNTQFTAV 500.0 qualitative \n",
- "239943 SLA-3*02:02 KTNTQFTAV 5000.0 qualitative \n",
- "239944 SLA-3*02:02 KVNTQFTAV 5000.0 qualitative \n",
- "239945 SLA-3*02:02 MQFSSLTV 100.0 qualitative \n",
- "239946 SLA-3*02:02 RRNYFTAEV 100.0 qualitative \n",
- "\n",
- " measurement_source original_allele \n",
- "0 Barlow - purified MHC/competitive/fluorescence BoLA-1*02101 \n",
- "1 Barlow - purified MHC/direct/fluorescence BoLA-1*02101 \n",
- "2 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "3 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "4 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "5 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "6 Barlow - purified MHC/competitive/fluorescence BoLA-2*00801 \n",
- "7 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "8 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "9 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "10 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "11 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "12 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "13 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "14 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "15 Barlow - purified MHC/direct/fluorescence BoLA-2*00801 \n",
- "16 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "17 Barlow - purified MHC/direct/fluorescence BoLA-2*01201 \n",
- "18 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "19 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "20 Barlow - purified MHC/direct/fluorescence BoLA-2*01201 \n",
- "21 Barlow - purified MHC/competitive/fluorescence BoLA-2*01201 \n",
- "22 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "23 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "24 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "25 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "26 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "27 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "28 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "29 Morrison - cellular MHC/T cell inhibition BoLA-2*01201 \n",
- "... ... ... \n",
- "239917 Golde - purified MHC SLA-2*0401 \n",
- "239918 Golde - purified MHC SLA-2*0401 \n",
- "239919 Golde - purified MHC SLA-2*0401 \n",
- "239920 Golde - purified MHC SLA-2*0401 \n",
- "239921 Golde - purified MHC SLA-2*0401 \n",
- "239922 Golde - purified MHC SLA-2*0401 \n",
- "239923 Golde - purified MHC SLA-2*0401 \n",
- "239924 Xia - purified MHC/direct/fluorescence SLA-3*0202 \n",
- "239925 Xia - purified MHC SLA-3*0202 \n",
- "239926 Xia - purified MHC SLA-3*0202 \n",
- "239927 Xia - purified MHC SLA-3*0202 \n",
- "239928 Xia - purified MHC SLA-3*0202 \n",
- "239929 Xia - purified MHC/direct/fluorescence SLA-3*0202 \n",
- "239930 Xia - purified MHC SLA-3*0202 \n",
- "239931 Xia - purified MHC SLA-3*0202 \n",
- "239932 Xia - purified MHC/direct/fluorescence SLA-3*0202 \n",
- "239933 Xia - purified MHC/direct/fluorescence SLA-3*0202 \n",
- "239934 Xia - purified MHC SLA-3*0202 \n",
- "239935 Xia - purified MHC SLA-3*0202 \n",
- "239936 Xia - purified MHC SLA-3*0202 \n",
- "239937 Xia - x-ray crystallography SLA-3*0202 \n",
- "239938 Xia - purified MHC SLA-3*0202 \n",
- "239939 Xia - purified MHC/direct/fluorescence SLA-3*0202 \n",
- "239940 Xia - purified MHC SLA-3*0202 \n",
- "239941 Xia - purified MHC SLA-3*0202 \n",
- "239942 Xia - purified MHC SLA-3*0202 \n",
- "239943 Xia - purified MHC SLA-3*0202 \n",
- "239944 Xia - purified MHC SLA-3*0202 \n",
- "239945 Xia - purified MHC SLA-3*0202 \n",
- "239946 Xia - purified MHC SLA-3*0202 \n",
- "\n",
- "[239947 rows x 6 columns]"
- ]
- },
- "execution_count": 13,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "data_df = pandas.read_csv(data_path)\n",
- "data_df"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "# Fit a model: Low level `Class1NeuralNetwork` interface"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 14,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "data": {
- "text/plain": [
- "{'activation': 'relu',\n",
- " 'batch_normalization': False,\n",
- " 'dense_layer_l1_regularization': 0.001,\n",
- " 'dense_layer_l2_regularization': 0.0,\n",
- " 'dropout_probability': 0.0,\n",
- " 'early_stopping': True,\n",
- " 'embedding_init_method': 'glorot_uniform',\n",
- " 'embedding_input_dim': 21,\n",
- " 'embedding_output_dim': 8,\n",
- " 'init': 'glorot_uniform',\n",
- " 'kmer_size': 15,\n",
- " 'layer_sizes': [16],\n",
- " 'left_edge': 4,\n",
- " 'locally_connected_layers': [{'activation': 'tanh',\n",
- " 'filters': 8,\n",
- " 'kernel_size': 3}],\n",
- " 'loss': 'mse',\n",
- " 'max_epochs': 500,\n",
- " 'min_delta': 0,\n",
- " 'minibatch_size': 128,\n",
- " 'mode': 'auto',\n",
- " 'monitor': 'val_loss',\n",
- " 'optimizer': 'rmsprop',\n",
- " 'output_activation': 'sigmoid',\n",
- " 'patience': 10,\n",
- " 'peptide_amino_acid_encoding': 'one-hot',\n",
- " 'pseudosequence_use_embedding': False,\n",
- " 'random_negative_affinity_max': 50000.0,\n",
- " 'random_negative_affinity_min': 20000.0,\n",
- " 'random_negative_constant': 25,\n",
- " 'random_negative_distribution_smoothing': 0.0,\n",
- " 'random_negative_match_distribution': True,\n",
- " 'random_negative_rate': 0.0,\n",
- " 'right_edge': 4,\n",
- " 'take_best_epoch': False,\n",
- " 'use_embedding': False,\n",
- " 'validation_split': 0.2,\n",
- " 'verbose': 1}"
- ]
- },
- "execution_count": 14,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "# We'll use mostly the default hyperparameters here. Could also specify them as kwargs.\n",
- "new_model = mhcflurry.Class1NeuralNetwork(layer_sizes=[16])\n",
- "new_model.hyperparameters"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 16,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "name": "stdout",
- "output_type": "stream",
- "text": [
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 36us/step - loss: 0.0173 - val_loss: 0.0270\n",
- "Epoch 0 / 500: loss=0.0173307. Min val loss (None) at epoch None\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 41us/step - loss: 0.0175 - val_loss: 0.0262\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 39us/step - loss: 0.0170 - val_loss: 0.0278\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 35us/step - loss: 0.0175 - val_loss: 0.0284\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 34us/step - loss: 0.0174 - val_loss: 0.0276\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 59us/step - loss: 0.0171 - val_loss: 0.0272\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 33us/step - loss: 0.0171 - val_loss: 0.0273\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 49us/step - loss: 0.0169 - val_loss: 0.0268\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 66us/step - loss: 0.0171 - val_loss: 0.0273\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 35us/step - loss: 0.0169 - val_loss: 0.0268\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 31us/step - loss: 0.0168 - val_loss: 0.0275\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 41us/step - loss: 0.0169 - val_loss: 0.0275\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 47us/step - loss: 0.0169 - val_loss: 0.0298\n",
- "Early stopping at epoch 12 / 500: loss=0.0168712. Min val loss (0.0261514389179) at epoch 1\n",
- "CPU times: user 1.92 s, sys: 167 ms, total: 2.09 s\n",
- "Wall time: 1.72 s\n"
- ]
- }
- ],
- "source": [
- "train_data = data_df.loc[\n",
- " (data_df.allele == \"HLA-B*57:01\") &\n",
- " (data_df.peptide.str.len() >= 8) &\n",
- " (data_df.peptide.str.len() <= 15)\n",
- "]\n",
- "% time new_model.fit(train_data.peptide.values, train_data.measurement_value.values)"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 17,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "data": {
- "text/plain": [
- "array([ 26802.58186135])"
- ]
- },
- "execution_count": 17,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "new_model.predict([\"SYNPEPII\"])"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "# Fit a model: high level `Class1AffinityPredictor` interface"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 18,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "name": "stdout",
- "output_type": "stream",
- "text": [
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 1s 204us/step - loss: 0.2511 - val_loss: 0.2062\n",
- "Epoch 0 / 10: loss=0.251054. Min val loss (None) at epoch None\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 56us/step - loss: 0.1951 - val_loss: 0.1758\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 52us/step - loss: 0.1658 - val_loss: 0.1494\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 50us/step - loss: 0.1401 - val_loss: 0.1259\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 48us/step - loss: 0.1171 - val_loss: 0.1059\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 52us/step - loss: 0.0965 - val_loss: 0.0869\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 40us/step - loss: 0.0783 - val_loss: 0.0716\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 46us/step - loss: 0.0631 - val_loss: 0.0586\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 54us/step - loss: 0.0514 - val_loss: 0.0486\n",
- "Train on 2491 samples, validate on 623 samples\n",
- "Epoch 1/1\n",
- "2491/2491 [==============================] - 0s 36us/step - loss: 0.0433 - val_loss: 0.0421\n"
- ]
- },
- {
- "data": {
- "text/plain": [
- "[<mhcflurry.class1_neural_network.Class1NeuralNetwork at 0x124ad7d30>]"
- ]
- },
- "execution_count": 18,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "affinity_predictor = mhcflurry.Class1AffinityPredictor()\n",
- "\n",
- "# This can be called any number of times, for example on different alleles, to build up the ensembles.\n",
- "affinity_predictor.fit_allele_specific_predictors(\n",
- " n_models=1,\n",
- " architecture_hyperparameters={\"layer_sizes\": [16], \"max_epochs\": 10},\n",
- " peptides=train_data.peptide.values,\n",
- " affinities=train_data.measurement_value.values,\n",
- " allele=\"HLA-B*57:01\",\n",
- ")"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 19,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "data": {
- "text/plain": [
- "array([ 17405.26823281])"
- ]
- },
- "execution_count": 19,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "affinity_predictor.predict([\"SYNPEPII\"], allele=\"HLA-B*57:01\")"
- ]
- },
- {
- "cell_type": "markdown",
- "metadata": {},
- "source": [
- "# Save and restore the fit model"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 20,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "name": "stdout",
- "output_type": "stream",
- "text": [
- "manifest.csv\r\n",
- "weights_HLA-B*57:01-0-7be58b1094489f2d.npz\r\n"
- ]
- }
- ],
- "source": [
- "!mkdir /tmp/saved-affinity-predictor\n",
- "affinity_predictor.save(\"/tmp/saved-affinity-predictor\")\n",
- "!ls /tmp/saved-affinity-predictor"
- ]
- },
- {
- "cell_type": "code",
- "execution_count": 21,
- "metadata": {
- "collapsed": false
- },
- "outputs": [
- {
- "data": {
- "text/plain": [
- "array([ 17405.26823281])"
- ]
- },
- "execution_count": 21,
- "metadata": {},
- "output_type": "execute_result"
- }
- ],
- "source": [
- "affinity_predictor2 = mhcflurry.Class1AffinityPredictor.load(\"/tmp/saved-affinity-predictor\")\n",
- "affinity_predictor2.predict([\"SYNPEPII\"], allele=\"HLA-B*57:01\")"
- ]
- }
- ],
- "metadata": {
- "anaconda-cloud": {},
- "kernelspec": {
- "display_name": "Python [py3k]",
- "language": "python",
- "name": "Python [py3k]"
- },
- "language_info": {
- "codemirror_mode": {
- "name": "ipython",
- "version": 3
- },
- "file_extension": ".py",
- "mimetype": "text/x-python",
- "name": "python",
- "nbconvert_exporter": "python",
- "pygments_lexer": "ipython3",
- "version": "3.5.2"
- }
- },
- "nbformat": 4,
- "nbformat_minor": 0
-}