__init__.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Wed Mar 25 10:42:22 2020

@author: Eric Latorre Crespo
"""

from MSclassifier import load, signature, model
import pickle


class signature_model:
    ''' Define a signature model class that will be filled when trained '''

    def __init__(self):
        pass


class signature_classifier:
    '''signature classifier is class defined to store entire projects,
       train projects and share them.
       '''

    def __init__(self, vcf, positive=None, negative=None,
                 project_name='MSclassifier', reference_genome='GRCh38',
                 exome=False, feature_list=['SBS96', 'ID83', 'DBS78'],
                 model=signature_model()):
        self.vcf = vcf
        self.positive = positive
        self.negative = negative
        self.project_name = project_name
        self.reference_genome = reference_genome
        self.exome = exome
        self.feature_list = feature_list
        self.model = model        # model will be of signature_model class

    def load_vcf(self):
        load.load(self)
        self.sample_names = load.load_names(self)
        self.data = load.dataset(self)

    def signature_train(self, ratio=0.7, end=3):
        self.model.signatures, self.model.features = signature.train(self, ratio, end)

    def signature_fit(self):
        signature.fit(self)

    def model_fit(self, model_type='neural'):
        model.fit(self, model_type)
        self.model.importances = model.importances(self)

    def test_check(self):
        model.test(self)

    def model_predict(self):
        model.predict(self)
        values = self.data['training']
        if not all(v == 0 for v in values):
            self.confusion_matrix, self.accuracy = model.plot_confusion(self)
            self.ROC_curve, self.AUC = model.ROC(self)
        else:
            self.accuracy = None
        self.plot = model.plot_regression(self)

    def export(self):
        pickle.dump(self, open(self.vcf+'/output/'+self.project_name+".p", "wb"))