get_learning_stats.py

# -*- coding: utf-8 -*-
import os
import json
import urllib2

DATA_URL = 'https://dl.dropboxusercontent.com/s/blq0xxvx5asfnkl/combined.json?token_hash=AAGZxGLemmo09Ib-_P6kyjnzEoTzTMcIVjJ5b95zdb0Eqw&dl=1'
DATA_FILE = 'combined_sample.json'


class ChemblStatsClient:
    def __init__(self):

        if not os.path.exists(DATA_FILE):
            self.data = json.loads(urllib2.urlopen(DATA_URL).read())
            with open(DATA_FILE, 'w') as fw:
                fw.write(json.dumps(self.data, indent=2))
        else:
            self.data = json.loads(open(DATA_FILE).read())


    def print_basic_stats(self):
        message = '''
         Targets: {0}
         Train compounds subset: {1}
         Test compounds subset: {2}
         Bioactivities: {3}
         '''
        print(message.format(
            len(self.data[u'targets']),
            len(self.data[u'train_compounds']),
            len(self.data[u'test_compounds']),
            len(self.data[u'bioactivities'])
        ))


    def print_prediction_stats(self):
        stats = reduce(
            lambda x, y: (x[0] + y[0], x[1] + y[1], x[2] + y[2]),
            map(lambda compound: (
                len(compound['predicted_targets']),
                len(compound['known_targets']),
                len(set(compound['predicted_targets']).intersection(set(compound['known_targets'])))),
                self.data[u'test_compounds']
            ),
            (0, 0, 0)
        )

        message = '''
        Accurately predicted: {0}  / Known targets: {1} | Ratio: {2}
        '''
        no_test_compounds = float(len(self.data[u'test_compounds']))
        print(message.format(stats[2], stats[1], float(stats[2]) / float(stats[1])))


    def get_known_vs_prediction(self, print_ = True):
        if print_:
            for line in map(lambda x: (x['chemblId'], x['known_targets'], x['predicted_targets']),
                            self.data[u'test_compounds']):
                print('{0}\t\t\t{1}\t\t{2}'.format(*line))

    def get_bioactivity(self, compound_chemblid, target_chemblid):
        return filter(lambda x: x['ingredient_cmpd_chemblid'] == compound_chemblid and x['target_chemblid'] == target_chemblid,
                      self.data[u'bioactivities'])

    def get_compound(self, compound_chemblid):
        return (
            filter(lambda x: x['chemblId'] == compound_chemblid, self.data[u'test_compounds']) +
            filter(lambda x: x['chemblId'] == compound_chemblid, self.data[u'train_compounds']))

    def get_target(self, target_chemblid):
        return filter(lambda x: x['chemblId'] == target_chemblid, self.data[u'targets'])


def main():
    client = ChemblStatsClient()
    client.print_basic_stats()
    client.print_prediction_stats()
    client.get_known_vs_prediction()


if __name__ == '__main__':
    main()

from get_learning_stats import ChemblStatsClient
client = ChemblStatsClient()
client.print_basic_stats()
client.print_prediction_stats()
compound_chemblid = 'CHEMBL234691'
client.get_compound(compound_chemblid)
target_chemblid = u'CHEMBL3979'
client.get_target(target_chemblid)
client.get_bioactivity(compound_chemblid, target_chemblid)