DeepOncoAI/master_script.py at master · sysbiolux/DeepOncoAI · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227

###################################################################################################################
# DEEP-ONCO-AI is an analysis pipeline that trains a metaclassifier on the predictions of machine-learning algos  #
# in order to predict the chemosensitivity of cell lines in the CCLE dataset. The first-level algos are trained   #
# independently for each drug and omic type. Most important features of the most contributing algos are retrieved #
# and used to form explainable classifiers. The contributions of the different omics is systematically examined.  #
###################################################################################################################

# Authors: Sebastien De Landtsheer: [email protected]
#          Prof Thomas Sauter, University of Luxembourg

# This version: February 2023

####################
### HOUSEKEEPING ###
####################

import logging
import numpy as np
import pandas as pd
import seaborn as sns
from matplotlib import pyplot as plt
from DBM_toolbox.data_manipulation import data_utils, dataset_class

from config import Config  # many operations are conducted from the Config class, as it has access to the config file

logging.basicConfig(
    filename="run_paper_rev_0.25.log",
    level=logging.INFO,
    filemode="w",
    format="%(asctime)s %(levelname)-8s %(message)s",
    datefmt="%H:%M:%S",
)

rng = np.random.default_rng(42)

outer_folds = 10
inner_folds = 10

config = Config("testall/config_paper_rev_0.25.yaml")  # here is the path to the config file to be used in the analysis

###################################
### READING AND PROCESSING DATA ###
###################################

logging.info("Reading data")
data, ActAreas, ic50s = config.read_data()

logging.info("Filtering data")
filtered_data, filters = config.filter_data(data)

config.save(to_save=filtered_data, name="REV_filtered_0.25")

#####

logging.info("Selecting subsets for feature engineering")
selected_subset = config.select_subsets(filtered_data)

logging.info("Engineering features")
engineered_features = config.engineer_features(filtered_data)

logging.info("Merging engineered features")
engineered_data = filtered_data.merge_with(engineered_features)

logging.info("Quantizing targets")
quantized_data = config.quantize(engineered_data, target_omic="DRUGS", ic50s=ic50s)

final_data = quantized_data.normalize().optimize_formats()
config.save(to_save=final_data, name="REV_preprocessed_data")

missing_data = final_data.dataframe.loc[:, final_data.dataframe.isnull().any(axis=0)]

######

logging.info("Getting optimized models")

trained_models = config.get_models(dataset=final_data, method="standard")
config.save(to_save=trained_models, name="REV_pre-models")

########################## to load previous data

# final_data = data_utils.unpickle_objects('FINAL_explain_preprocessed_data_2024-06-25-21-07-10-041579.pkl')
# trained_models = data_utils.unpickle_objects('FINAL_explain_pre-models_2024-06-25-22-59-37-868197.pkl')


final_results = dict()
feature_importances = dict()
base_models = dict()
fprs = dict()
tprs = dict()
roc_aucs = dict()

data = [x for x in final_data.dataframe.columns if 'ActArea' not in x]
l = len(data)
targets = [x for x in final_data.dataframe.columns if 'ActArea' in x]
omics = final_data.omic[data]

for target in targets:
    base_models[target] = dict()
    omics[target] = 'DRUGS'
    this_df = pd.concat([final_data.dataframe[data], final_data.dataframe[target]], axis=1)
    this_df = dataset_class.Dataset(this_df, omic=omics, database='x')
    this_df = this_df.to_binary(target=target)

    n_samples = len(this_df.dataframe.index)
    split_size = np.floor(n_samples / outer_folds)
    outer_mixed = rng.choice(this_df.dataframe.index, size=n_samples, replace=False)

    outer_train_idx = dict()
    outer_test_idx = dict()
    inner_train_idx = dict()
    inner_test_idx = dict()
    feature_importances[target] = dict()
    final_results[target] = pd.DataFrame(this_df.dataframe[target].copy())  # truth

    for outer_loop in range(outer_folds):
        feature_importances[target][outer_loop] = dict()
        if outer_loop == outer_folds - 1:  # if it is the last split, put all remaining samples in
            outer_test_idx[outer_loop] = outer_mixed[int(outer_loop * split_size): -1]
        else:
            outer_test_idx[outer_loop] = outer_mixed[int(outer_loop * split_size): int((outer_loop + 1) * split_size)]
        outer_train_idx[outer_loop] = [x for x in outer_mixed if x not in outer_test_idx[outer_loop]]

        rest_dataset, valid_dataset = this_df.split(train_index=outer_train_idx[outer_loop],
                                                    test_index=outer_test_idx[outer_loop])

        n_inner_samples = len(rest_dataset.dataframe.index)
        split_inner_size = np.floor(n_inner_samples / inner_folds)
        inner_mixed = rng.choice(rest_dataset.dataframe.index, size=n_inner_samples, replace=False)

        inner_train_idx[outer_loop] = dict()
        inner_test_idx[outer_loop] = dict()
        base_models[target][outer_loop] = dict()

        first_level_preds = pd.DataFrame(this_df.dataframe[target].copy())  # truth

        for inner_loop in range(inner_folds):
            print(f"target: {target}, out: {outer_loop + 1}/{outer_folds}, in: {inner_loop + 1}/{inner_folds}")
            if inner_loop == inner_folds - 1:
                inner_test_idx[outer_loop][inner_loop] = inner_mixed[int(inner_loop * split_inner_size): -1]
            else:
                inner_test_idx[outer_loop][inner_loop] = inner_mixed[int(inner_loop * split_inner_size): int(
                    (inner_loop + 1) * split_inner_size)]
            inner_train_idx[outer_loop][inner_loop] = [x for x in inner_mixed if
                                                       x not in inner_test_idx[outer_loop][inner_loop]]

            train_dataset, test_dataset = rest_dataset.split(train_index=inner_train_idx[outer_loop][inner_loop],
                                                             test_index=inner_test_idx[outer_loop][inner_loop])

            base_models[target][outer_loop][inner_loop] = dict()

            for omic in trained_models[target].keys():  # for each omic type
                if omic != 'complete':
                    print(f'omic: {omic}')
                    train_features = train_dataset.to_pandas(omic=omic)
                    train_labels = train_dataset.to_pandas(omic='DRUGS').values.ravel()
                    test_features = test_dataset.to_pandas(omic=omic)
                    test_labels = test_dataset.to_pandas(omic='DRUGS').values.ravel()

                    base_models[target][outer_loop][inner_loop][omic] = dict()

                    for algo in trained_models[target][omic].keys():  # for each algo
                        print(f'algo: {algo}')
                        this_model = trained_models[target][omic][algo]['estimator']
                        this_model.fit(train_features, train_labels)
                        try:
                            this_predictions = this_model.predict_proba(test_features)
                            this_predictions = this_predictions[:, 1]
                        except:
                            this_predictions = this_model.predict(test_features)
                        first_level_preds.loc[
                            inner_test_idx[outer_loop][inner_loop], 'pred_' + omic + '_' + algo] = this_predictions
                        try:
                            base_models[target][outer_loop][inner_loop][omic][algo] = this_model.feature_importances_
                        except:
                            try:
                                base_models[target][outer_loop][inner_loop][omic][algo] = this_model.coef_
                            except:
                                base_models[target][outer_loop][inner_loop][omic][algo] = 'None'

        for omic in trained_models[target].keys():  # for each omic type
            if omic != 'complete':
                print(f'omic: {omic}')
                train_features = rest_dataset.to_pandas(omic=omic)
                train_labels = rest_dataset.to_pandas(omic='DRUGS').values.ravel()
                test_features = valid_dataset.to_pandas(omic=omic)
                test_labels = valid_dataset.to_pandas(omic='DRUGS').values.ravel()
                for algo in trained_models[target][omic].keys():  # for each algo
                    print(f'algo: {algo}')
                    this_model = trained_models[target][omic][algo]['estimator']
                    this_model.fit(train_features, train_labels)
                    try:
                        this_predictions = this_model.predict_proba(test_features)
                        this_predictions = this_predictions[:, 1]
                    except:
                        this_predictions = this_model.predict(test_features)
                    first_level_preds.loc[outer_test_idx[outer_loop], 'pred_' + omic + '_' + algo] = this_predictions

        cols = first_level_preds.columns
        train_features = first_level_preds.loc[outer_train_idx[outer_loop], cols[1:]]
        train_labels = first_level_preds.loc[outer_train_idx[outer_loop], cols[0]]
        test_features = first_level_preds.loc[outer_test_idx[outer_loop], cols[1:]]
        test_labels = first_level_preds.loc[outer_test_idx[outer_loop], cols[0]]
        train_features, train_labels = data_utils.merge_and_clean(train_features, train_labels)

        for algo in trained_models[target][omic].keys():
            if algo == 'RFC':
                this_model = trained_models[target][omic][algo]['estimator']
                this_model.fit(train_features, train_labels)
                try:
                    this_predictions = this_model.predict_proba(test_features)
                    this_predictions = this_predictions[:, 1]
                except:
                    this_predictions = this_model.predict(test_features)
                try:
                    f_imp = this_model.feature_importances_
                except:
                    f_imp = np.nan
                feature_importances[target][outer_loop][algo] = pd.DataFrame(f_imp, index=this_model.feature_names_in_,
                                                                             columns=[algo])
                final_results[target].loc[outer_test_idx[outer_loop], 'pred2_' + algo] = this_predictions

    omics = omics.drop(target)

config.save(to_save=final_results, name="REV_results")
config.save(to_save=feature_importances, name="REV_featimp")
config.save(to_save=base_models, name="REV_base-models")