forked from BioDataLearning/PredSynRNA
-
Notifications
You must be signed in to change notification settings - Fork 0
/
prediction.py
170 lines (152 loc) · 6.52 KB
/
prediction.py
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
import argparse
import pandas as pd
import numpy as np
from sklearn.utils import shuffle
import tensorflow as tf
from keras.models import load_model
from keras import backend as K
import joblib
import random
random.seed(30)
from sklearn.preprocessing import MinMaxScaler
def getparas():
parser = argparse.ArgumentParser(description="progrom usage")
parser.add_argument("-t", "--train", type=str, help="Training data")
parser.add_argument("-f", "--data", type=str, help="Prediction data")
parser.add_argument("-s", "--svm", type=str, help="SVM model")
parser.add_argument("-r", "--rf", type=str, help="RF model")
parser.add_argument("-a", "--ann", type=str, help="ANN model")
parser.add_argument("-p", "--path", type=str, help="File path to save the results")
args = parser.parse_args()
trainfile = args.train
predfile = args.data
svm = args.svm
rf = args.rf
ann = args.ann
path = args.path
return trainfile, predfile, svm, rf, ann, path
''' PFA, prob false alert for binary classifier'''
def binary_PFA(y_true, y_pred, threshold=K.variable(value = 0.5)):
y_pred = K.cast(y_pred >= threshold, 'float32')
# N = total number of negative labels
N = K.sum(1 - y_true)
# FP = total number of false alerts, alerts from the negative class labels
FP = K.sum(y_pred - y_pred * y_true)
return FP/N
''' P_TA prob true alerts for binary classifier'''
def binary_PTA(y_true, y_pred, threshold=K.variable(value = 0.5)):
y_pred = K.cast(y_pred >= threshold, 'float32')
# P = total number of positive labels
P = K.sum(y_true)
# TP = total number of correct alerts, alerts from the positive class labels
TP = K.sum(y_pred * y_true)
return TP/P
'''Calculate ROC AUC during model training, obtained from <https://github.com/nathanshartmann/NILC-at-CWI-2018>'''
def roc_auc(y_true, y_pred):
ptas = tf.stack([binary_PTA(y_true,y_pred,k) for k in np.linspace(0, 1, 1000)],axis=0)
pfas = tf.stack([binary_PFA(y_true,y_pred,k) for k in np.linspace(0, 1, 1000)],axis=0)
pfas = tf.concat([tf.ones((1,)) ,pfas],axis=0)
binSizes = -(pfas[1:]-pfas[:-1])
s = ptas*binSizes
return K.sum(s, axis=0)
def get_train_exp_data(file):
full = pd.read_csv(file)
full = shuffle(full, random_state = 42)
x_exp = full.iloc[:, 2:526].reset_index(drop = True)
# log2(RPKM + 1) transformation
x_exp = round(np.log2(x_exp + 1), 2)
x_exp = np.array(x_exp)
return x_exp
def get_pred_exp_data(file):
full = pd.read_csv(file)
full = shuffle(full, random_state = 42)
x_exp = full.iloc[:, 2:526].reset_index(drop = True)
# log2(RPKM + 1) transformation
x_exp = round(np.log2(x_exp + 1), 2)
x_exp = np.array(x_exp)
return x_exp
def prediction(svm, rf, ann, to_be_predicted, info, path):
svm_predicted = svm.predict(to_be_predicted)
svm_count = 0
for item in svm_predicted:
if item == 1:
svm_count = svm_count + 1
print ("SVM predicted: " + str(svm_count))
rf_predicted = rf.predict(to_be_predicted)
rf_count = 0
for item in rf_predicted:
if item == 1:
rf_count = rf_count + 1
print ("RF predicted: " + str(rf_count))
ann_predicted = ann.predict(to_be_predicted)
ann_rounded = [round(x[0]) for x in ann_predicted]
ann_count = 0
for item in ann_rounded:
if item == 1:
ann_count = ann_count + 1
print ("ANN predicted: " + str(ann_count))
# svm: sort the probabilities in descending order and stored the instances predicted as positive
svm_proba = svm.predict_proba(to_be_predicted)
g1 = []
i1 = []
p1 = []
for j in range(0, len(svm_proba)):
g1.append(info[j][0])
i1.append(info[j][1])
p1.append(svm_proba[j][1])
df1 = pd.DataFrame(list(zip(i1, g1, p1)), columns =['gene_id', 'gene_symbol', 'predict_proba'])
df1.sort_values(by=['predict_proba'],ascending=False, inplace=True)
# rf: sort the probabilities in descending order and stored the instances predicted as positive
rf_proba = rf.predict_proba(to_be_predicted)
g2 = []
i2 = []
p2 = []
for j in range(0, len(rf_proba)):
g2.append(info[j][0])
i2.append(info[j][1])
p2.append(rf_proba[j][1])
df2 = pd.DataFrame(list(zip(i2, g2, p2)), columns =['gene_id', 'gene_symbol', 'predict_proba'])
df2.sort_values(by=['predict_proba'],ascending=False, inplace=True)
# ann: sort the probabilities in descending order and stored the instances predicted as positive
ann_proba = [x[0] for x in ann.predict(to_be_predicted)]
g3 = []
i3 = []
p3 = []
for j in range(0, len(ann_proba)):
g3.append(info[j][0])
i3.append(info[j][1])
p3.append(ann_proba[j])
df3 = pd.DataFrame(list(zip(i3, g3, p3)), columns =['gene_id', 'gene_symbol', 'predict_proba'])
df3.sort_values(by=['predict_proba'],ascending=False, inplace=True)
# write the prioritization results to a single file
print ("Writting prioritization results...")
writer = pd.ExcelWriter(path + "prioritization_result.xlsx", engine='xlsxwriter')
df1.head(svm_count).to_excel(writer, sheet_name='svm_proba_positive', index = False)
df2.head(rf_count).to_excel(writer, sheet_name='rf_proba_positive', index = False)
df3.head(ann_count).to_excel(writer, sheet_name='ann_proba_positive', index = False)
writer.save()
print ("Writting whole prioritization results...")
writer2 = pd.ExcelWriter(path + "full_prioritization_result.xlsx", engine='xlsxwriter')
df1.to_excel(writer2, sheet_name='svm_proba_positive', index = False)
df2.to_excel(writer2, sheet_name='rf_proba_positive', index = False)
df3.to_excel(writer2, sheet_name='ann_proba_positive', index = False)
writer2.save()
def main():
trainingfile, predfile, svmfile, rffile, annfile, path = getparas()
full_data = pd.read_csv(predfile)
full_data = shuffle(full_data, random_state = 42)
info = []
for index, row in full_data.iterrows():
info.append((row['gene_symbol'], row['ensembl_gene_id']))
scaler = MinMaxScaler()
scaler.fit_transform(get_train_exp_data(trainingfile))
to_be_predicted = scaler.transform(get_pred_exp_data(predfile))
dependencies = {'roc_auc': roc_auc}
# load machine learning models
svm = joblib.load(svmfile)
rf = joblib.load(rffile)
ann = load_model(annfile, custom_objects = dependencies)
# prediction, prioritization, and writing results
prediction(svm, rf, ann, to_be_predicted, info, path)
if __name__ == '__main__':
main()