tcRIP/ML_EnsembleNet.py at master · innate2adaptive/tcRIP · 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
#==============================================================================
# IMPORTS
#==============================================================================
import tensorflow as tf
import numpy as np
import os
import dataProcessing as dp
import sklearn as sk

from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report
from sklearn.metrics import accuracy_score

import keras
from keras.models import Model
from keras.layers import Input, Dense


#==============================================================================
# ORGANIZE THE DATA
#==============================================================================


mb_size = 32
Z_dim = 16

# load in sequences and vj info for all patients with naive and beta chains
seqs, vj = dp.loadAllPatients()

# filter out joint sequences
seqs[0], seqs[1], vj[0], vj[1], joint = dp.removeDup(seqs[0], seqs[1], vj[0], vj[1])

print("Number of Shared Seqs: {}".format(len(joint)))
print("Shared Percent: %.2f%%" % (len(joint)/(len(seqs[0])+len(seqs[1])) * 100.0))

# add extrac cds
cddict=dp.extraCDs()
seqs[0], vj[0]=dp.addCDextra(seqs[0],vj[0],cddict)
seqs[1], vj[1]=dp.addCDextra(seqs[1],vj[1],cddict)


# replace the seqeunces with their atchely vectors and add the v region as the
# final region
for idx, seqz in enumerate(seqs):
    seqs[idx]=dp.seq2fatch(seqz)
    for idx2, seq in enumerate(seqs[idx]):
        new=np.zeros((1))
        new[0]=float(vj[idx][idx2][0])
        seqs[idx][idx2]=np.concatenate((seq,new))


# filter to a set length
length=14+5+6 # 14 is the most abundant
seqs[0]=np.array(dp.filtr(seqs[0], length*5+1))
seqs[1]=np.array(dp.filtr(seqs[1], length*5+1))

# use function to create data
X, y = dp.dataCreator(seqs[0],seqs[1])

# shuffle data
X, y = sk.utils.shuffle(X,y)


# print class balances
dp.printClassBalance(y)


# because of keras I need to expand the y so that its one-hot encoded
def quickExpand(y):
    newY=[]
    for val in y:
        small=[0,0]
        small[int(val)]+=1
        newY.append(small)
    newY=np.array(newY)
    return newY

y=quickExpand(y)


# 25% Validation set
xTrain, xVal, yTrain, yVal= train_test_split(X, y, test_size=0.25)

print("======================================")
print("Running Classification using Keras NN")


# keras is lacking in that it cant handle splitting

cdr1t, cdr2t, cdr3t, v_genet, _ = np.split(xTrain,(25,55,125,126),1)
cdr1v, cdr2v, cdr3v, v_genev, _ = np.split(xVal,(25,55,125,126),1)


# defining the keras model
cdr1in = Input(shape=(25,))
cdr2in = Input(shape=(30,))
cdr3in = Input(shape=(70,))
v_genein = Input(shape=(1,))

#CDR1 pipeline
cdr1 = Dense(64, activation='elu')(cdr1in)
cdr1_2 = Dense(64, activation='elu')(cdr1)
cdr1_3 = Dense(2, activation='softmax')(cdr1_2)

#CDR2 pipeline
cdr2 = Dense(64, activation='elu')(cdr2in)
cdr2_2 = Dense(64, activation='elu')(cdr2)
cdr2_3 = Dense(2, activation='softmax')(cdr2_2)

#CDR3 pipeline
cdr3 = Dense(128, activation='elu')(cdr3in)
cdr3_2 = Dense(64, activation='elu')(cdr3)
cdr3_3 = Dense(2, activation='softmax')(cdr3_2)

#CDR1 pipeline
v_gene = Dense(64, activation='elu')(v_genein)
v_gene_2 = Dense(64, activation='elu')(v_gene)
v_gene_3 = Dense(2, activation='softmax')(v_gene_2)

# bring it back together
#x = tf.concat((cdr1_3, cdr2_3, cdr3_3, v_gene_3),1)

x_1 = keras.layers.concatenate([cdr1_3, cdr2_3, cdr3_3, v_gene_3],1)

x = Dense(64, activation='relu')(x_1)
predictions = Dense(2, activation='softmax')(x)


# define the model
model = Model(inputs=[cdr1in, cdr2in, cdr3in, v_genein], outputs=[predictions])


model.compile(optimizer='adam',
              loss='categorical_crossentropy',
              metrics=['accuracy'])


model.fit(x=[cdr1t, cdr2t, cdr3t, v_genet],
          y=yTrain,
          batch_size=32,
          epochs=10,
          verbose=2,
          validation_split=0.1)

# evaluate the model
scores = model.evaluate([cdr1v, cdr2v, cdr3v, v_genev], yVal)
print("\nTesting Accuracy: %.2f%%" % ( scores[1]*100))