-
Notifications
You must be signed in to change notification settings - Fork 0
/
neurally_test.py
137 lines (114 loc) · 6.5 KB
/
neurally_test.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
import tensorflow as tf
import tensorflow.keras.backend as K
config = tf.compat.v1.ConfigProto()
config.gpu_options.allow_growth = True
session = tf.compat.v1.Session(config=config)
import errno
import numpy as np
import math
import argparse
from pyfaidx import Fasta
import matplotlib.pyplot as plt
import re
from sklearn.metrics import roc_auc_score, auc, precision_recall_curve
from neurally_train import *
def model_testing(test_dataset, args, encfiles, test_model, ref_genome):
roc_auc = []
baseline_pr = []
pr_auc = []
test_steps = len(test_dataset) // args.batch_size
label_num = len(encfiles)
#metric calculation is carried out on a subset of targets because the entire testing samples and labels won't fit in memory..
#test_bin determines the subset target used for testing
test_bin = (label_num * 10) // 100
for cls in range(0, label_num, test_bin):
for sample in range(test_steps):
print(f'Running test samples:{str(sample*args.batch_size) + ":" + str((sample+1)*args.batch_size)}')
test_batch = test_dataset[sample*args.batch_size:(sample+1)*args.batch_size]
x_test, y_test = process_element(test_batch, ref_genome, args)
y_pred = test_model(input_data=x_test)
if (cls + test_bin) > label_num:
y_test = y_test[:,cls:label_num]
y_pred = y_pred[:,cls:label_num]
else:
y_test = y_test[:,cls:cls+test_bin]
y_pred = y_pred[:,cls:cls+test_bin]
if sample == 0:
test_true = y_test.flatten().tolist()
test_pred = y_pred.numpy().flatten().tolist()
else:
test_true.extend(y_test.flatten().tolist())
test_pred.extend(y_pred.numpy().flatten().tolist())
num_class = y_test.shape[-1]
total_peaks = len(test_true) // num_class
test_true = np.array(test_true).reshape(total_peaks,num_class)
test_pred = np.array(test_pred).reshape(total_peaks,num_class)
for num in range(num_class):
true_labels = test_true[:,num]
pred_labels = test_pred[:,num]
roc_auc.append(roc_auc_score(true_labels,pred_labels))
baseline_pr.append(np.mean(true_labels))
pre, rec, _ = precision_recall_curve(true_labels, pred_labels)
sort_pre = np.argsort(pre)
pre = np.array(pre)[sort_pre]
rec = np.array(rec)[sort_pre]
pr_auc.append(auc(pre, rec))
out_path = args.model_dir+"/"+args.model_name+"/"+args.out_file
with open(out_path, "w") as tsv_file:
tsv_file.write("Sample" + "\t" + "Assay" + "\t" + "Bed_file" + "\t" + "AUROC" + "\t" + "PR-AUC" + "\t" + "Baseline_PR-AUC" + "\n")
for cls in range(label_num):
sample_type, assay, bed_file = encfiles[cls].split("_")[0:3]
tsv_file.write(sample_type + "\t" + assay + "\t" + bed_file + "\t" + str(roc_auc[cls]) + "\t" + str(pr_auc[cls]) + "\t" + str(baseline_pr[cls]) + "\n")
tsv_file.write("" + "\t" + "" + "\t" + "Mean_values" + "\t" + str(np.mean(roc_auc)) + "\t" + str(np.mean(pr_auc)) + "\t" + str(np.mean(baseline_pr)) + "\n")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Neur-Ally Testing using sklearn metrics')
parser.add_argument("--epoch_num", type=int, help="epoch number of weights file having minimum validation loss")
parser.add_argument('--dim', type=int, default=128, help='dimensions of input after embedding')
parser.add_argument('--batch_size', type=int, default=160, help='specify the batch_size needed for training')
parser.add_argument('--seq_len', type=int, default=2000, help='total flanking+bin sequence length for each input')
parser.add_argument('--heads', type=int, default=4, help='number of multi head attention heads')
parser.add_argument('-d', '--dataset_files', dest = "dataset_files" , default="datasets/encfiles.txt", help='text file containing names of encode dataset files')
parser.add_argument('--model_dir', dest = "model_dir", default = "Models", help='specify the output folder name for saving the model')
parser.add_argument('-r', '--ref_fasta', dest = "ref_fasta", default = "datasets/hg38.fa", help='specify the reference genome fasta file')
parser.add_argument('--test_file', default="datasets/input_bins_test.txt", help='testing data file')
parser.add_argument('--out_file', default="out_file.txt", help='name of output file having testing results')
parser.add_argument('--model_name', default="neurally", help='specify the name of the model under study')
args, unknown = parser.parse_known_args()
#raises exception if the input testing file does not exist in the current location
if not os.path.isfile(args.test_file):
raise FileNotFoundError(errno.ENOENT, os.strerror(errno.ENOENT), args.test_file)
#raises exception if the reference genome fasta file does not exist in the current location
if not os.path.isfile(args.ref_fasta):
raise FileNotFoundError(errno.ENOENT, os.strerror(errno.ENOENT), args.ref_fasta)
print("Saving reference genome to variable...")
ref_genome = Fasta(args.ref_fasta)
print("Extracting filenames...")
encfiles = extract_names(args)
label_num = len(encfiles)
#create positional encoding array
pos = pos_encode(args.seq_len, args.dim)
print("Initializing the model...")
test_model = Modelsubclass(args, label_num, pos)
print("Reading testing dataset from file...")
with open(args.test_file) as f:
test_dataset = f.readlines()
#create one testing batch for running model once
x_single, _ = one_train_batch(args, test_dataset, ref_genome)
test_model(input_data=x_single)
print("Loading weights from file...")
model_dir = args.model_dir
dir_path = weights_filepath(model_dir+"/"+args.model_name)
file_names = os.listdir(dir_path)
idx_files = []
for names in file_names:
if names.endswith(".index"):
idx_files.append(names)
for idx in idx_files:
if int(re.search('weights.(.+?)-', idx).group(1)) == args.epoch_num:
weights_file = idx[0:-6]
weights_path = dir_path + "/" + weights_file
status = test_model.load_weights(weights_path).expect_partial()
print("Asserting matching of weights...")
status.assert_existing_objects_matched()
print("Testing model...")
model_testing(test_dataset, args, encfiles, test_model, ref_genome)