def export(args):
raw.run(args)
def test_chiron_input():
DATA_FORMAT = np.dtype([('start','<i4'),
('length','<i4'),
('base','S1')])
### Generate dummy dataset and check input ###
dummy_dir = './Dummy_data/'
if not os.path.isdir(dummy_dir):
os.makedirs(dummy_dir)
dummy_fast5 = os.path.join(dummy_dir,'fast5s')
if not os.path.isdir(dummy_fast5):
os.makedirs(dummy_fast5)
file_num = 10
base_signal = {'A':100,'C':200,'G':300,'T':400}
bases = ['A','C','G','T']
for i in range(file_num):
file_n = os.path.join(dummy_fast5,'dummy_' + str(i) + '.fast5')
length = np.random.randint(40000,50000)
start = 0
start_list = []
length_list = []
base_list = []
raw_signal = []
while start < length-1:
start_list.append(start)
step = min(length-start-1, np.random.randint(5,150))
length_list.append(step)
start = start + step
base = bases[np.random.randint(len(bases))]
base_list.append(base)
raw_signal = raw_signal + [base_signal[base]] + [base_signal[base]-1]*(step-1)
event_matrix = np.asarray(list(zip(start_list,length_list,base_list)),dtype = DATA_FORMAT)
with h5py.File(file_n,'w') as root:
if '/Raw' in root:
del root['/Raw']
raw_h = root.create_dataset('/Raw/Reads/Read_'+ str(i)+'/Signal',
shape = (len(raw_signal),),
dtype = np.int16)
channel_h=root.create_dataset('/UniqueGlobalKey/channel_id/',shape=[],dtype=np.int16)
channel_h.attrs['offset']=0
channel_h.attrs['range']=1
channel_h.attrs['digitisation']=1
raw_h[...] = raw_signal[::-1]
if '/Analyses' in root:
del root['/Analyses']
event_h = root.create_dataset('/Analyses/Corrected_000/BaseCalled_template/Events',
shape = (len(event_matrix),),
maxshape=(None,),
dtype = DATA_FORMAT)
event_h[...] = event_matrix
event_h.attrs['read_start_rel_to_raw'] = 0
class Args(object):
def __init__(self):
self.input = dummy_fast5
self.output = dummy_dir
self.basecall_group = 'Corrected_000'
self.mode = 'rna'
self.tffile = 'train.tfrecords'
self.basecall_subgroup = 'BaseCalled_template'
self.unit=True
from chiron.utils import raw
args = Args()
raw.run(args)
train = read_tfrecord(dummy_dir,"train.tfrecords",seq_length=1000,h5py_file_path=os.path.join(dummy_dir,'cache.fast5'))
for i in range(100):
inputX, sequence_length, label = train.next_batch(10,shuffle=False)
accum_len = 0
for idx,x in enumerate(inputX):
x = inputX[idx][:sequence_length[idx]]
y = list()
for x_idx, signal in enumerate(x):
if x_idx==0:
y.append(signal)
else:
if (abs(signal - x[x_idx-1]) >2) or (signal > x[x_idx-1]):
y.append(signal)
corr = np.corrcoef(y, label[1][accum_len:accum_len + len(y)])[0, 1]
for loc in label[0][accum_len:accum_len + len(y)]:
assert(loc[0] == idx)
accum_len += len(y)
assert abs(corr - 1)< 1e-6
print("Input pipeline dummy data test passed!")