def test_tasks():
set_default(assembly_path=f'{os.getcwd()}/genomes')
cell_line = 'HEK293'
en, lab_en = data_retrieval(cell_line, 'enhancers')
pr, lab_pr = data_retrieval(cell_line, 'promoters')
en = fit_neighbours(en, 5)
pr = fit_neighbours(pr, 5)
epigenomes = {'enhancers': en, 'promoters': pr}
labels = {'enhancers': lab_en, 'promoters': lab_pr}
sequences = {'enhancers': to_bed(en), 'promoters': to_bed(pr)}
_, _, _ = get_tasks(epigenomes, labels, sequences)
def test_checks():
cell_line = 'HEK293'
en, lab_en = data_retrieval(cell_line, 'enhancers')
pr, lab_pr = data_retrieval(cell_line, 'promoters')
epigenomes = {'enhancers': en, 'promoters': pr}
labels = {'enhancers': lab_en, 'promoters': lab_pr}
overfitting_risk(epigenomes)
nan_check(epigenomes)
check_class_balance(labels)
rmtree('datasets')
def test_data_retrieval():
cell_line = 'HepG2'
region = 'promoters'
epigenomes, labels = data_retrieval(cell_line=cell_line, region=region)
epigenomes = fit_neighbours(epigenomes, 5)
scores = drop_too_correlated(epigenomes)
show({'promoters': epigenomes}, {'promoters': labels}, {'promoters': scores})
def test_data_prediction():
set_default(cell_line='HEK293',
region='enhancers',
epochs=2,
splits=2,
batch_size=1024,
boruta_iterations=2,
results_path=f'{os.getcwd()}/results')
input_data, output_data = data_retrieval(get_default('cell_line'),
get_default('region'))
input_data_epi = fit_neighbours(input_data, 5)
input_data_epi = apply_z_scoring(input_data_epi)
input_data_epi = drop_constant_features(get_default('region'),
input_data_epi)
input_data_epi = drop_uncorrelated(input_data_epi, output_data)
input_data_epi = get_filtered_with_boruta(input_data_epi, output_data,
get_default('cell_line'),
get_default('region'))
shape = (input_data_epi.shape[1], )
epi_models = [
get_mlp_epigenomics()(shape, validation_split=0.1, name="MLP"),
get_ffnn_epigenomics_v1()(shape, validation_split=0.1, name="FFNN")
]
results = predict_epigenomics(input_data_epi.values,
output_data.values.ravel(), epi_models)
show_barplots(results, 'epi')
def test_data_retrieval():
cell_line = 'HepG2'
region = 'promoters'
epigenomes, labels = data_retrieval(cell_line=cell_line, region=region)
epigenomes = fit_neighbours(epigenomes, 5)
xs = [*[epigenomes.values]]
ys = [*[labels.values.ravel()]]
titles = ['Epigenomes promoters']
show_decomposed_data(xs, ys, titles)
def test_data_prediction():
set_default(
cell_line='HepG2',
region='promoters',
epochs=2,
splits=2,
batch_size=1024,
results_path=f'{os.getcwd()}/results',
assembly_path=f'{os.getcwd()}/genomes'
)
input_data_seq, output_data = data_retrieval(get_default('cell_line'), get_default('region'))
input_data_seq = to_bed(input_data_seq)
shape = (get_default('window_size'), len(get_default('nucleotides')))
seq_models = [
get_mlp_sequential()(shape, name="MLP")
]
results = predict_sequences(input_data_seq, output_data.values.ravel(), seq_models)
show_barplots(results, 'seq')
def test_data_retrieval():
cell_line = 'HepG2'
region = 'promoters'
data_retrieval(cell_line=cell_line, region=region)
from bioinformatica.data_manipulation import fit_neighbours, apply_z_scoring, drop_constant_features, drop_uncorrelated
from bioinformatica.data_prediction import predict_epigenomics, predict_sequences, show_barplots
from bioinformatica.data_retrieval import data_retrieval, to_bed
from bioinformatica.defaults import set_default, get_default
from bioinformatica.models import get_mlp_epigenomics, get_ffnn_epigenomics_v1, get_ffnn_epigenomics_v2, get_ffnn_epigenomics_v3, \
get_mlp_sequential, get_ffnn_sequential, get_cnn_sequential_v1
set_default(
assembly='hg19', # path
cell_line='HEK293',
region='promoters',
dataset_path=
r'C:\Users\matte\Documents\GitHub\bioinformatica\HepG2\datasets')
if __name__ == '__main__':
input_data_o, output_data = data_retrieval(get_default('cell_line'),
get_default('region'))
input_data_seq = to_bed(
input_data_o
) # annotate genome using index extracted from epigenomic data
# epigenomic data's preproceccing
input_data_epi = fit_neighbours(input_data_o, 5) # NaN imputation
input_data_epi = apply_z_scoring(input_data_epi) # Normalizing
# feature selection
input_data_epi = drop_constant_features(get_default('region'),
input_data_epi)
input_data_epi = drop_uncorrelated(input_data_epi, output_data)
input_data_epi = get_filtered_with_boruta(input_data_epi, output_data,
get_default('cell_line'),
get_default('region'))
def test_data_retrieval():
cell_line = 'HepG2'
region = 'enhancers'
input_data, output_data = data_retrieval(cell_line=cell_line, region=region)
get_sequences(input_data)
