def main():
os.system("module load tensorflow")
project = test_CNN.get_project_and_check_arguments(sys.argv, 'display_graphs.py')
model_ids = []
checkpoints_folder = project.checkpoints_folder_tmp
for filename in os.listdir(checkpoints_folder):
if filename.endswith(".tar"):
model_ids.append(filename[:-len(".tar")])
print("len model_ids: ", len(model_ids))
draw_k_graph(project, model_ids)
exit()
###############################################################################
colors = sns.color_palette("hls", project.MAXIMAL_K)
sorted_ids = [None] * project.MAXIMAL_K
for model_validation_id in model_ids:
split_id = model_validation_id.split("_")
if len(split_id) == 4:
print("split_id: ", split_id)
k = int(model_validation_id.split("_")[2])
else:
k = 4
sorted_ids[k-1] = model_validation_id
print("sorted_ids: ", sorted_ids)
for k in range(1, project.MAXIMAL_K+1):
model_validation_id = sorted_ids[k-1]
draw_ROC_graphs(project, model_validation_id, k, colors[k-1])
print("End!!!")
def main():
project = test_CNN.get_project_and_check_arguments(
sys.argv, 'read_filters_and_run_Homer_compare_motifs.py')
sorted_models_list, map_model_ids = test_CNN.get_sorted_models_list(
project)
for best_model_validation_id in sorted_models_list:
train_species = map_model_ids[best_model_validation_id]
if trained_on_all_species_only:
trained_species_index = len(project.species) - 2 # TODO update
if train_species != project.species[trained_species_index]:
continue
print("train_name = ", train_species)
model_dir = test_CNN.create_directories(project,
best_model_validation_id)
conv_results_dir = os.path.join(model_dir, 'convolution_results')
filters_folder = os.path.join(conv_results_dir, "filters", "layer_1")
num_filters_in_first_layer = project.CNN_structure.get_kernels_shape_and_number(
1)[1]
create_filters_file_for_Homer_compareMotifs(
filters_folder, num_filters_in_first_layer, train_species)
# run Homer:
script = "~tommy/Work/HOMER/bin/compareMotifs.pl"
motifs_file = os.path.join(filters_folder, "filters_file.txt")
output_directory = os.path.join(filters_folder, "Homer_compareMotifs/")
known_motifs = "/cs/cbio/tommy/HOMER/data/knownTFs/vertebrates/all.motifs"
homer_results_file = os.path.join(filters_folder, "homer_results.txt")
os.system(script + " " + motifs_file + " " + output_directory +
" -known " + known_motifs + " -cpu 2 2>&1 | tee " +
homer_results_file)
print("End!")
def main():
project = test_CNN.get_project_and_check_arguments(
sys.argv, 'run_Homer_find_denovo_motifs.py')
for species_name in project.species:
print("species_name = ", species_name, " : ")
if species_name != "simulated":
# species_dir = os.path.join(project.text_samples_base_dir, species_name)
Homer_dir = create_data_for_Homer.create_Homer_directory(
project, species_name)
else:
# species_dir = project.text_samples_base_dir
Homer_dir = create_data_for_Homer.create_Homer_directory(project)
# run Homer findMotifs:
script = "~tommy/Work/HOMER/bin/findMotifs.pl"
positive_samples_path = os.path.join(Homer_dir,
species_name + "_positive.fa")
negative_samples_path = os.path.join(Homer_dir,
species_name + "_negative.fa")
output_directory = os.path.join(Homer_dir, "motifResults/")
homer_results_file = os.path.join(Homer_dir, "homer_results.txt")
os.system(script + " " + positive_samples_path + " fasta " +
output_directory + " -fasta " + negative_samples_path +
" -S 5 -p 16 -bits -mset vertebrates 2>&1 | tee " +
homer_results_file)
print("End!")
def main():
num_times_negative_data_is_taken = 0
project = test_CNN.get_project_and_check_arguments(
sys.argv,
"run_data_loader.py",
num_times_negative_data_is_taken=num_times_negative_data_is_taken)
print "start creating data for project: ", project.project_name
for species in project.species:
species_text_samples_dir = os.path.join(project.text_samples_base_dir,
species)
species_npy_samples_dir = os.path.join(project.samples_base_dir,
species)
if not os.path.exists(species_text_samples_dir) and not os.path.isdir(
species_text_samples_dir):
print("make directory: ", species_text_samples_dir)
os.makedirs(species_text_samples_dir)
if not os.path.exists(species_npy_samples_dir) and not os.path.isdir(
species_npy_samples_dir):
print("make directory: ", species_npy_samples_dir)
os.makedirs(species_npy_samples_dir)
data_loader = DataLoaderH3K27acvsExpandedNeg(project)
data_loader.get_all_positive_and_negative_samples(
num_times_negative_data_is_taken)
data_loader.create_data_for_each_species()
data_loader.create_data_from_all_species_together()
print "End!"
def main():
project = test_CNN.get_project_and_check_arguments(
sys.argv, 'create_data_for_Homer.py')
for species_name in project.species:
print("species_name = ", species_name, " : ")
if species_name != "simulated":
species_dir = os.path.join(project.text_samples_base_dir,
species_name)
Homer_dir = create_Homer_directory(project, species_name)
else:
species_dir = project.text_samples_base_dir
Homer_dir = create_Homer_directory(project)
out_positive_samples_path = os.path.join(Homer_dir,
species_name + "_positive.fa")
out_negative_samples_path = os.path.join(Homer_dir,
species_name + "_negative.fa")
input_positive_samples_path = os.path.join(species_dir,
"positive_samples")
if project.k:
input_negative_samples_path = os.path.join(
species_dir, project.k_let_dirs[project.k - 1],
"negative_samples")
else:
input_negative_samples_path = os.path.join(species_dir,
"negative_samples")
write_data_file_for_Homer(True, input_positive_samples_path,
out_positive_samples_path)
write_data_file_for_Homer(False, input_negative_samples_path,
out_negative_samples_path)
print("End!")
def main():
project = test_CNN.get_project_and_check_arguments(sys.argv,
"run_data_loader.py")
print "start creating data for project: ", project.project_name
data_loader = DataLoader(project, motifs_base_path)
data_loader.create_npy_files()
print "End!"
def main():
project = test_CNN.get_project_and_check_arguments(sys.argv,
"run_data_loader.py")
print "start creating data for project: ", project.project_name
data_loader = DataLoaderNegDatavsShuffle(project)
data_loader.get_all_positive_and_negative_samples()
data_loader.create_data_from_all_species_together()
print "End!"
def main():
project = test_CNN.get_project_and_check_arguments(sys.argv, "run_data_loader.py")
# data_handle.remove_files(project.text_samples_base_dir, ".txt", project.species)
print "start creating data for project: ", project.project_name
data_loader = DataLoaderH3K27acvsNeg(project)
data_loader.get_all_positive_and_negative_samples()
data_loader.create_data_for_each_species()
data_loader.create_data_from_all_species_together()
print "End!"
def main():
project = test_CNN.get_project_and_check_arguments(sys.argv, 'create_species_dirs.py')
project_data_dir = os.path.join(base_path, project.project_name, 'data')
for dir_name in ['npy_files', 'samples']:
dir_path = os.path.join(project_data_dir, dir_name)
for species in project.species:
species_dir = os.path.join(dir_path, species)
if not os.path.exists(species_dir) and not os.path.isdir(species_dir):
print("make directory: ", species_dir)
os.makedirs(species_dir)
def main():
project = test_CNN.get_project_and_check_arguments(sys.argv,
"run_data_loader.py")
print "start creating data for project: ", project.project_name
for species in project.species:
species_text_samples_dir = os.path.join(
project.text_samples_base_dir, species,
project.k_let_dirs[project.k - 1])
species_npy_samples_dir = os.path.join(project.samples_base_dir,
species)
if not os.path.exists(species_text_samples_dir) and not os.path.isdir(
species_text_samples_dir):
print("make directory: ", species_text_samples_dir)
os.makedirs(species_text_samples_dir)
if not os.path.exists(species_npy_samples_dir) and not os.path.isdir(
species_npy_samples_dir):
print("make directory: ", species_npy_samples_dir)
os.makedirs(species_npy_samples_dir)
data_loader = DataLoaderTFvsShuffle(project)
data_loader.get_all_positive_and_negative_samples()
data_loader.create_data_from_all_species_together()
print "End!"
def main():
project = test_CNN.get_project_and_check_arguments(
sys.argv, 'tensor_visualization.py')
sorted_models_list, map_model_ids = test_CNN.get_sorted_models_list(
project)
number_of_species = len(project.species)
# with open(conv1_results_output_file, 'w') as conv_file:
for index_train_species in range(len(project.species)):
if trained_on_all_species_only:
trained_species_index = number_of_species - 2 # all species 238000 # change if needed
train_species = project.species[trained_species_index]
best_model_validation_id = (list(map_model_ids.keys()))[0]
if index_train_species != 0: break
else:
best_model_validation_id = sorted_models_list[index_train_species]
train_species = map_model_ids[best_model_validation_id]
print("start show convolution on species: ", train_species)
model_dir = test_CNN.create_directories(project,
best_model_validation_id)
conv_results_dir = os.path.join(model_dir, 'convolution_results')
get_filters(project, train_species, conv_results_dir,
best_model_validation_id)
def main():
project = test_CNN.get_project_and_check_arguments(
sys.argv, 'display_heatmap_enhancers.py')
test_results_file_path = os.path.join(project.CNN_output_dir,
"test_results.txt")
number_of_test_species = len(project.species) - 2
tested_on_all_species = []
for i in range(number_of_test_species):
tested_on_one_species = [None] * len(project.species)
tested_on_all_species.append(tested_on_one_species)
test_counter = 0
number_of_values = 0
sum_v = 0
with open(test_results_file_path) as results_file:
for line in results_file:
if "\n" in line:
new_line = line[:-1]
if new_line in project.species:
train_index = project.species.index(new_line)
continue
split_line = new_line.split()
for value in split_line:
number_of_values += 1
sum_v += float(value)
old_test_species = old_species_name_order[test_counter]
new_test_index = project.species.index(old_test_species)
tested_on_all_species[new_test_index][train_index] = float(
value)
test_counter += 1
if test_counter == number_of_test_species:
test_counter = 0
figure_path = os.path.join(project.CNN_output_dir,
"heatmap_enhancers_vs_negative_data_cbar.pdf")
create_heatmap(project, tested_on_all_species, figure_path)
def main():
project = test_CNN.get_project_and_check_arguments(sys.argv,
'show_convolution.py')
sorted_models_list, map_model_ids = test_CNN.get_sorted_models_list(
project)
number_of_species = len(project.species)
if trained_on_all_species_only:
trained_species_index = number_of_species - 2 # all species 238000
train_species = project.species[trained_species_index]
best_model_validation_id = (list(map_model_ids.keys()))[0]
test_species = train_species
array_true_labels, array_prediction_scores = test_CNN.import_model_and_test(
project, best_model_validation_id, test_species, train_species)
max_score = max(array_prediction_scores)
print("max_score = ", max_score)
argmax = np.argmax(array_prediction_scores)
print("argmax = ", argmax)
label_max = array_true_labels[argmax]
print("label_max = ", label_max)
min_score = min(array_prediction_scores)
print("min_score = ", min_score)
argmin = np.argmin(array_prediction_scores)
print("argmin = ", argmin)
label_min = array_true_labels[argmin]
print("label_min = ", label_min)
conv_results_all_layers_max, sample_max_score, conv_results_all_layers_min, sample_min_score = \
restore_model_and_get_conv_results(project, train_species,
best_model_validation_id, argmin, argmax)
create_convolution_figures_and_text_files(project,
best_model_validation_id,
train_species,
conv_results_all_layers_max,
sample_max_score, True)
create_convolution_figures_and_text_files(project,
best_model_validation_id,
train_species,
conv_results_all_layers_min,
sample_min_score, False)
else:
for best_model_validation_id in sorted_models_list:
train_species = map_model_ids[best_model_validation_id]
test_species = train_species
array_true_labels, array_prediction_scores = test_CNN.import_model_and_test(
project, best_model_validation_id, test_species, train_species)
max_score = max(array_prediction_scores)
print("max_score = ", max_score)
argmax = np.argmax(array_prediction_scores)
print("argmax = ", argmax)
label_max = array_true_labels[argmax]
print("label_max = ", label_max)
min_score = min(array_prediction_scores)
print("min_score = ", min_score)
argmin = np.argmin(array_prediction_scores)
print("argmin = ", argmin)
label_min = array_true_labels[argmin]
print("label_min = ", label_min)
conv_results_all_layers_max, sample_max_score, conv_results_all_layers_min, sample_min_score =\
restore_model_and_get_conv_results(project, train_species,
best_model_validation_id, argmin, argmax)
create_convolution_figures_and_text_files(
project, best_model_validation_id, train_species,
conv_results_all_layers_max, sample_max_score, True)
create_convolution_figures_and_text_files(
project, best_model_validation_id, train_species,
conv_results_all_layers_min, sample_min_score, False)
print("End!!!")
def main():
fig = plt.figure(1)
x = [0, 1]
plt.plot(x, x, 'k--')
project = test_CNN.get_project_and_check_arguments(sys.argv,
"roc_comparison.py")
TF_name = project.PWM.split("_")[0] # CEBPA for example
figure_roc_path = os.path.join(
project.basic_output_dir, "ROC_comparison_between_models_" +
project.distribution_samples_center_dir + "_sigma_" +
str(project.sigma) + ".pdf")
# CNN:
sorted_models_list, map_model_ids = test_CNN.get_sorted_models_list(
project)
index_train_species = 0
best_model_validation_id = sorted_models_list[index_train_species]
train_species = map_model_ids[best_model_validation_id]
print("train_species: ", train_species)
model_dir = test_CNN.create_directories(project, best_model_validation_id)
CNN_scores, CNN_labels = read_scores_and_labels_files(
model_dir, project, best_model_validation_id=best_model_validation_id)
model_label = "CNN: , AUC: "
add_roc_curve(CNN_labels, CNN_scores, fig, model_label)
# PSSM models:
for pssm_model in PSSM_models: # ["denovo", "JASPAR"]
for pr in ["with_prior", "without_prior"]:
dir_name = "CEBPA_" + pssm_model
PSSM_output_dir = os.path.join(
project.PSSM_output_dir,
project.distribution_samples_center_dir, dir_name)
PSSM_scores, PSSM_labels = read_scores_and_labels_files(
PSSM_output_dir, project, pssm_model=pssm_model, pr=pr)
if pr == "with_prior":
# model_label = "PSSM, " + pssm_model + ", with location prior, AUC: "
if pssm_model == "denovo":
model_label = "denovo motif w/ location prior, AUC: "
elif pssm_model == "JASPAR":
model_label = "Gold standard model - True motif w/ location prior, AUC: "
else:
# model_label = "PSSM, " + pssm_model + ", AUC: "
if pssm_model == "denovo":
model_label = "denovo motif w/o location prior, AUC: "
elif pssm_model == "JASPAR":
model_label = "True motif w/o location prior, AUC: "
add_roc_curve(PSSM_labels, PSSM_scores, fig, model_label)
plt.xlabel('False positive rate', fontsize=20)
plt.ylabel('True positive rate', fontsize=20)
if titles:
plt.suptitle('ROC curve')
data_name = " ".join(project.project_name.split("_"))
plt.title('Comparison between models, ' + data_name +
' of single TF: ' + TF_name)
if legend:
plt.legend(loc='best')
new_figure_path = figure_roc_path[:-len(".pdf")] + "_with_legend.pdf"
else:
new_figure_path = figure_roc_path
plt.savefig(new_figure_path, format='pdf')
print("saving figure: ", new_figure_path, "\n\n")
def main():
project = test_CNN.get_project_and_check_arguments(
sys.argv, 'display_heatmap_TF.py')
sorted_models_list, map_model_ids = test_CNN.get_sorted_models_list(
project)
if "Gallus_gallus" in project.species:
train_indices_map = {
"Canis_familiaris": 2,
"Gallus_gallus": 4,
"Homo_sapiens": 0,
"Monodelphis_domestica": 3,
"Mus_musculus": 1,
"All_species_12000": 5,
"All_species_60000": 6
}
else:
train_indices_map = {
"Canis_familiaris": 2,
"Homo_sapiens": 0,
"Monodelphis_domestica": 3,
"Mus_musculus": 1,
"All_species_12000": 4,
"All_species_60000": 5
}
test_results_file_1 = project.test_file
tested_on_Cfam_results = [None] * len(project.species)
tested_on_Mmus_results = [None] * len(project.species)
if "Gallus_gallus" in project.species:
tested_on_Ggal_results = [None] * len(project.species)
tested_on_Mdom_results = [None] * len(project.species)
tested_on_Hsap_results = [None] * len(project.species)
matrix_results_CEBPA_k_4 = []
auc = None
for test_results_file in [test_results_file_1]:
with open(test_results_file) as results_file:
for line in results_file:
if re.match("^\s$", line):
continue
elif line.startswith("train:") or \
line.startswith("finish test"):
continue
split_line = line.split()
if line.startswith("best_model_validation_id"):
model_id = split_line[2]
train = map_model_ids[model_id]
elif line.startswith("test:"):
test = split_line[1]
elif line.startswith("auc:"):
auc = float(split_line[1])
if auc:
train_index = train_indices_map[train]
if test == "Canis_familiaris":
tested_on_Cfam_results[train_index] = auc
elif test == "Mus_musculus":
tested_on_Mmus_results[train_index] = auc
elif test == "Gallus_gallus":
tested_on_Ggal_results[train_index] = auc
elif test == "Monodelphis_domestica":
tested_on_Mdom_results[train_index] = auc
elif test == "Homo_sapiens":
tested_on_Hsap_results[train_index] = auc
if project.k:
figure_path = os.path.join(
project.CNN_output_dir,
"heatmap_TF_vs_k_shuffle_k_" + str(project.k) + ".pdf")
else:
figure_path = os.path.join(project.CNN_output_dir,
"heatmap_TF_vs_negative_data.pdf")
results_path = os.path.join(project.CNN_output_dir,
"heatmap_results_TF_vs_negative_data.txt")
with open(results_path, "w") as out_results:
if "Gallus_gallus" in project.species:
matrix_results_CEBPA_k_4.append(tested_on_Ggal_results)
string_list = "Chicken: "
for i in tested_on_Ggal_results:
string_list += str(i) + ","
out_results.write(string_list + "\n")
matrix_results_CEBPA_k_4.append(tested_on_Mdom_results)
string_list = "Opossum: "
for i in tested_on_Mdom_results:
string_list += str(i) + ","
out_results.write(string_list + "\n")
matrix_results_CEBPA_k_4.append(tested_on_Cfam_results)
string_list = "Dog: "
for i in tested_on_Cfam_results:
string_list += str(i) + ","
out_results.write(string_list + "\n")
matrix_results_CEBPA_k_4.append(tested_on_Mmus_results)
string_list = "Mouse: "
for i in tested_on_Mmus_results:
string_list += str(i) + ","
out_results.write(string_list + "\n")
matrix_results_CEBPA_k_4.append(tested_on_Hsap_results)
string_list = "Human: "
for i in tested_on_Hsap_results:
string_list += str(i) + ","
out_results.write(string_list + "\n")
create_heatmap(project, matrix_results_CEBPA_k_4, figure_path)
def main():
# os.system("module load tensorflow;")
project = test_CNN.get_project_and_check_arguments(sys.argv,
'run_test_CNN.py')
project.num_times_negative_data_is_taken = n
sorted_models_list, map_model_ids = test_CNN.get_sorted_models_list(
project)
sum_auc = 0
with open(project.test_file, 'a+') as out_file:
number_of_species = len(project.species)
number_species_tested_on = 0
for index_to_test_on in range(number_of_species + 1):
if index_to_test_on != number_of_species:
test_species = project.species[index_to_test_on]
if "All_species" in test_species:
continue
number_species_tested_on += 1
print("test on species: ", test_species)
for best_model_validation_id in sorted_models_list:
# last_iteration - just plot the average auc of all tested species on the ROC figure
if index_to_test_on == number_of_species:
average_auc_returned = test_CNN.draw_roc_curve(
array_true_labels,
array_prediction_scores,
project,
model_dir,
best_model_validation_id,
out_file,
train_species,
test_species,
plot_ROC=plot_ROC,
average_auc=average_auc)
print("average_auc_returned : ", average_auc_returned)
out_file.write("average_auc_returned: {0:.3f}".format(
average_auc_returned) + "\n")
break
train_species = map_model_ids[best_model_validation_id]
if trained_on_one_species_only:
if train_on_human:
if "Homo_sapiens" in project.species:
trained_species_index = project.species.index(
"Homo_sapiens")
elif "Human" in project.species:
trained_species_index = project.species.index(
"Human")
elif train_on_all_samples:
trained_species_index = number_of_species - 2 # all species 238000
elif train_on_dog:
if "Canis_familiaris" in project.species:
trained_species_index = project.species.index(
"Canis_familiaris")
elif "Dog" in project.species:
trained_species_index = project.species.index(
"Dog")
train_species = project.species[trained_species_index]
print("train_species: ", train_species)
# if project.project_name == "TF_vs_k_shuffle":
# model_dir = os.path.join(project.checkpoints_folder, best_model_validation_id)
# else:
model_dir = os.path.join(project.checkpoints_folder_tmp,
best_model_validation_id)
if not os.path.exists(model_dir) and not os.path.isdir(
model_dir):
tar = tarfile.open(model_dir + ".tar")
tar.extractall(path=model_dir)
tar.close()
print("train on species: ", train_species)
model_dir = test_CNN.create_directories(
project, best_model_validation_id)
array_true_labels, array_prediction_scores = \
test_CNN.import_model_and_test(project, best_model_validation_id,
test_species, train_species, out_file)
test_CNN.write_labels_and_scores(model_dir,
best_model_validation_id,
array_prediction_scores,
array_true_labels,
test_species)
auc = test_CNN.draw_roc_curve(array_true_labels,
array_prediction_scores,
project,
model_dir,
best_model_validation_id,
out_file,
train_species,
test_species,
plot_ROC=plot_ROC)
print("auc: ", auc)
# print("sum_auc: ", sum_auc)
sum_auc += auc
average_auc = (sum_auc / number_species_tested_on)
if index_to_test_on != number_of_species:
out_file.write('finish test on species: ' + test_species +
'\n\n\n')
out_file.flush()
print()
# print("average_auc : ", average_auc)
# out_file.write("average AUC: {0:.3f}".format(average_auc) + "\n")
print("end")
