def generate_scatterplots(config_params, outfolder, mean_control_profile, raw_dataset, lowess_dataset, deviation_dataset, scaled_dev_dataset):
scatter_outfolder = os.path.join(outfolder, 'scatterplots')
if not os.path.isdir(scatter_outfolder):
os.makedirs(scatter_outfolder)
if config_params['scatter_label_scheme'] != '0':
assert config_params['scatter_label_scheme'].isdigit(), '"scatter_label_scheme" must be an integer!'
try:
float(config_params['scatter_label_cutoff'])
except ValueError:
assert False, "'scatter_label_cutoff' must be a float!"
# Load in sample table
sample_table = get_sample_table(config_params)
# Load in barcode table
barcode_table = get_barcode_table(config_params)
# Get formatted condition names
condition_ids = raw_dataset[1]
if config_params['scatter_condition_columns'] == '':
condition_fmt_string = 'screen_name,expt_id'
else:
condition_fmt_string = config_params['scatter_condition_columns']
condition_ids_final = customize_conditions(condition_ids, sample_table, condition_fmt_string)
# If specified, get formatted strain names
# If the strain labeling scheme is 0, then all the labels are blank!
if config_params['scatter_label_scheme'] == '0':
pass
# strain_ids_final = ['' for i in range(raw_dataset[0].shape[0])]
elif config_params['scatter_label_scheme'] in ['1', '2']:
strain_ids = raw_dataset[0]
strain_fmt_string = config_params['scatter_strain_columns']
assert strain_fmt_string != '', 'scatter_strain_columns must be specified if scatter_label_scheme is not "0"'
strain_ids_custom = customize_strains(strain_ids, barcode_table, strain_fmt_string)
else:
assert False, '"scatter_label_scheme" must be one of [0, 1, 2]!'
# Loop over the columns of the datasets and make plots!
for i in range(raw_dataset[2].shape[1]):
# Extract the profiles to work with
raw_prof = np.log(raw_dataset[2][:, i].squeeze())
lowess_prof = lowess_dataset[2][:, i].squeeze()
deviation_prof = deviation_dataset[2][:, i].squeeze()
scaled_dev_prof = scaled_dev_dataset[2][:, i].squeeze()
if get_verbosity(config_params) >= 2:
print raw_prof
print lowess_prof
print deviation_prof
print scaled_dev_prof
# If labeling is to occur, get which strains to label from the final profile
if config_params['scatter_label_scheme'] == '1':
strains_to_label_inds = np.abs(scaled_dev_prof) > float(config_params['scatter_label_cutoff'])
if get_verbosity(config_params) >= 2:
print np.sum(strains_to_label_inds)
strain_ids_final = strain_ids_custom.copy()
strain_ids_final[np.invert(strains_to_label_inds)] = ''
if get_verbosity(config_params) >= 2:
print strain_ids_final[strains_to_label_inds]
elif config_params['scatter_label_scheme'] == '2':
strains_to_label_inds = rankdata(-np.abs(scaled_dev_prof), method = 'min') <= float(config_params['scatter_label_cutoff'])
if get_verbosity(config_params) >= 2:
print np.sum(strains_to_label_inds)
strain_ids_final = strain_ids_custom.copy()
strain_ids_final[np.invert(strains_to_label_inds)] = ''
if get_verbosity(config_params) >= 2:
print strain_ids_final[strains_to_label_inds]
#strains_to_label = scaled_dev_dataset[0][strains_to_label_inds, :]
# Functions to plot both individual versions of the data processing
# plots and a combined version!
make_individual_plots(mean_control_profile, raw_prof, lowess_prof, deviation_prof, scaled_dev_prof,
strain_ids_final, condition_ids_final[i], config_params, scatter_outfolder)
make_2x2_plot(mean_control_profile, raw_prof, lowess_prof, deviation_prof, scaled_dev_prof,
strain_ids_final, condition_ids_final[i], config_params, scatter_outfolder)
def generate_scatterplots(config_params, outfolder, mean_control_profile, raw_dataset, lowess_dataset, deviation_dataset, scaled_dev_dataset):
scatter_outfolder = os.path.join(outfolder, 'scatterplots')
if not os.path.isdir(scatter_outfolder):
os.makedirs(scatter_outfolder)
if config_params['scatter_label_scheme'] != '0':
assert config_params['scatter_label_scheme'].isdigit(), '"scatter_label_scheme" must be an integer!'
try:
float(config_params['scatter_label_cutoff'])
except ValueError:
assert False, "'scatter_label_cutoff' must be a float!"
# Load in sample table
sample_table = get_sample_table(config_params)
# Load in barcode table
barcode_table = get_barcode_table(config_params)
# Get formatted condition names
condition_ids = raw_dataset[1]
if config_params['scatter_condition_columns'] == '':
condition_fmt_string = 'screen_name,expt_id'
else:
condition_fmt_string = config_params['scatter_condition_columns']
condition_ids_final = customize_conditions(condition_ids, sample_table, condition_fmt_string)
# If specified, get formatted strain names
# If the strain labeling scheme is 0, then all the labels are blank!
if config_params['scatter_label_scheme'] == '0':
pass
# strain_ids_final = ['' for i in range(raw_dataset[0].shape[0])]
elif config_params['scatter_label_scheme'] in ['1', '2']:
strain_ids = raw_dataset[0]
strain_fmt_string = config_params['scatter_strain_columns']
assert strain_fmt_string != '', 'scatter_strain_columns must be specified if scatter_label_scheme is not "0"'
strain_ids_custom = customize_strains(strain_ids, barcode_table, strain_fmt_string)
else:
assert False, '"scatter_label_scheme" must be one of [0, 1, 2]!'
# Loop over the columns of the datasets and make plots!
for i in range(raw_dataset[2].shape[1]):
# Extract the profiles to work with
raw_prof = np.log(raw_dataset[2][:, i].squeeze())
lowess_prof = lowess_dataset[2][:, i].squeeze()
deviation_prof = deviation_dataset[2][:, i].squeeze()
scaled_dev_prof = scaled_dev_dataset[2][:, i].squeeze()
if get_verbosity(config_params) >= 2:
print raw_prof
print lowess_prof
print deviation_prof
print scaled_dev_prof
# If labeling is to occur, get which strains to label from the final profile
if config_params['scatter_label_scheme'] == '1':
strains_to_label_inds = np.abs(scaled_dev_prof) > float(config_params['scatter_label_cutoff'])
strain_ids_final = strain_ids_custom.copy()
strain_ids_final[np.invert(strains_to_label_inds)] = ''
elif config_params['scatter_label_scheme'] == '2':
strains_to_label_inds = rankdata(-np.abs(scaled_dev_prof), method = 'min') <= float(config_params['scatter_label_cutoff'])
strain_ids_final = strain_ids_custom.copy()
strain_ids_final[np.invert(strains_to_label_inds)] = ''
#strains_to_label = scaled_dev_dataset[0][strains_to_label_inds, :]
# Set up and draw the 2x2 plot!
f, axarr = plt.subplots(2, 2, sharex = True)
# also create a big plot so x axis label is common
#ax = f.add_subplot(111)
# Turn off axis lines and ticks of the big subplot
#ax.spines['top'].set_color('none')
#ax.spines['bottom'].set_color('none')
#ax.spines['left'].set_color('none')
#ax.spines['right'].set_color('none')
#ax.tick_params(labelcolor='w', top='off', bottom='off', left='off', right='off')
# Draw scatterplots
axarr[0, 0].scatter(mean_control_profile, raw_prof)
axarr[0, 1].scatter(mean_control_profile, lowess_prof)
axarr[1, 0].scatter(mean_control_profile, deviation_prof)
axarr[1, 1].scatter(mean_control_profile, scaled_dev_prof)
# Add point labels
#if config_params['scatter_label_scheme'] == '0':
# for j, lab in enumerate(strain_ids_final):
# axarr[0, 0].annotate(lab, xy = (mean_control_profile[j], raw_prof[j]))
# for j, lab in enumerate(strain_ids_final):
# axarr[0, 1].annotate(lab, xy = (mean_control_profile[j], lowess_prof[j]))
# for j, lab in enumerate(strain_ids_final):
# axarr[1, 0].annotate(lab, xy = (mean_control_profile[j], deviation_prof[j]))
# for j, lab in enumerate(strain_ids_final):
# axarr[1, 1].annotate(lab, xy = (mean_control_profile[j], scaled_dev_prof[j]))
# Add common x label
#ax.set_xlabel('mean control profile (log counts)')
# Add title!
#ax.set_title(condition_ids_final[i])
# Add individual y labels
axarr[0, 0].set_ylabel('Read counts')
axarr[0, 1].set_ylabel('Lowess-normalized read counts')
axarr[1, 0].set_ylabel('Deviation from normalized counts')
axarr[1, 1].set_ylabel('z-score')
outfile = os.path.join(scatter_outfolder, '{}.pdf'.format(condition_ids_final[i]))
plt.savefig(outfile, bbox_inches = 'tight')
plt.close()
def main(dataset, dataset_file, table, val_name, strain_table_f, strain_columns, sample_table_f, condition_columns, verbosity):
strains, conditions, matrix = dataset
full_fname = os.path.abspath(dataset_file)
out_folder = full_fname.replace('.dump.gz', '')
if not os.path.isdir(out_folder):
os.makedirs(out_folder)
# Here is where the script determines what modifications to make to the row/columns labels
if strain_table_f is not None:
strain_table = read_sample_table(strain_table_f)
custom_strains = clus_wrap.customize_strains(strains, strain_table, strain_columns)
# If the table boolean is not True, write out three fiies: matrix, rownames, and colnames
if not table:
strain_fname = os.path.join(out_folder, 'strains.txt')
cond_fname = os.path.join(out_folder, 'conditions.txt')
mat_fname = os.path.join(out_folder, 'matrix.txt.gz')
# Write out the strains!
with open(strain_fname, 'wt') as strain_f:
strain_f.write('Strain_ID\tBarcode\n')
for strain in strains:
strain_f.write('\t'.join(strain) + '\n')
# Write out the conditions!
with open(cond_fname, 'wt') as cond_f:
cond_f.write('screen_name\texpt_id\n')
for cond in conditions:
cond_f.write('\t'.join(cond) + '\n')
# Write out the matrix!
with gzip.open(mat_fname, 'wb') as mat_f:
np.savetxt(mat_f, matrix, delimiter = '\t')
# If the table boolean is True, then reshape into a table and write everything out as one table
else:
# Create containers for the rowname, colname, and value columns of the new table
Strain_IDs = []
Barcodes = []
screen_names = []
expt_ids = []
values = []
# Iterate over the rows and columns of the matrix, and fill in the lists!
for i in range(matrix.shape[0]):
for j in range(matrix.shape[1]):
Strain_IDs.append(strains[i][0])
Barcodes.append(strains[i][1])
screen_names.append(conditions[j][0])
expt_ids.append(conditions[j][1])
values.append(matrix[i, j])
# Create the data frame!
if val_name is None:
val_name = 'value'
table = pd.DataFrame({'Strain_ID': Strain_IDs, 'Barcode': Barcodes, 'screen_name': screen_names, 'expt_id': expt_ids, val_name: values})
# Reorder the data frame columns!
table = table.reindex(columns = ['Strain_ID', 'Barcode', 'screen_name', 'expt_id', val_name])
# Write the table out to file!
tab_fname = os.path.join(out_folder, 'data_table.txt.gz')
with gzip.open(tab_fname, 'wb') as tab_f:
table.to_csv(tab_f, sep = '\t', header = True, index = False)