def _check_diff_headers(header_row, sig_label='', pval_label=''):
""" check the additional significance header label and
p_value label. Has be able to deal with there having been
no probesets in column 2.
"""
rr = RunRecord('_check_diff_headers')
try:
sig_col = header_row.index(sig_label)
except ValueError:
rr.dieOnCritical('Significance column header not found in', header_row)
try:
pval_col = header_row.index(pval_label)
except ValueError:
rr.addCritical('Expected to see P-value column header', pval_label)
rr.dieOnCritical('P-value column header not found in', header_row)
return sig_col, pval_col
def _read_data_file(data_path, sep='\t', stable_id_label='ENSEMBL',
probeset_label='probeset', exp_label='exp', sig_label='sig',
pval_label='p_val', is_diff=False):
"""
Get the data out of the file - shared by both:
gene_expr_to_table() and gene_expr_diff_to_table()
"""
rr = RunRecord('_read_data_file')
rows = []
if '.gz' in data_path.lower():
with GzipFile(data_path, 'r') as data_file:
for data in data_file:
rows.append(str(data).strip().split(sep))
else:
with open(data_path) as data_file:
for data in data_file:
rows.append(str(data).strip().split(sep))
# check that headers are valid and whether probesets are present
gene_col, probe_col, exp_col, probes_present = _check_expr_headers(rows[0],
stable_id_label=stable_id_label, probeset_label=probeset_label,
exp_label=exp_label)
if not probes_present:
rr.addInfo('No probeset header found. Reading as', 'RNA-seq')
else:
rr.addInfo('Probesets found. Reading as', 'Micro-array')
if is_diff:
sig_col, pval_col = _check_diff_headers(rows[0], sig_label=sig_label,
pval_label=pval_label)
genes = []; probes = []; exp = []; sigs = []; pvals = []
# get data from each row
for i, row in enumerate(rows):
if i==0:
continue # skip header line
if probes_present:
tmp_probes = list(row[probe_col].split('|'))
else:
tmp_probes = 'P'+str(i) # give RNAseq data a fake probe id
exp_strs = row[exp_col].split('|')
# Nuke any scores and probes marked with 'NA' by R
while 'NA' in exp_strs:
exp_strs.remove('NA')
while 'NA' in tmp_probes:
tmp_probes.remove('NA')
if not len(exp_strs) or not len(tmp_probes):
continue
try:
tmp_expr = map(float, exp_strs)
except ValueError:
rr.addCritical('Expected expression score float on line', i)
rr.dieOnCritical('Line has incorrect format', row)
genes.append(str(row[gene_col]))
probes.append(tmp_probes)
exp.append(tmp_expr)
if is_diff:
sigs.append(str(row[sig_col]))
pvals.append(float(row[pval_col]))
if is_diff:
return genes, probes, exp, sigs, pvals, probes_present
else:
return genes, probes, exp, probes_present
def div_plots(plot_lines, div_study_name, div_by='all'):
""" Divides the counts values in plot_lines by those in the divisor
lines """
rr = RunRecord('div_plots')
# build two matching sized dicts of lines indexed by rank
ranked_plot_lines = {}
dividing_plot_lines = {}
for line in plot_lines:
if line.study == div_study_name:
dividing_plot_lines[line.rank] = line
else:
if not line.rank in ranked_plot_lines.keys():
ranked_plot_lines[line.rank] = []
ranked_plot_lines[line.rank].append(line)
# sanity check
if len(ranked_plot_lines.keys()) == 0:
rr.dieOnCritical('No plot lines.', 'Same study as div plot?')
# default: divide scores line for line - maybe important if trends change
out_lines = []
if div_by.lower() == 'all':
for ranked_index, div_index in zip(sorted(ranked_plot_lines.keys()),
sorted(dividing_plot_lines.keys())):
if ranked_index != div_index:
rr.addCritical('Div and study plot lines do not match',
[len(ranked_plot_lines), len(dividing_plot_lines)])
rr.addCritical('Differences in filtering may be responsible')
rr.dieOnCritical('Try choosing a representative group instead',
['mean-counts', 'median-counts',
'top-expr', 'median-expr', 'bottom-expr'])
for line in ranked_plot_lines[ranked_index]:
line.counts = safe_line_division(line.counts,
dividing_plot_lines[ranked_index].counts)
out_lines.append(line)
else: # divide all lines by the same counts array
# build counts array to choose dividing counts from
counts_lines = []
div_counts = None
for key in sorted(dividing_plot_lines.keys()):
counts_lines.append(dividing_plot_lines[key].counts)
counts_array = numpy.array(counts_lines)
if div_by.lower() == 'mean-counts': # mean counts for all div
div_counts = numpy.mean(counts_array, axis=0)
elif div_by.lower() == 'median-counts': # median counts for all div
div_counts = numpy.median(counts_array, axis=0)
elif div_by.lower() == 'top-expr': # top expressed genes counts line from div
div_counts = counts_array[0]
elif div_by.lower() == 'median-expr': # median expressed genes counts line from div
div_counts = counts_array[int(len(counts_array)/2)]
elif div_by.lower() == 'bottom-expr': # bottom expressed genes counts line from div
div_counts = counts_array[-1]
else:
rr.dieOnCritical('Unrecognised div_by type', div_by)
for ranked_index in ranked_plot_lines:
for line in ranked_plot_lines[ranked_index]:
line.counts = safe_line_division(line.counts, div_counts)
out_lines.append(line)
return out_lines