def __call__(self, fasta):
"Run the method."
start_time = time.time()
ensure_dir_exists(self.options.output_dir)
predictions = []
# run MEME
self.meme_cmd_args, self.stdoutdata, self.starts, self.Zs, self.thetas, self.lambdas = run_meme(
fasta, self.options)
# parse output
from Bio import Motif
for motif in Motif.parse(open(os.path.join(self.options.output_dir, 'meme.txt')), "MEME"):
for instance in motif.instances:
# MEME parser seems to count from 1, not 0
start = instance.start - 1
prediction = instance.sequence_name, Interval(
start, start + motif.length), instance.strand == '-'
predictions.append(prediction)
logger.info('MEME took %.1f seconds', time.time() - start_time)
return predictions
def run_dataset(method_name, suite_name, data_set, fasta, options):
"Run data set."
logger = logging.getLogger()
logger.setLevel(logging.INFO)
options.output_dir = output_dir_for_dataset(suite_name, data_set)
ensure_dir_exists(options.output_dir)
handler = logging.FileHandler(
os.path.join(options.output_dir, '%s.log' % method_name))
handler.setLevel(logging.INFO)
handler.setFormatter(
logging.Formatter("%(asctime)s - %(levelname)s - %(message)s"))
logger.addHandler(handler)
logging.info('%s is analysing data set: %s', method_name, data_set)
predictions = method_for_name(method_name).Algorithm(options)(fasta)
logging.info('%s predicted %d binding sites.',
method_name, len(predictions))
predictions = list(rationalise_predictions(predictions))
return predictions
W], label='STEME W=%d' % W, ls='-', color=colour)
P.loglog(fasta_sizes, meme_timings[
W], label='MEME W=%d' % W, ls='-.', color=colour)
P.legend(loc='upper left')
P.xlabel('\\# bases in data set')
P.ylabel('seconds')
P.savefig(os.path.join(output_dir, 'timings.eps'))
P.savefig(os.path.join(output_dir, 'timings.png'))
P.close()
#
# do the timings.
#
pylab_utils.set_rcParams_for_latex()
stem_timings = DictOf(list)
meme_timings = DictOf(list)
fasta_sizes = []
for fasta in fastas:
for W in Ws:
options.min_w = options.max_w = W
options.output_dir = os.path.join(
output_dir, 'W=%02d-%s' % (W, stempy.basename_wo_ext(fasta)))
stempy.ensure_dir_exists(options.output_dir)
stem_algorithm = stem.Algorithm(options)
meme_algorithm = meme.Algorithm(options)
stem_timings[W].append(timeit(lambda: stem_algorithm(fasta)))
meme_timings[W].append(timeit(lambda: meme_algorithm(fasta)))
fasta_sizes.append(stem_algorithm.num_bases)
save_timings()
# for each number of sites to use
for num_sites in [
2,
5,
10,
20,
50
]:
# don't try if we don't have enough sequences
if num_sites > num_seqs:
continue
options.min_num_sites = options.max_num_sites = num_sites
# for each width
for min_w, max_w in [
# ( 6, 6),
# ( 8, 8),
# (12, 12),
(6, 16),
# (16, 16),
]:
options.min_w = min_w
options.max_w = max_w
options.output_dir = dir_for_options(options, fasta)
ensure_dir_exists(options.output_dir)
logging.info(
'Comparing MEME to STEM. Output dir=%s', options.output_dir)
meme_algorithm, stem_algorithm = compare_meme_stem(options, fasta)
def create_figures(motifs, occs, by_motif, seq_infos, options):
"""Create figures.
"""
from stempy import ensure_dir_exists
ensure_dir_exists(os.path.join(options.results_dir, 'scan-stats'))
# Size of figlegend
if len(motifs) > 30:
size = 6
elif len(motifs) > 16:
size = 8
elif len(motifs) > 10:
size = 10
else:
size = 12
figlegendprops = {'size': size}
# Format cycler for line plots
format_cycler = create_format_cycler(
linestyle=['--', '-.', '-', ':'],
c=("#E69F00", "#56B4E9", "#009E73", "#F0E442", "#0072B2",
"#D55E00", "#CC79A7"))
# Format cycler for marker plots
# format_cycler_marker = create_format_cycler(
# marker=simple_marker_styles,
# c=("#E69F00", "#56B4E9", "#009E73", "#F0E442", "#0072B2",
# "#D55E00", "#CC79A7"))
# Scan scores
pylab.figure(figsize=(6, 4))
lines = plot_scores_per_motif(motifs, by_motif, format_cycler)
savefig('scan-scores', options)
pylab.close()
# Scan legend
pylab.figure(figsize=(4.25, 4))
pylab.figlegend(lines, motifs, 'center', prop=figlegendprops)
savefig('scan-legend', options)
pylab.close()
# Best Z for each motif/sequence combination
pylab.figure(figsize=(6, 4))
best_Z = calculate_motif_best_Z_per_sequence(
motifs, by_motif, len(seq_infos))
plot_best_Z(motifs, best_Z)
savefig('scan-best-Z', options)
pylab.close()
# Scan motif cooccurrences
pylab.figure(figsize=(6, 4))
# pylab.figlegend(lines, motifs, 'center')
plot_collinearity(motifs, best_Z)
savefig('scan-collinearity', options)
pylab.close()
# Scan positions
pylab.figure(figsize=(6, 4))
lines = plot_site_positions(motifs, occs, by_motif, seq_infos,
format_cycler)
savefig('scan-positions', options)
pylab.close()
# Scan legend with all
pylab.figure(figsize=(4.25, 4))
pylab.figlegend(
lines, ['ALL MOTIFS'] + motifs, 'center', prop=figlegendprops)
savefig('scan-legend-with-all', options)
pylab.close()
# Sequence coverage
pylab.figure(figsize=(6, 4))
plot_seq_coverage(best_Z, format_cycler)
savefig('scan-seq-coverage', options)
pylab.close()
# Scan sequences
pylab.figure(figsize=(6, 4))
lines = plot_seq_distribution(motifs, by_motif, seq_infos, format_cycler)
savefig('scan-sequences', options)
pylab.close()
# Scan legend with markers
# fig = pylab.figure(figsize=(4.25, 4))
# pylab.figlegend(lines, motifs, 'center', prop=figlegendprops)
# savefig('scan-legend-marker', options)
# pylab.close()
# Scan lengths
pylab.figure(figsize=(6, 4))
plot_seq_lengths(seq_infos)
savefig('scan-lengths', options)
pylab.close()
# Scan occurrences by motif
pylab.figure(figsize=(6, len(by_motif) / 4.))
pylab.subplots_adjust(left=.3, bottom=.1, right=.96, top=.98)
plot_occs_by_motif(by_motif)
savefig('scan-occs-by-motif', options)
pylab.close()
