def genes_from_sites(sites, rho):
"Convert samples sites to data structure ready to pass to class Data."
result = []
for (_v_g, _z_g, x_g), rho_g in zip(sites, rho):
_tmp = DictOf(list)
for x_gi, rho_gi in zip(x_g, rho_g):
_tmp[x_gi].append(rho_gi)
result.append(list(_tmp.items()))
return result
def count_joint_features(data, num_features, num_values, pseudo_count=.35):
"""
Counts the co-occurence of values of pairs of features in the data.
@arg data: Each datum in data is a list of integers. Each element in each list is the value
of the corresponding feature or None for missing data.
@return: A map from tuples of features (f1, f2) to arrays of counts (c1, c2).
"""
def initial_counts(key):
f1, f2 = key
if f1 == f2:
return N.zeros(num_values) + pseudo_count
else:
return N.zeros((num_values, num_values)) + 4. * pseudo_count
result = DictOf(initial_counts, take_key_as_arg=True)
for datum in data:
if num_features != len(datum):
raise RuntimeError('Datum has wrong number (%d) of features (should be %d)' % (len(datum), num_features))
for f1, v1 in enumerate(datum):
if None != v1:
assert v1 < num_values
# update marginal prob of f1 feature
result[f1,f1][v1] += 1.
for f2, v2 in enumerate(datum[:f1]):
if None != v2:
assert v2 < num_values
result[f1,f2][v1,v2] += 1.
return result
def __init__(self, options):
"Construct."
self.options = options
"Options for the test harness."
# put defaults in if not specified
if not len(options.fragments):
self.options.fragments = default_fragments
if not len(options.backgrounds):
self.options.backgrounds = default_backgrounds
self.lazy_sequences = DictOf(load_sequences, take_key_as_arg=True)
"Reads sequences lazily."
self.lazy_sequences_for_hmm = DictOf(self.get_sequence_for_hmm, take_key_as_arg=True)
"Converts sequences to HMM format lazily."
def do_analysis(ucsc_promoters):
"Analyse the promoters."
logging.info('Analysing %d UCSC promoters with %d bases',
len(ucsc_promoters), num_bases_in_promoters(ucsc_promoters))
sequence_analyser = analysis.get_sequence_analyser()
ucsc_analysis = DictOf(list)
for ensembl, promoters in ucsc_promoters.iteritems():
for promoter in promoters:
ucsc_analysis[ensembl].append(sequence_analyser(promoter))
return ucsc_analysis
def group_output_files(directory):
"Find all glam2 output files in a directory and group them by run."
all_files = glob.glob(os.path.join(directory, 'glam2*.out'))
basenames = map(os.path.basename, all_files)
#logging.info('\n'.join(basenames))
result = DictOf(set)
for filename in basenames:
fragment, cross_fold_index, seed = interpret_output_filename(filename)
#logging.info(fragment, cross_fold_index, seed)
result[(fragment, cross_fold_index)].add(seed)
return result
def get_promoters():
"Load the promoters into a dict keyed by ensembl gene."
logging.info('Loading UCSC promoters')
ucsc_promoters = DictOf(list)
for fasta_file in get_fasta_files(options.ucsc_use_masked_seqs):
for seq in corebio.seq_io.fasta_io.iterseq(open(fasta_file),
dna_alphabet):
ucsc, ensembl = seq.description.split()
ucsc_promoters[ensembl].append(seq)
logging.info('Loaded %d promoters', len(ucsc_promoters))
return ucsc_promoters
mouse_promoters = transpro.get_mouse_promoters()
for ref_type in [
'MGI',
'ENTREZGENE',
'REFSEQ',
'UNIGENE',
'ENSMUSG'
]:
print ref_type, len(filter(has_ref_fn(ref_type), mouse_promoters))
print map(mgi_id_for, mouse_promoters[:20])
mgi_to_ensembl = get_mgi_to_ensembl_map()
ensembl_promoters = DictOf(list)
for p in mouse_promoters:
ensembl = ensembl_for(p)
if ensembl:
ensembl_promoters[ensembl].append(p)
sequence_analyser = get_sequence_analyser()
analysis = DictOf(list)
for ensembl, remos in ensembl_promoters.iteritems():
for remo in remos:
analysis[ensembl].append(sequence_analyser(remo.sequence))
fasta = open('mouse-promoters.fa', 'w')
for ensembl, remos in ensembl_promoters.iteritems():
for i, remo in enumerate(remos):
numpy_seqs = map(seq_to_numpy, sequences)
logging.info('Loaded %d sequences', len(sequences))
logging.info('Parsing PSSMs: %s', model_file)
pssms = list(parse_models(open(model_file)))
logging.info('Building models')
models = [
build_hmm_from_semi_parsed(parsed, p_binding_site=p_binding_site)
for parsed in pssms
]
def nucleotide_dist():
return numpy.zeros(4) + .25
base_dists = DictOf(nucleotide_dist)
min_site_length = 20
logging.info('Analysing sequences')
for hmm, traits in models:
sites = []
for sequence in numpy_seqs:
# analyse the sequence for its most likely state sequence
LL, states = hmm.viterbi(sequence)
# for each site
for site_seq, site_states in sites_from_states(
states, sequence, traits.background_states):
# is it long enough?
# Parse arguments
#
methods = args
logging.info('Methods are: %s', ' '.join(methods))
if not len(methods):
raise RuntimeError('No methods specified on command line')
#
# Set up the test harness
#
harness = TestHarness(options)
#
# Merge scores
#
scores = DictOf(
list) # indexed by (method, fragment) or (method, fragment, bg)
def add_results(key, results):
scores[key] += results
scores[key].sort()
for method in methods:
for fragment in harness.options.fragments:
for fold in harness.folds():
dataset = (fragment, fold)
results = harness.results(dataset, method)
add_results((method, fragment), results)
add_results((method, 'Overall'), results)
for bg in harness.options.backgrounds:
def get_matrix_by_name(name):
for m in T.Matrix.all():
if -1 != m.name.lower().find(name.lower()):
return m
return None
pastaa_matrices = {
'muscle' : [ 'SRF_Q5_01', 'SRF_01', 'SRF_Q5_02', 'SRF_C', 'MTATA_B' ],
'heart' : [ 'MEF2_Q6_01', 'SRF_C', 'RSRFC4_01', 'MTATA_B', 'MEF2_02' ],
'liver' : [ 'HNF4_Q6_01', 'HNF1_01', 'HNF4_01', 'HNF1_Q6', 'HNF1_C', ],
'retina' : [ 'CRX_Q4', 'CHX10_01' ],
'leukocyte' : [ 'NFKAPPAB65_01', 'NFKAPPAB_01', 'NFKB_Q6_01', 'CREL_01', 'ETS_Q6' ]
}
pssm_map = get_pssm_to_ensembl_map_min_range()
pastaa_gene_sets = DictOf(set)
for group, names in pastaa_matrices.iteritems():
for name in names:
m = get_matrix_by_name(name)
if m:
ensembl = pssm_map.get(str(m.acc), '<unknown>')
if str(m.acc) in pssm_map:
pastaa_gene_sets[group].add(ensembl)
logging.info(
'%20s : %16s = %s (%16s) - %s (%s)',
group,
name,
m.acc,
m.name,
ensembl,
