def compute_module_count(inf, is_mmod):
if is_mmod:
mods = eval_util.read_mmods(inf)
else:
mods = eval_util.read_rmods(inf)
return len(mods)
def compute_efficacy(inf, is_rmod, is_partial_module, S, mets, reacts):
if is_rmod:
mods = eval_util.read_mmods(inf)
else:
mods = eval_util.read_rmods(inf)
sizes = []
for mod_name, mod_metabs in mods.iteritems():
sizes.append(len(mod_metabs))
if is_partial_module:
pass
# TODO:PARTIAL: should create one big module of unused ones!? NO! DO NOTHING JUST IGNORE THEM!
N = sum(sizes)
p = efficacy_empirical_p(N)
if len(sizes) != 0:
efficacy = 1.0 * (math.log(efficacy_f(N, alpha, p)) - math.log(
efficacy_F(sizes, len(sizes), alpha, p))) * 1.0 / (
math.log(efficacy_f(N, alpha, p)) -
math.log(2 * efficacy_f(math.sqrt(N), alpha, p)))
else:
efficacy = 0.0
return efficacy
def general_compute_distance(cached_reaction_similarities, similarity_table, table_gene_order, inf, species, criteria, factor1, factor2, is_mmod, is_partial_module, S, mets, reacts, revs, go_loaded_model, moredata_loaded_model):
if is_mmod:
mmods = eval_util.read_mmods(inf)
mods = eval_util.convert_metabmod_to_reactmod(mmods, is_partial_module, S, mets, reacts)
else:
mods = eval_util.read_rmods(inf)
# TODO:PARTIAL: all reactions that are not counted in modules (NO: are assigned to external module
# and) -1 (like blocked reactions) is placed for couplings (in associated rows and columns)
if is_partial_module:
done_reacts = set()
for mod_n, mod_rs in mods.iteritems():
done_reacts.update(mod_rs)
not_done_reacts = set(reacts) - set(done_reacts)
if len(not_done_reacts) > 0:
mods[eval_constants.EXTERNAL_MODULE_ID] = not_done_reacts
if species == 'arabidopsis_irs1597':
mods = filter_out_reactions_by_type(mods, [(0, 'R_EX')])
else:
mods = filter_out_reactions_by_type(mods, [(0, 'R_EX_')])
mods_similarity_table = {}
for mod1_name, mod1 in mods.iteritems():
mods_similarity_table[mod1_name] = {}
for mod2_name, mod2 in mods.iteritems():
sim = 0
for r1 in mod1:
for r2 in mod2:
if mod1_name != mod2_name or r1 != r2: # designed to work for overlapping modules. only filters out a gene-similarity-with-itself for intra-module
sim += compute_similarity_of_gene_pair(cached_reaction_similarities, r1, r2, similarity_table, species, criteria, factor1, factor2, go_loaded_model, moredata_loaded_model)
mods_similarity_table[mod1_name][mod2_name] = sim
# print 'mod simetable done'
# TODO:BUG: for overlapping modules this fails because (len(rxns) - len(mods[mod_name])) is wrong
intra_score = 0
inter_score = 0
for mod_name, mods_sim in mods_similarity_table.iteritems():
intra_piece = mods_sim[mod_name] * 1.0 / pow(len(mods[mod_name]), 2)
inter_piece = sum([mods_sim[other_mod_name] * 1.0 / (len(mods[mod_name]) * len(mods[other_mod_name])) for other_mod_name in mods_similarity_table.iterkeys() if other_mod_name != mod_name])
if mod_name == eval_constants.EXTERNAL_MODULE_ID:
intra_score -= intra_piece
inter_score -= inter_piece
else:
intra_score += intra_piece
inter_score += inter_piece
# print 'done'
return intra_score - inter_score
def compute_cohesion_coupling(inf, is_mmod, is_partial_module, S, mets, reacts,
revs):
couplings, blocks = compute_couplings(
S, mets, reacts, revs,
max(map(lambda s: len(s), reacts)) + 1,
max(map(lambda s: len(s), mets)) + 1)
if is_mmod:
mmods = eval_util.read_mmods(inf)
mods = eval_util.convert_metabmod_to_reactmod(mmods, is_partial_module,
S, mets, reacts)
else:
mods = eval_util.read_rmods(inf)
# TODO:PARTIAL: all reactions that are not counted in modules (NO: are assigned to external module
# and) -1 (like blocked reactions) is placed for couplings (in associated rows and columns)
if is_partial_module:
done_reacts = set()
for mod_n, mod_rs in mods.iteritems():
done_reacts.update(mod_rs)
not_done_reacts = set(reacts) - set(done_reacts)
if len(not_done_reacts) > 0:
mods[eval_constants.EXTERNAL_MODULE_ID] = not_done_reacts
couplings = copy.deepcopy(couplings)
for r in not_done_reacts:
r_idx = reacts.index(r)
for i in range(len(reacts)):
couplings[i][r_idx] = -1
couplings[r_idx][i] = -1
module_coupling_table, module_order_in_header = compute_module_coupling_table(
couplings, blocks, mods, reacts)
return compute_coupling_uncoupling_score(module_coupling_table,
module_order_in_header)
def compute_and_print_size_distribution(inf, is_mmod, species, method):
out_dir = '%s/sizes' % my_constants.evalResultPath
my_util.mkdir_p(out_dir)
if is_mmod:
mods = eval_util.read_mmods(inf)
else:
mods = eval_util.read_rmods(inf)
sizes = []
for mod_name, mod_metabs in mods.iteritems():
sizes.append(len(mod_metabs))
result_file = '%s/%s_%s.png' % (out_dir, species, method)
alt_result_file = '%s/sc_%s_%s.png' % (out_dir, species, method)
f = open('size_distrib.r', 'w')
write_line(f, 'png("%s")' % result_file)
write_line(
f, "hist(c(%s), main='%d modules (%s)', xlab='', ylab='')" %
(','.join([str(s) for s in sizes]), len(mods), species + ' ' + method))
write_line(f, 'dev.off()')
write_line(f, 'png("%s")' % alt_result_file)
write_line(
f,
"stripchart(c(%s), main='%d modules (%s)', method='stack', offset=0.5, pch=1)"
% (','.join([str(s)
for s in sizes]), len(mods), species + ' ' + method))
write_line(f, 'dev.off()')
f.close()
res = os.system('%s size_distrib.r' % my_constants.rScriptPath)
while not os.path.isfile(result_file):
time.sleep(5)
return '=HYPERLINK("%s")' % result_file
def compute_modularity(inf, S, mets, reacts, is_partial_module):
mods = eval_util.read_mmods(
inf
) #khotoote dakhele final module ro mikhone mirize toye in dictionary harkodoom az met ha dakhele yek cell
metab_edges = {}
# the following IF merely fills metabmod_orgranized_reacts
# TODO: do_bipartite_explode + is_rmod not supported yet! do_bipartite_explode is simply ignored in the case
#if is_rmod:
#mods, metabmod_orgranized_reacts = eval_util.convert_reactmod_to_overlapping_metabmod(mods, is_partial_module, S, mets, reacts)
#elif do_bipartite_explode:
# metabmod_orgranized_reacts = {}
# graph_edges = my_util.graph_by_explode_reactions_to_complete_bipartite(S, mets)
# the following computes metabmod_orgranized_reacts
#metab_to_module = {v: k for k in mods.keys() for v in mods[k]}
#for e in graph_edges:
# r_mods = {metab_to_module[e[0]], metab_to_module[e[1]]}
# if len(r_mods) == 1:
# mod = r_mods.pop()
# if mod not in metabmod_orgranized_reacts:
# metabmod_orgranized_reacts[mod] = ([], [], [])
# metabmod_orgranized_reacts[mod][0].append(e)
#else:
#for mod in r_mods:
#if mod not in metabmod_orgranized_reacts:
# metabmod_orgranized_reacts[mod] = ([], [], [])
#metabmod_orgranized_reacts[mod][2].append(e)
#if not is_rmod:
metabmod_orgranized_reacts = {}
metab_to_module = {v: k
for k in mods.keys() for v in mods[k]
} #moshakhas mikone har metabi baraye che moduli hast
if is_partial_module:
done_metabs = set(metab_to_module.keys())
for m in mets:
if m not in done_metabs:
metab_to_module[m] = eval_constants.EXTERNAL_MODULE_ID
for r in reacts:
r_metabs = eval_util.get_metabolites_of_reaction_idx(
S, mets, reacts.index(r)
) #metab hayi ke dakhele in reaction hastand ra barmigardanad
r_mods = set(
[metab_to_module[m] for m in r_metabs]
) #moshakhas mikonad ke har metab dakhele in reaction dar che moduli hast
# TODO: with the following if, reactions where all nodes are in either one module or external space, are counted as outside-reaction
if len(r_mods) == 1: # inner-edge detected
mod = r_mods.pop()
if is_partial_module and mod == eval_constants.EXTERNAL_MODULE_ID: # TODO:PARTIAL: skip reactions completely outside all modules
continue
if mod not in metabmod_orgranized_reacts:
metabmod_orgranized_reacts[mod] = ([], [], [])
metabmod_orgranized_reacts[mod][0].append(r)
else: # outer-edge detected
for mod in r_mods:
if mod not in metabmod_orgranized_reacts:
metabmod_orgranized_reacts[mod] = ([], [], [])
metabmod_orgranized_reacts[mod][2].append(r)
# compute metab_edges
#if do_bipartite_explode:
#for e in graph_edges:
#for ee in e:
#if ee not in metab_edges:
#metab_edges[ee] = []
#metab_edges[ee].append(e)
#if not do_bipartite_explode:
for ri, r in enumerate(reacts):
for re_idx in range(len(S)):
if S[re_idx][ri] == 0:
continue
m = mets[re_idx]
if m not in metab_edges:
metab_edges[m] = []
metab_edges[m].append(
r
) # ta inja miad reaction haye marbot be har metab ra migzare dakhele dictinary ba kelide metab
# compute total_links
#if do_bipartite_explode:
#total_links = len(graph_edges)
#if not do_bipartite_explode:
total_links = 0
for m_row in S: # TODO: what about metabolites that are counted more than one time because of overlapping modules?
m_degree = sum([1 for ri in m_row
if ri != 0]) #tedad azaye mokhalefe 0 ro mishmare
total_links += m_degree
total_links /= 2.0 #inja darajeye kole graph ro bedast miare
modularity = 0.0
for mod_name, mod_metabs in mods.iteritems():
print mod_name, mod_metabs
ls = len(
metabmod_orgranized_reacts[mod_name][0]
) # + len(metabmod_orgranized_reacts[mod_name][1]) # TODO: wrongly-inner edges should be counted?
ds = 0
for m in mod_metabs:
# metab_row = S[mets.index(m)]
# ds += sum([1 for ri in metab_row if ri != 0])
ds += len(
metab_edges[m]
) # TODO: for hyperarcs this causes a reaction to be counted more than once for a module
sum_term = ls * 1.0 / total_links - pow(ds / (2.0 * total_links), 2)
if mod_name == eval_constants.EXTERNAL_MODULE_ID:
modularity -= sum_term
else:
modularity += sum_term
return modularity
def compute_modularity(inf, is_rmod, do_bipartite_explode, is_partial_module, S, mets, reacts):
mods = eval_util.read_mmods(inf)
metab_edges = {}
# the following IF merely fills metabmod_orgranized_reacts
# TODO: do_bipartite_explode + is_rmod not supported yet! do_bipartite_explode is simply ignored in the case
if is_rmod:
mods, metabmod_orgranized_reacts = eval_util.convert_reactmod_to_overlapping_metabmod(mods, is_partial_module, S, mets, reacts)
elif do_bipartite_explode:
metabmod_orgranized_reacts = {}
graph_edges = my_util.graph_by_explode_reactions_to_complete_bipartite(S, mets)
# the following computes metabmod_orgranized_reacts
metab_to_module = {v: k for k in mods.keys() for v in mods[k]}
for e in graph_edges:
r_mods = {metab_to_module[e[0]], metab_to_module[e[1]]}
if len(r_mods) == 1:
mod = r_mods.pop()
if mod not in metabmod_orgranized_reacts:
metabmod_orgranized_reacts[mod] = ([], [], [])
metabmod_orgranized_reacts[mod][0].append(e)
else:
for mod in r_mods:
if mod not in metabmod_orgranized_reacts:
metabmod_orgranized_reacts[mod] = ([], [], [])
metabmod_orgranized_reacts[mod][2].append(e)
else:
metabmod_orgranized_reacts = {}
metab_to_module = {v: k for k in mods.keys() for v in mods[k]}
if is_partial_module:
done_metabs = set(metab_to_module.keys())
for m in mets:
if m not in done_metabs:
metab_to_module[m] = eval_constants.EXTERNAL_MODULE_ID
for r in reacts:
r_metabs = eval_util.get_metabolites_of_reaction_idx(S, mets, reacts.index(r))
r_mods = set([metab_to_module[m] for m in r_metabs])
# TODO: with the following if, reactions where all nodes are in either one module or external space, are counted as outside-reaction
if len(r_mods) == 1: # inner-edge detected
mod = r_mods.pop()
if is_partial_module and mod == eval_constants.EXTERNAL_MODULE_ID: # TODO:PARTIAL: skip reactions completely outside all modules
continue
if mod not in metabmod_orgranized_reacts:
metabmod_orgranized_reacts[mod] = ([], [], [])
metabmod_orgranized_reacts[mod][0].append(r)
else: # outer-edge detected
for mod in r_mods:
if mod not in metabmod_orgranized_reacts:
metabmod_orgranized_reacts[mod] = ([], [], [])
metabmod_orgranized_reacts[mod][2].append(r)
# compute metab_edges
if do_bipartite_explode:
for e in graph_edges:
for ee in e:
if ee not in metab_edges:
metab_edges[ee] = []
metab_edges[ee].append(e)
else:
for ri, r in enumerate(reacts):
for re_idx in range(len(S)):
if S[re_idx][ri] == 0:
continue
m = mets[re_idx]
if m not in metab_edges:
metab_edges[m] = []
metab_edges[m].append(r)
# compute total_links
if do_bipartite_explode:
total_links = len(graph_edges)
else:
total_links = 0
for m_row in S: # TODO: what about metabolites that are counted more than one time because of overlapping modules?
m_degree = sum([1 for ri in m_row if ri != 0])
total_links += m_degree
total_links /= 2.0
modularity = 0.0
for mod_name, mod_metabs in mods.iteritems():
ls = len(metabmod_orgranized_reacts[mod_name][0]) # + len(metabmod_orgranized_reacts[mod_name][1]) # TODO: wrongly-inner edges should be counted?
ds = 0
for m in mod_metabs:
# metab_row = S[mets.index(m)]
# ds += sum([1 for ri in metab_row if ri != 0])
ds += len(metab_edges[m]) # TODO: for hyperarcs this causes a reaction to be counted more than once for a module
sum_term = ls * 1.0 / total_links - pow(ds / (2.0 * total_links), 2)
if mod_name == eval_constants.EXTERNAL_MODULE_ID:
modularity -= sum_term
else:
modularity += sum_term
return modularity
def do_compute_distance(cached_metabolite_similarities,
compound_similarity_table, table_compound_order, inf,
species, factor1, is_rmod, is_partial_module, S, mets,
reacts, revs, chebi_loaded_model):
if is_rmod:
mods = eval_util.read_mmods(inf)
mods, metabmod_orgranized_reacts = eval_util.convert_reactmod_to_overlapping_metabmod(
mods, is_partial_module, S, mets, reacts)
else:
mods = eval_util.read_mmods(inf)
if is_partial_module:
done_mets = set()
for mod_n, mod_mets in mods.iteritems():
done_mets.update(mod_mets)
not_done_metabs = set(mets) - set(done_mets)
if len(not_done_metabs) > 0:
mods[eval_constants.EXTERNAL_MODULE_ID] = not_done_metabs
# mods = filter_out_reactions_by_type(mods, [(0, 'R_EX_')])
mods_similarity_table = {}
for mod1_name, mod1 in mods.iteritems():
mods_similarity_table[mod1_name] = {}
for mod2_name, mod2 in mods.iteritems():
sim = 0
for m1 in mod1:
for m2 in mod2:
if mod1_name != mod2_name or m1 != m2: # designed to work for overlapping modules. only filters out a gene-similarity-with-itself for intra-module
sim += compute_similarity_of_metabolite_pair(
cached_metabolite_similarities, m1, m2,
compound_similarity_table, species, factor1,
chebi_loaded_model)
mods_similarity_table[mod1_name][mod2_name] = sim
if eval_constants.VERBOSE >= 2:
print 'compute_similarity_of_metabolite_pair stat for species %s: done: %d, missed_for_no_compound: %d, missed_for_not_in_chebi: %d' % (
species, metabolite_pair_similarity_done,
metabolite_pair_similarity_missed_reason1,
metabolite_pair_similarity_missed_reason2)
# print 'mod simetable done'
# TODO:BUG: for overlapping modules this fails because (len(rxns) - len(mods[mod_name])) is wrong
intra_score = 0
inter_score = 0
for mod_name, mods_sim in mods_similarity_table.iteritems():
intra_piece = mods_sim[mod_name] * 1.0 / pow(len(mods[mod_name]), 2)
inter_piece = sum([
mods_sim[other_mod_name] * 1.0 /
(len(mods[mod_name]) * len(mods[other_mod_name]))
for other_mod_name in mods_similarity_table.iterkeys()
if other_mod_name != mod_name
])
if mod_name == eval_constants.EXTERNAL_MODULE_ID:
intra_score -= intra_piece
inter_score -= inter_piece
else:
intra_score += intra_piece
inter_score += inter_piece
# print 'done'
return intra_score - inter_score