def generalize_species(model, s_id2chebi_id, ub_s_ids, chebi, ub_chebi_ids, threshold=UBIQUITOUS_THRESHOLD,
r_ids_to_ignore=None):
"""
Groups metabolites of the model into clusters.
:param model: libsbml.Model model of interest
:param s_id2chebi_id: dict {metabolite_id: ChEBI_term_id}
:param ub_s_ids: collection of ubiquitous metabolite ids
:param chebi: mod_sbml.onto.obo_ontology.Ontology ChEBI ontology
:param ub_chebi_ids: collection of ubiquitous ChEBI term ids
:param threshold: threshold for a metabolite to be considered as frequently participating in reactions
and therefore ubiquitous
:param r_ids_to_ignore: (optional) ids of reactions whose stoichiometry preserving constraint can be ignores
:return:
"""
unmapped_s_ids = {s.getId() for s in model.getListOfSpecies() if s.getId() not in s_id2chebi_id}
term_id2clu = find_term_clustering(model, chebi, s_id2chebi_id, unmapped_s_ids, ub_chebi_ids,
r_ids_to_ignore=r_ids_to_ignore)
if term_id2clu:
term_id2clu = select_representative_terms(term_id2clu, chebi)
s_id2clu = compute_s_id2clu(unmapped_s_ids, model, s_id2chebi_id, term_id2clu)
clu2s_ids = invert_map(s_id2clu)
for s_ids in clu2s_ids.values():
if len(s_ids) == 1:
del s_id2clu[s_ids.pop()]
else:
s_id2clu = {}
if not ub_s_ids:
frequent_ch_ids = get_frequent_term_ids(model, threshold)
ub_s_ids = select_metabolite_ids_by_term_ids(model, frequent_ch_ids) - set(s_id2clu.keys())
# unmapped_s_ids = {s_id for s_id in unmapped_s_ids if s_id not in s_id2clu}
# infer_clusters(model, unmapped_s_ids, s_id2clu, species_id2chebi_id, ub_chebi_ids)
return s_id2clu, ub_s_ids
def maximize(unmapped_s_ids, model, term_id2clu, species_id2term_id, ub_chebi_ids, r_ids_to_ignore=None):
clu2term_ids = invert_map(term_id2clu)
s_id2clu = compute_s_id2clu(unmapped_s_ids, model, species_id2term_id, term_id2clu)
r_id2clu = generalize_reactions(model, s_id2clu, species_id2term_id, ub_chebi_ids, r_ids_to_ignore=r_ids_to_ignore)
thrds = []
for (clu, term_ids) in clu2term_ids.items():
if len(term_ids) <= 1:
continue
thread = MaximizingThread(model, term_ids, species_id2term_id, clu, term_id2clu,
s_id2clu, ub_chebi_ids, r_id2clu, r_ids_to_ignore=r_ids_to_ignore)
thrds.append(thread)
thread.start() # This actually causes the thread to run
for th in thrds:
th.join() # This waits until the thread has completed
return term_id2clu
def cover_t_ids(model, species_id2term_id, ubiquitous_t_ids, t_ids, onto, clu=None, r_ids_to_ignore=None):
"""
Find ancestor terms that cover (generalize) given terms.
:param model: libsbml.Model model of interest
:param species_id2term_id: dict {species_id: term_id}
:param ubiquitous_t_ids: collection of ubiquitous term ids
:param t_ids: collection of term ids to be covered
:param onto: mod_sbml.onto.obo_ontology.Ontology ChEBI ontology
:param clu: current cluster to which the terms belong
:param r_ids_to_ignore: collection of reaction ids to ignore (don't fix their Stoichiometry preserving constraints)
:return: dictionary {term_id: cluster}
"""
term_id2clu = {}
real_terms = {onto.get_term(t_id) for t_id in t_ids if onto.get_term(t_id)}
# If there is no term for t_id in the ontology, we assume it is a metabolite id instead
unmapped_s_ids = {t_id for t_id in t_ids if not onto.get_term(t_id)}
roots = onto.common_points(real_terms, relationships=EQUIVALENT_RELATIONSHIPS)
if roots:
root_id = roots[0].get_id()
new_clu = clu + (root_id, ) if clu else (root_id, )
return {t_id: new_clu for t_id in t_ids}
roots = set()
for term in real_terms:
roots |= onto.get_generalized_ancestors_of_level(term, set(), None, 4)
terms2root = {tuple(sorted(t.get_id() for t in onto.get_sub_tree(root))): root.get_id() for root in roots}
for t_set, root_id in greedy({t.get_id() for t in real_terms}, terms2root, {it: 1 for it in terms2root}):
new_clu = clu + (root_id, ) if clu else (root_id, )
term_id2clu.update({t_id: new_clu for t_id in t_set})
s_id2clu = compute_s_id2clu(set(), model, species_id2term_id, term_id2clu)
infer_clusters(model, unmapped_s_ids, s_id2clu, species_id2term_id, ubiquitous_t_ids,
r_ids_to_ignore=r_ids_to_ignore)
# for s_id in unmapped_s_ids:
# if s_id in s_id2clu:
# term_id2clu[s_id] = (s_id2clu[s_id][1], )
return term_id2clu
