def fix_stoichiometry(model, term_id2clu, species_id2term_id, ub_chebi_ids, onto, r_ids_to_ignore=None):
clu2term_ids = invert_map(term_id2clu)
thrds = []
conflicts = []
for r in model.getListOfReactions():
if r_ids_to_ignore and r.getId() in r_ids_to_ignore:
continue
t_ids = {species_id2term_id[s_id] if s_id in species_id2term_id else s_id
for s_id in chain((species_ref.getSpecies() for species_ref in r.getListOfReactants()),
(species_ref.getSpecies() for species_ref in r.getListOfProducts()))}
if len(t_ids) > 1:
conflicts.append(t_ids)
for clu, term_ids in clu2term_ids.items():
if len(term_ids) <= 1:
continue
clu_conflicts = [set(it) for it in {tuple(t_ids & term_ids) for t_ids in conflicts} if len(it) > 1]
real_term_ids = {t_id for t_id in term_ids if onto.get_term(t_id)}
unmapped_s_ids = {s_id for s_id in term_ids if not onto.get_term(s_id)}
if clu_conflicts:
thread = StoichiometryFixingThread(model, species_id2term_id, ub_chebi_ids, unmapped_s_ids, real_term_ids,
clu_conflicts, onto, clu, term_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
def select_representative_terms(term_id2clu, onto):
"""
Replaces each cluster with a tuple containing one ChEBI term id: (term_id, )
:param term_id2clu: dict {term_id: cluster}
:param onto: mod_sbml.onto.obo_ontology.Ontology ChEBI ontology
:return: updated (inplace) dict term_id2clu {term_id: (cluster_representative_term_id, )}
"""
clu2term_ids = invert_map(term_id2clu)
used = set()
i = 0
for clu, term_ids in clu2term_ids.items():
terms = {onto.get_term(t) for t in term_ids if onto.get_term(t)}
common_ancestors = \
{t for t in onto.common_points(terms, relationships=EQUIVALENT_RELATIONSHIPS)} if terms else set()
options = common_ancestors - used
if options:
common_ancestor_term = options.pop()
else:
name = common_ancestors.pop().get_name() + " (another)" if common_ancestors else 'fake term'
common_ancestor_term = Term(onto=onto, t_id="chebi:unknown_{0}".format(i), name=name)
onto.add_term(common_ancestor_term)
i += 1
used.add(common_ancestor_term)
for t in term_ids:
term_id2clu[t] = (common_ancestor_term.get_id(), )
return term_id2clu
def __init__(self, st_matrix=None, pws=None,
N=None, V=None, m_id2i=None, r_id2i=None, efm_id2i=None, boundary_m_ids=None,
r_ids=None, m_ids=None, efm_ids=None,
r_id2gr_id=None, gr_id2r_id2c=None, efm_id2gr_id=None, m_id2gr_id=None,):
if not r_id2i:
r_id2i = st_matrix.r_id2i if st_matrix else (pws.r_id2i if pws else {})
self.st_matrix = st_matrix
if N is not None and m_id2i is not None and r_id2i is not None:
self.st_matrix = StoichiometricMatrix(N, m_id2i, r_id2i, boundary_m_ids)
self.pws = pws
if V is not None and r_id2i is not None:
self.pws = PathwaySet(V, r_id2i, efm_id2i)
self.r_ids = set(r_ids) if r_ids else set(self.r_id2i.keys())
self.m_ids = set(m_ids) if m_ids else set(self.m_id2i.keys())
self.efm_ids = set(efm_ids) if efm_ids else set(self.efm_id2i.keys())
self.r_id2gr_id = r_id2gr_id if r_id2gr_id else {}
self.gr_id2r_id2c = gr_id2r_id2c if gr_id2r_id2c else {}
self.efm_id2gr_id = efm_id2gr_id if efm_id2gr_id else {}
self.gr_id2efm_ids = invert_map(self.efm_id2gr_id, list)
self.m_id2gr_id = m_id2gr_id if m_id2gr_id else {}
self.coupled_rs = set()
self.r_types = set()
self.folded_efms = set()
self.pathways = set()
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 write_detailed_r_id2c(model, r_id2c, f):
c2r_ids = invert_map(r_id2c)
for c, r_ids in sorted(c2r_ids.items(), key=lambda it: (-abs(it[0]), -it[0])):
for r_id in sorted(r_ids):
f.write('%g\t%s:\t%s\n'
% (c, r_id, get_sbml_r_formula(model, model.getReaction(r_id), show_compartments=False,
show_metabolite_ids=True)))
f.write('\n')
def in_species_conflict(term, candidate_sps, proposal_s_id2clu):
proposal_clu2s_ids = invert_map(proposal_s_id2clu)
for s in candidate_sps:
s_clu = s.getCompartment(), term
rs = {r_id for r_id in s_id2r_ids[s.getId()]}
clu_s_ids = clu2s_ids[s_clu] | proposal_clu2s_ids[s_clu]
for clu_s_id in clu_s_ids:
if {r_id for r_id in s_id2r_ids[clu_s_id]} & rs:
return True
return False
def in_species_conflict(term, candidate_sps, proposal_s_id2clu):
proposal_clu2s_ids = invert_map(proposal_s_id2clu)
for s in candidate_sps:
s_clu = s.getCompartment(), term
rs = {r_id for r_id in s_id2r_ids[s.getId()]}
clu_s_ids = clu2s_ids[s_clu] | proposal_clu2s_ids[s_clu]
for clu_s_id in clu_s_ids:
if {r_id for r_id in s_id2r_ids[clu_s_id]} & rs:
return True
return False
def serialize_generalization(r_id2clu, s_id2clu, sbml, chebi, path):
doc = libsbml.SBMLReader().readSBML(sbml)
model = doc.getModel()
groups_plugin = model.getPlugin("groups")
clu2r_ids, clu2s_ids = invert_map(r_id2clu), invert_map(s_id2clu)
wb = openpyxl.Workbook()
ws = wb.create_sheet(0, "Metabolite Groups")
row = 1
add_values(ws, row, 1, ["Group Id", "Group Name", "Group CHEBI", "Id", "Name", "Compartment", "CHEBI"], HEADER_STYLE)
row += 1
processed_s_ids = set()
for (g_id, ch_term), s_ids in sorted(clu2s_ids.items(), key=lambda ((g_id, _), s_ids): g_id):
group = groups_plugin.getGroup(g_id)
add_values(ws, row, 1, [g_id, group.getName(), ch_term.get_id() if ch_term else ''])
for s_id in sorted(s_ids, key=lambda s_id: s_id[s_id.find('__'):]):
species = model.getSpecies(s_id)
ch_term = get_chebi_term_by_annotation(species, chebi)
add_values(ws, row, 4, [s_id, species.getName(), model.getCompartment(species.getCompartment()).getName(),
ch_term.get_id() if ch_term else ''])
row += 1
processed_s_ids |= s_ids
def _log_clusters(term_id2clu, onto, model):
clu2term = invert_map(term_id2clu)
blueprint = []
logging.info("-------------------\nquotient species sets:\n-------------------")
for clu in sorted(clu2term.keys(), key=lambda k: -len(clu2term[k])):
term_ids = clu2term[clu]
if len(term_ids) == 1:
continue
blueprint.append(len(term_ids))
logging.info("(%d)\t%s\n" % (len(term_ids), [
onto.get_term(it).get_name() if onto.get_term(it) else model.getSpecies(
it).getName() if model.getSpecies(it) else it for it in term_ids]))
logging.info("Cluster sizes: %s\n-------------------\n\n" % sorted(blueprint, key=lambda s: -s))
def cover_with_onto_terms(model, onto, species_id2chebi_id, term_id2clu, ubiquitous_chebi_ids, r_ids_to_ignore=None):
onto_updated = update_onto(onto, term_id2clu)
if onto_updated:
for clu, t_ids in invert_map(term_id2clu).items():
if len(t_ids) == 1:
del term_id2clu[t_ids.pop()]
else:
new_t_id2clu = cover_t_ids(model, species_id2chebi_id, ubiquitous_chebi_ids, t_ids, onto, clu,
r_ids_to_ignore=r_ids_to_ignore)
for t_id in t_ids:
if t_id in new_t_id2clu:
term_id2clu[t_id] = new_t_id2clu[t_id]
else:
del term_id2clu[t_id]
return onto_updated
def generalize_model(in_sbml, chebi, groups_sbml, out_sbml, ub_s_ids=None, ub_chebi_ids=None, ignore_biomass=True):
"""
Generalizes a model.
:param in_sbml: str, path to the input SBML file
:param chebi: mod_sbml.onto.obo_ontology.Ontology ChEBI ontology
:param groups_sbml: str, path to the output SBML file (with groups extension)
:param out_sbml: str, path to the output SBML file (generalized)
:param ub_s_ids: optional, ids of ubiquitous species (will be inferred if set to None)
:param ub_chebi_ids: optional, ids of ubiquitous ChEBI terms (will be inferred if set to None)
:param ignore_biomass: boolean, whether to ignore the biomass reaction (and its stoichiometry preserving constraint)
:return: tuple (r_id2g_eq, s_id2gr_id, s_id2chebi_id, ub_s_ids):
dict {reaction_id: reaction_group_id}, dict {species_id: species_group_id}, dict {species_id: ChEBI_term_id},
collection of ubiquitous species_ids.
"""
# input_model
input_doc = libsbml.SBMLReader().readSBML(in_sbml)
input_model = input_doc.getModel()
r_ids_to_ignore = get_biomass_r_ids(input_model) if ignore_biomass else None
remove_is_a_reactions(input_model)
annotate_metabolites(input_model, chebi)
# TODO: fix comp separation
# separate_boundary_metabolites(input_model)
remove_unused_elements(input_model)
logging.info("mapping species to ChEBI")
s_id2chebi_id = get_species_id2chebi_id(input_model)
ub_chebi_ids, ub_s_ids = get_ub_elements(input_model, chebi, s_id2chebi_id, ub_chebi_ids, ub_s_ids)
terms = (t for t in (chebi.get_term(t_id) for t_id in s_id2chebi_id.values()) if t)
old_onto_len = len(chebi)
filter_ontology(chebi, terms, relationships=EQUIVALENT_RELATIONSHIPS, min_deepness=3)
logging.info('Filtered the ontology from %d terms to %d' % (old_onto_len, len(chebi)))
threshold = min(max(3, int(0.1 * input_model.getNumReactions())), UBIQUITOUS_THRESHOLD)
s_id2clu, ub_s_ids = generalize_species(input_model, s_id2chebi_id, ub_s_ids, chebi, ub_chebi_ids, threshold,
r_ids_to_ignore=r_ids_to_ignore)
logging.info("generalized species")
r_id2clu = generalize_reactions(input_model, s_id2clu, s_id2chebi_id, ub_chebi_ids,
r_ids_to_ignore=r_ids_to_ignore)
logging.info("generalized reactions")
clu2s_ids = {(c_id, term): s_ids for ((c_id, (term, )), s_ids) in invert_map(s_id2clu).items()}
r_id2g_eq, s_id2gr_id = save_as_comp_generalized_sbml(input_model, out_sbml, groups_sbml, r_id2clu, clu2s_ids,
ub_s_ids, chebi)
return r_id2g_eq, s_id2gr_id, s_id2chebi_id, 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 update_onto(onto, term_id2clu):
ancestors = []
clu2t_ids = invert_map(term_id2clu)
for clu, t_ids in clu2t_ids.items():
if len(t_ids) <= 1:
continue
terms = {onto.get_term(t_id) for t_id in t_ids if onto.get_term(t_id)}
if terms:
ancestors.extend(set(onto.common_points(terms, relationships=EQUIVALENT_RELATIONSHIPS)))
removed_something = False
count = Counter(ancestors)
for t in (t for t in count.keys() if count[t] > 1):
# if this term has been already removed as an ancestor/equivalent of another term
if not onto.get_term(t.get_id()):
continue
for it in onto.get_generalized_ancestors(t, relationships=EQUIVALENT_RELATIONSHIPS):
onto.remove_term(it, True)
for it in onto.get_equivalents(t, relationships=EQUIVALENT_RELATIONSHIPS):
onto.remove_term(it, True)
onto.remove_term(t, True)
removed_something = True
return removed_something
def infer_clusters(model,
unmapped_s_ids,
s_id2clu,
s_id2term_id,
ubiquitous_chebi_ids,
r_ids_to_ignore=None):
# TODO: double check it
return
if not unmapped_s_ids:
return
term_id2s_ids = invert_map(s_id2term_id)
clu2s_ids = invert_map(s_id2clu)
vk2r_ids = get_vk2r_ids(model,
s_id2clu,
s_id2term_id,
ubiquitous_chebi_ids,
r_ids_to_ignore=r_ids_to_ignore)
vk2r_ids = {
vk: r_ids
for (vk, r_ids) in vk2r_ids.items() if len(r_ids) > 1
}
simplified_vk2vk_set = defaultdict(set)
for vk in vk2r_ids.keys():
simplified_vk2vk_set[vertical_key2simplified_vertical_key(vk)].add(vk)
s_id2r_ids = defaultdict(list)
for r in (r for r in model.getListOfReactions()
if r.getNumReactants() + r.getNumProducts() > 2):
if r_ids_to_ignore and r.getId() in r_ids_to_ignore:
continue
r_id = r.getId()
for s_id in chain((species_ref.getSpecies()
for species_ref in r.getListOfReactants()),
(species_ref.getSpecies()
for species_ref in r.getListOfProducts())):
s_id2r_ids[s_id].append(r_id)
def in_species_conflict(term, candidate_sps, proposal_s_id2clu):
proposal_clu2s_ids = invert_map(proposal_s_id2clu)
for s in candidate_sps:
s_clu = s.getCompartment(), term
rs = {r_id for r_id in s_id2r_ids[s.getId()]}
clu_s_ids = clu2s_ids[s_clu] | proposal_clu2s_ids[s_clu]
for clu_s_id in clu_s_ids:
if {r_id for r_id in s_id2r_ids[clu_s_id]} & rs:
return True
return False
processed_r_ids = reduce(lambda s1, s2: s1 | s2, vk2r_ids.values(), set())
for r in model.getListOfReactions():
if r.getId(
) in processed_r_ids or not unmapped_s_ids & get_metabolites(r):
continue
if r_ids_to_ignore and r.getId() in r_ids_to_ignore:
continue
ub_rs, ub_ps, rs, ps = get_vertical_key(model, r, s_id2clu,
s_id2term_id,
ubiquitous_chebi_ids)
vk = ub_rs, ub_ps, rs, ps
rs, ps = set(rs), set(ps)
partial_rs, partial_ps = {(s_id, c_id) for (s_id, c_id) in rs if s_id not in unmapped_s_ids}, \
{(s_id, c_id) for (s_id, c_id) in ps if s_id not in unmapped_s_ids}
if len(ub_rs) + len(ub_ps) + len(partial_rs) + len(partial_ps) < 2:
continue
simplified_vk = vertical_key2simplified_vertical_key(vk)
if simplified_vk in simplified_vk2vk_set:
ub_rs, ub_ps = tuple(sorted(ub_rs)), tuple(sorted(ub_ps))
for (vk_ub_rs, vk_ub_ps, vk_rs,
vk_ps) in simplified_vk2vk_set[simplified_vk]:
vk_rs, vk_ps = {(s_id if s_id not in s_id2clu else s_id2clu[s_id], c_id) for (s_id, c_id) in vk_rs}, \
{(s_id if s_id not in s_id2clu else s_id2clu[s_id], c_id) for (s_id, c_id) in vk_ps}
proposal = {}
if vk_ub_rs == ub_ps and vk_ub_ps == ub_rs and not partial_rs - vk_ps and not partial_ps - vk_rs:
vk_ub_rs, vk_ub_ps, partial_rs, partial_ps = vk_ub_ps, vk_ub_rs, partial_ps, partial_rs
if vk_ub_rs == ub_rs and vk_ub_ps == ub_ps and not partial_rs - vk_rs and not partial_ps - vk_ps:
r_s_ids = rs - vk_rs
p_s_ids = ps - vk_ps
if 0 < len(r_s_ids) <= 1 and 0 < len(
p_s_ids) <= 1 and r_s_ids or p_s_ids:
if r_s_ids and vk_rs - rs:
s_id, c_id = r_s_ids.pop()
# if it is not a species id but a cluster, continue
if not isinstance(s_id, str):
continue
clu, c_id = (vk_rs - rs).pop()
# if it is a species id instead of a cluster, continue
if not isinstance(clu, tuple):
continue
candidate_sps = {
model.getSpecies(it)
for it in (term_id2s_ids[s_id2term_id[s_id]]
if s_id in s_id2term_id else {s_id})
}
comp, term = clu
if not in_species_conflict(term, candidate_sps,
proposal):
for s in candidate_sps:
proposal[
s.getId()] = s.getCompartment(), term
else:
continue
if p_s_ids and vk_ps - ps:
s_id, c_id = p_s_ids.pop()
# if it is not a species id but a cluster, continue
if not isinstance(s_id, str):
continue
clu, c_id = (vk_ps - ps).pop()
# if it is a species id instead of a cluster, continue
if not isinstance(clu, tuple):
continue
candidate_sps = {
model.getSpecies(it)
for it in (term_id2s_ids[s_id2term_id[s_id]]
if s_id in s_id2term_id else {s_id})
}
comp, term = clu
if not in_species_conflict(term, candidate_sps,
proposal):
for s in candidate_sps:
proposal[
s.getId()] = s.getCompartment(), term
else:
continue
if proposal:
s_id2clu.update(proposal)
for s_id, clu in proposal.items():
clu2s_ids[clu].add(s_id)
unmapped_s_ids -= set(proposal.keys())
def suggest_clusters(model, unmapped_s_ids, term_id2clu, s_id2term_id, ubiquitous_chebi_ids, r_ids_to_ignore=None):
# TODO: double check it
return
if not unmapped_s_ids:
return
s_id2clu = compute_s_id2clu(set(), model, s_id2term_id, term_id2clu)
term_id2s_ids = invert_map(s_id2term_id)
vk2r_ids = get_vk2r_ids(model, s_id2clu, s_id2term_id, ubiquitous_chebi_ids, r_ids_to_ignore=r_ids_to_ignore)
vk2r_ids = {vk: r_ids for (vk, r_ids) in vk2r_ids.items() if len(r_ids) > 1}
processed_r_ids = reduce(lambda s1, s2: s1 | s2, vk2r_ids.values(), set())
s_vk2vk = defaultdict(set)
for vk in vk2r_ids.keys():
s_vk2vk[vertical_key2simplified_vertical_key(vk)].add(vk)
s_id2r_ids = defaultdict(list)
for r in (r for r in model.getListOfReactions() if r.getNumReactants() + r.getNumProducts() > 2):
if r_ids_to_ignore and r.getId() in r_ids_to_ignore:
continue
r_id = r.getId()
for s_id in chain((species_ref.getSpecies() for species_ref in r.getListOfReactants()),
(species_ref.getSpecies() for species_ref in r.getListOfProducts())):
s_id2r_ids[s_id].append(r_id)
for r in model.getListOfReactions():
if r.getId() in processed_r_ids or not unmapped_s_ids & get_metabolites(r):
continue
if r_ids_to_ignore and r.getId() in r_ids_to_ignore:
continue
ub_rs, ub_ps, rs, ps = get_vertical_key(model, r, s_id2clu, s_id2term_id, ubiquitous_chebi_ids)
vk = ub_rs, ub_ps, rs, ps
rs, ps = set(rs), set(ps)
partial_rs, partial_ps = {(s_id, c_id) for (s_id, c_id) in rs if s_id not in unmapped_s_ids}, \
{(s_id, c_id) for (s_id, c_id) in ps if s_id not in unmapped_s_ids}
if vk in vk2r_ids or len(ub_rs) + len(ub_ps) + len(partial_rs) + len(partial_ps) < 2:
continue
s_vk = vertical_key2simplified_vertical_key(vk)
if s_vk in s_vk2vk:
ub_rs, ub_ps = tuple(sorted(ub_rs)), tuple(sorted(ub_ps))
for (vk_ub_rs, vk_ub_ps, vk_rs, vk_ps) in s_vk2vk[s_vk]:
vk_rs, vk_ps = {(s_id if s_id not in s_id2clu else s_id2clu[s_id], c_id) for (s_id, c_id) in vk_rs}, \
{(s_id if s_id not in s_id2clu else s_id2clu[s_id], c_id) for (s_id, c_id) in vk_ps}
proposal = {}
if vk_ub_rs == ub_ps and vk_ub_ps == ub_rs and not partial_rs - vk_ps and not partial_ps - vk_rs:
vk_ub_rs, vk_ub_ps, partial_rs, partial_ps = vk_ub_ps, vk_ub_rs, partial_ps, partial_rs
if vk_ub_rs == ub_rs and vk_ub_ps == ub_ps and not partial_rs - vk_rs and not partial_ps - vk_ps:
r_s_ids = rs - vk_rs
p_s_ids = ps - vk_ps
if 0 < len(r_s_ids) <= 1 and 0 < len(p_s_ids) <= 1 and r_s_ids or p_s_ids:
if r_s_ids and vk_rs - rs:
s_id, c_id = r_s_ids.pop()
clu, c_id = (vk_rs - rs).pop()
# if it is a species id instead of a cluster, continue
if not isinstance(clu, tuple):
continue
# if it is not a species id but a cluster, continue
if not isinstance(s_id, str):
continue
candidate_sps = {model.getSpecies(sp_id) for sp_id in
(term_id2s_ids[s_id2term_id[s_id]] if s_id in s_id2term_id else [s_id])}
comp, term = clu
for s in candidate_sps:
proposal[s.getId()] = term
if p_s_ids and vk_ps - ps:
s_id, c_id = p_s_ids.pop()
clu, c_id = (vk_ps - ps).pop()
# if it is a species id instead of a cluster, continue
if not isinstance(clu, tuple):
continue
# if it is not a species id but a cluster, continue
if not isinstance(s_id, str):
continue
candidate_sps = {model.getSpecies(it) for it in
(term_id2s_ids[s_id2term_id[s_id]] if s_id in s_id2term_id else {s_id})}
comp, term = clu
for s in candidate_sps:
proposal[s.getId()] = term
if proposal:
for s_id, clu in proposal.items():
term_id2clu[s_id] = (clu, ) if not (isinstance(clu, tuple)) else clu
unmapped_s_ids -= {s_id}
def save_as_comp_generalized_sbml(input_model, out_sbml, groups_sbml, r_id2clu, clu2s_ids, ub_sps, onto):
logging.info("serializing generalization")
s_id_increment, r_id_increment = 0, 0
if groups_sbml:
doc = convert_to_lev3_v1(input_model)
groups_model = doc.getModel()
groups_plugin = groups_model.getPlugin("groups")
if groups_plugin:
logging.info(" saving ubiquitous species annotations")
s_group = groups_plugin.createGroup()
s_group.setId("g_ubiquitous_sps")
s_group.setKind(libsbml.GROUP_KIND_COLLECTION)
s_group.setSBOTerm(SBO_CHEMICAL_MACROMOLECULE)
s_group.setName("ubiquitous species")
for s_id in ub_sps:
member = s_group.createMember()
member.setIdRef(s_id)
add_annotation(s_group, libsbml.BQB_IS_DESCRIBED_BY, GROUP_TYPE_UBIQUITOUS)
if out_sbml:
# generalized model
generalized_doc = libsbml.SBMLDocument(input_model.getSBMLNamespaces())
generalized_model = generalized_doc.createModel()
copy_elements(input_model, generalized_model)
r_id2g_eq, s_id2gr_id = {}, {}
if not clu2s_ids:
logging.info(" nothing to serialize")
else:
clu2r_ids = invert_map(r_id2clu)
logging.info(" creating species groups")
for ((c_id, t), s_ids) in clu2s_ids.items():
comp = input_model.getCompartment(c_id)
if len(s_ids) > 1:
t = onto.get_term(t)
t_name, t_id = (t.get_name(), t.get_id()) if t \
else (' or '.join(input_model.getSpecies(s_id).getName() for s_id in s_ids), None)
if not t_id:
t = t_name
if out_sbml:
new_species = create_species(model=generalized_model, compartment_id=comp.getId(), type_id=None,
name="{0} ({1}) [{2}]".format(t_name, len(s_ids), comp.getName()))
add_annotation(new_species, libsbml.BQB_IS, t_id, CHEBI_PREFIX)
new_s_id = new_species.getId()
else:
s_id_increment += 1
new_s_id = generate_unique_id(input_model, "s_g_", s_id_increment)
for s_id in s_ids:
s_id2gr_id[s_id] = new_s_id, t
if groups_sbml and groups_plugin:
# save as a group
s_group = groups_plugin.createGroup()
s_group.setId(new_s_id)
s_group.setKind(libsbml.GROUP_KIND_CLASSIFICATION)
s_group.setSBOTerm(SBO_CHEMICAL_MACROMOLECULE)
g_name = "{0} [{1}]".format(t_name, comp.getName())
s_group.setName(g_name)
# logging.info("%s: %d" % (g_name, len(s_ids)))
if t_id:
add_annotation(s_group, libsbml.BQB_IS, t_id, CHEBI_PREFIX)
for s_id in s_ids:
member = s_group.createMember()
member.setIdRef(s_id)
add_annotation(s_group, libsbml.BQB_IS_DESCRIBED_BY, GROUP_TYPE_EQUIV)
generalize_species = lambda species_id: s_id2gr_id[species_id][0] if (species_id in s_id2gr_id) else species_id
s_id_to_generalize = set(s_id2gr_id.keys())
logging.info(" creating reaction groups")
for clu, r_ids in clu2r_ids.items():
representative = input_model.getReaction(list(r_ids)[0])
r_name = "generalized %s" % representative.getName()
if out_sbml:
reactants = dict(get_reactants(representative, stoichiometry=True))
products = dict(get_products(representative, stoichiometry=True))
if (len(r_ids) == 1) and \
not ((set(reactants.keys()) | set(products.keys())) & s_id_to_generalize):
generalized_model.addReaction(representative)
continue
r_id2st = {generalize_species(it): st for (it, st) in reactants.items()}
p_id2st = {generalize_species(it): st for (it, st) in products.items()}
reversible = next((False for r_id in r_ids if not input_model.getReaction(r_id).getReversible()), True)
new_r_id = create_reaction(generalized_model, r_id2st, p_id2st, name=r_name, reversible=reversible,
id_=representative.getId() if len(r_ids) == 1 else None).getId()
elif len(r_ids) > 1:
r_id_increment += 1
new_r_id = generate_unique_id(input_model, "r_g_", r_id_increment)
if len(r_ids) > 1:
for r_id in r_ids:
r_id2g_eq[r_id] = new_r_id, r_name
if groups_sbml and groups_plugin:
# save as a group
r_group = groups_plugin.createGroup()
r_group.setId(new_r_id)
r_group.setKind(libsbml.GROUP_KIND_COLLECTION)
r_group.setSBOTerm(SBO_BIOCHEMICAL_REACTION)
r_group.setName(r_name)
for r_id in r_ids:
member = r_group.createMember()
member.setIdRef(r_id)
add_annotation(r_group, libsbml.BQB_IS_DESCRIBED_BY, GROUP_TYPE_EQUIV)
if out_sbml:
remove_unused_elements(generalized_model)
save_as_sbml(generalized_model, out_sbml)
if groups_sbml and groups_model:
save_as_sbml(groups_model, groups_sbml)
logging.info("serialized to " + groups_sbml)
return r_id2g_eq, s_id2gr_id
def infer_clusters(model, unmapped_s_ids, s_id2clu, s_id2term_id, ubiquitous_chebi_ids, r_ids_to_ignore=None):
# TODO: double check it
return
if not unmapped_s_ids:
return
term_id2s_ids = invert_map(s_id2term_id)
clu2s_ids = invert_map(s_id2clu)
vk2r_ids = get_vk2r_ids(model, s_id2clu, s_id2term_id, ubiquitous_chebi_ids, r_ids_to_ignore=r_ids_to_ignore)
vk2r_ids = {vk: r_ids for (vk, r_ids) in vk2r_ids.items() if len(r_ids) > 1}
simplified_vk2vk_set = defaultdict(set)
for vk in vk2r_ids.keys():
simplified_vk2vk_set[vertical_key2simplified_vertical_key(vk)].add(vk)
s_id2r_ids = defaultdict(list)
for r in (r for r in model.getListOfReactions() if r.getNumReactants() + r.getNumProducts() > 2):
if r_ids_to_ignore and r.getId() in r_ids_to_ignore:
continue
r_id = r.getId()
for s_id in chain((species_ref.getSpecies() for species_ref in r.getListOfReactants()),
(species_ref.getSpecies() for species_ref in r.getListOfProducts())):
s_id2r_ids[s_id].append(r_id)
def in_species_conflict(term, candidate_sps, proposal_s_id2clu):
proposal_clu2s_ids = invert_map(proposal_s_id2clu)
for s in candidate_sps:
s_clu = s.getCompartment(), term
rs = {r_id for r_id in s_id2r_ids[s.getId()]}
clu_s_ids = clu2s_ids[s_clu] | proposal_clu2s_ids[s_clu]
for clu_s_id in clu_s_ids:
if {r_id for r_id in s_id2r_ids[clu_s_id]} & rs:
return True
return False
processed_r_ids = reduce(lambda s1, s2: s1 | s2, vk2r_ids.values(), set())
for r in model.getListOfReactions():
if r.getId() in processed_r_ids or not unmapped_s_ids & get_metabolites(r):
continue
if r_ids_to_ignore and r.getId() in r_ids_to_ignore:
continue
ub_rs, ub_ps, rs, ps = get_vertical_key(model, r, s_id2clu, s_id2term_id, ubiquitous_chebi_ids)
vk = ub_rs, ub_ps, rs, ps
rs, ps = set(rs), set(ps)
partial_rs, partial_ps = {(s_id, c_id) for (s_id, c_id) in rs if s_id not in unmapped_s_ids}, \
{(s_id, c_id) for (s_id, c_id) in ps if s_id not in unmapped_s_ids}
if len(ub_rs) + len(ub_ps) + len(partial_rs) + len(partial_ps) < 2:
continue
simplified_vk = vertical_key2simplified_vertical_key(vk)
if simplified_vk in simplified_vk2vk_set:
ub_rs, ub_ps = tuple(sorted(ub_rs)), tuple(sorted(ub_ps))
for (vk_ub_rs, vk_ub_ps, vk_rs, vk_ps) in simplified_vk2vk_set[simplified_vk]:
vk_rs, vk_ps = {(s_id if s_id not in s_id2clu else s_id2clu[s_id], c_id) for (s_id, c_id) in vk_rs}, \
{(s_id if s_id not in s_id2clu else s_id2clu[s_id], c_id) for (s_id, c_id) in vk_ps}
proposal = {}
if vk_ub_rs == ub_ps and vk_ub_ps == ub_rs and not partial_rs - vk_ps and not partial_ps - vk_rs:
vk_ub_rs, vk_ub_ps, partial_rs, partial_ps = vk_ub_ps, vk_ub_rs, partial_ps, partial_rs
if vk_ub_rs == ub_rs and vk_ub_ps == ub_ps and not partial_rs - vk_rs and not partial_ps - vk_ps:
r_s_ids = rs - vk_rs
p_s_ids = ps - vk_ps
if 0 < len(r_s_ids) <= 1 and 0 < len(p_s_ids) <= 1 and r_s_ids or p_s_ids:
if r_s_ids and vk_rs - rs:
s_id, c_id = r_s_ids.pop()
# if it is not a species id but a cluster, continue
if not isinstance(s_id, str):
continue
clu, c_id = (vk_rs - rs).pop()
# if it is a species id instead of a cluster, continue
if not isinstance(clu, tuple):
continue
candidate_sps = {model.getSpecies(it) for it in
(term_id2s_ids[s_id2term_id[s_id]] if s_id in s_id2term_id else {s_id})}
comp, term = clu
if not in_species_conflict(term, candidate_sps, proposal):
for s in candidate_sps:
proposal[s.getId()] = s.getCompartment(), term
else:
continue
if p_s_ids and vk_ps - ps:
s_id, c_id = p_s_ids.pop()
# if it is not a species id but a cluster, continue
if not isinstance(s_id, str):
continue
clu, c_id = (vk_ps - ps).pop()
# if it is a species id instead of a cluster, continue
if not isinstance(clu, tuple):
continue
candidate_sps = {model.getSpecies(it) for it in
(term_id2s_ids[s_id2term_id[s_id]] if s_id in s_id2term_id else {s_id})}
comp, term = clu
if not in_species_conflict(term, candidate_sps, proposal):
for s in candidate_sps:
proposal[s.getId()] = s.getCompartment(), term
else:
continue
if proposal:
s_id2clu.update(proposal)
for s_id, clu in proposal.items():
clu2s_ids[clu].add(s_id)
unmapped_s_ids -= set(proposal.keys())
def generalize_model(in_sbml,
chebi,
groups_sbml,
out_sbml,
ub_s_ids=None,
ub_chebi_ids=None,
ignore_biomass=True):
"""
Generalizes a model.
:param in_sbml: str, path to the input SBML file
:param chebi: mod_sbml.onto.obo_ontology.Ontology ChEBI ontology
:param groups_sbml: str, path to the output SBML file (with groups extension)
:param out_sbml: str, path to the output SBML file (generalized)
:param ub_s_ids: optional, ids of ubiquitous species (will be inferred if set to None)
:param ub_chebi_ids: optional, ids of ubiquitous ChEBI terms (will be inferred if set to None)
:param ignore_biomass: boolean, whether to ignore the biomass reaction (and its stoichiometry preserving constraint)
:return: tuple (r_id2g_eq, s_id2gr_id, s_id2chebi_id, ub_s_ids):
dict {reaction_id: reaction_group_id}, dict {species_id: species_group_id}, dict {species_id: ChEBI_term_id},
collection of ubiquitous species_ids.
"""
# input_model
input_doc = libsbml.SBMLReader().readSBML(in_sbml)
input_model = input_doc.getModel()
r_ids_to_ignore = get_biomass_r_ids(
input_model) if ignore_biomass else None
remove_is_a_reactions(input_model)
annotate_metabolites(input_model, chebi)
# TODO: fix comp separation
# separate_boundary_metabolites(input_model)
remove_unused_elements(input_model)
logging.info("mapping species to ChEBI")
s_id2chebi_id = get_species_id2chebi_id(input_model)
ub_chebi_ids, ub_s_ids = get_ub_elements(input_model, chebi, s_id2chebi_id,
ub_chebi_ids, ub_s_ids)
terms = (t for t in (chebi.get_term(t_id)
for t_id in s_id2chebi_id.values()) if t)
old_onto_len = len(chebi)
filter_ontology(chebi,
terms,
relationships=EQUIVALENT_RELATIONSHIPS,
min_deepness=3)
logging.info('Filtered the ontology from %d terms to %d' %
(old_onto_len, len(chebi)))
threshold = min(max(3, int(0.1 * input_model.getNumReactions())),
UBIQUITOUS_THRESHOLD)
s_id2clu, ub_s_ids = generalize_species(input_model,
s_id2chebi_id,
ub_s_ids,
chebi,
ub_chebi_ids,
threshold,
r_ids_to_ignore=r_ids_to_ignore)
logging.info("generalized species")
r_id2clu = generalize_reactions(input_model,
s_id2clu,
s_id2chebi_id,
ub_chebi_ids,
r_ids_to_ignore=r_ids_to_ignore)
logging.info("generalized reactions")
clu2s_ids = {(c_id, term): s_ids
for ((c_id, (term, )), s_ids) in invert_map(s_id2clu).items()}
r_id2g_eq, s_id2gr_id = save_as_comp_generalized_sbml(
input_model, out_sbml, groups_sbml, r_id2clu, clu2s_ids, ub_s_ids,
chebi)
return r_id2g_eq, s_id2gr_id, s_id2chebi_id, ub_s_ids
def filter_clu_to_terms(term2clu):
clu2term = invert_map(term2clu)
for clu, terms in clu2term.items():
if len(terms) == 1:
del term2clu[terms.pop()]
def suggest_clusters(model,
unmapped_s_ids,
term_id2clu,
s_id2term_id,
ubiquitous_chebi_ids,
r_ids_to_ignore=None):
# TODO: double check it
return
if not unmapped_s_ids:
return
s_id2clu = compute_s_id2clu(set(), model, s_id2term_id, term_id2clu)
term_id2s_ids = invert_map(s_id2term_id)
vk2r_ids = get_vk2r_ids(model,
s_id2clu,
s_id2term_id,
ubiquitous_chebi_ids,
r_ids_to_ignore=r_ids_to_ignore)
vk2r_ids = {
vk: r_ids
for (vk, r_ids) in vk2r_ids.items() if len(r_ids) > 1
}
processed_r_ids = reduce(lambda s1, s2: s1 | s2, vk2r_ids.values(), set())
s_vk2vk = defaultdict(set)
for vk in vk2r_ids.keys():
s_vk2vk[vertical_key2simplified_vertical_key(vk)].add(vk)
s_id2r_ids = defaultdict(list)
for r in (r for r in model.getListOfReactions()
if r.getNumReactants() + r.getNumProducts() > 2):
if r_ids_to_ignore and r.getId() in r_ids_to_ignore:
continue
r_id = r.getId()
for s_id in chain((species_ref.getSpecies()
for species_ref in r.getListOfReactants()),
(species_ref.getSpecies()
for species_ref in r.getListOfProducts())):
s_id2r_ids[s_id].append(r_id)
for r in model.getListOfReactions():
if r.getId(
) in processed_r_ids or not unmapped_s_ids & get_metabolites(r):
continue
if r_ids_to_ignore and r.getId() in r_ids_to_ignore:
continue
ub_rs, ub_ps, rs, ps = get_vertical_key(model, r, s_id2clu,
s_id2term_id,
ubiquitous_chebi_ids)
vk = ub_rs, ub_ps, rs, ps
rs, ps = set(rs), set(ps)
partial_rs, partial_ps = {(s_id, c_id) for (s_id, c_id) in rs if s_id not in unmapped_s_ids}, \
{(s_id, c_id) for (s_id, c_id) in ps if s_id not in unmapped_s_ids}
if vk in vk2r_ids or len(ub_rs) + len(ub_ps) + len(partial_rs) + len(
partial_ps) < 2:
continue
s_vk = vertical_key2simplified_vertical_key(vk)
if s_vk in s_vk2vk:
ub_rs, ub_ps = tuple(sorted(ub_rs)), tuple(sorted(ub_ps))
for (vk_ub_rs, vk_ub_ps, vk_rs, vk_ps) in s_vk2vk[s_vk]:
vk_rs, vk_ps = {(s_id if s_id not in s_id2clu else s_id2clu[s_id], c_id) for (s_id, c_id) in vk_rs}, \
{(s_id if s_id not in s_id2clu else s_id2clu[s_id], c_id) for (s_id, c_id) in vk_ps}
proposal = {}
if vk_ub_rs == ub_ps and vk_ub_ps == ub_rs and not partial_rs - vk_ps and not partial_ps - vk_rs:
vk_ub_rs, vk_ub_ps, partial_rs, partial_ps = vk_ub_ps, vk_ub_rs, partial_ps, partial_rs
if vk_ub_rs == ub_rs and vk_ub_ps == ub_ps and not partial_rs - vk_rs and not partial_ps - vk_ps:
r_s_ids = rs - vk_rs
p_s_ids = ps - vk_ps
if 0 < len(r_s_ids) <= 1 and 0 < len(
p_s_ids) <= 1 and r_s_ids or p_s_ids:
if r_s_ids and vk_rs - rs:
s_id, c_id = r_s_ids.pop()
clu, c_id = (vk_rs - rs).pop()
# if it is a species id instead of a cluster, continue
if not isinstance(clu, tuple):
continue
# if it is not a species id but a cluster, continue
if not isinstance(s_id, str):
continue
candidate_sps = {
model.getSpecies(sp_id)
for sp_id in (
term_id2s_ids[s_id2term_id[s_id]]
if s_id in s_id2term_id else [s_id])
}
comp, term = clu
for s in candidate_sps:
proposal[s.getId()] = term
if p_s_ids and vk_ps - ps:
s_id, c_id = p_s_ids.pop()
clu, c_id = (vk_ps - ps).pop()
# if it is a species id instead of a cluster, continue
if not isinstance(clu, tuple):
continue
# if it is not a species id but a cluster, continue
if not isinstance(s_id, str):
continue
candidate_sps = {
model.getSpecies(it)
for it in (term_id2s_ids[s_id2term_id[s_id]]
if s_id in s_id2term_id else {s_id})
}
comp, term = clu
for s in candidate_sps:
proposal[s.getId()] = term
if proposal:
for s_id, clu in proposal.items():
term_id2clu[s_id] = (
clu, ) if not (isinstance(clu, tuple)) else clu
unmapped_s_ids -= {s_id}
def save_as_comp_generalized_sbml(input_model, out_sbml, groups_sbml, r_id2clu, clu2s_ids, ub_sps, onto):
logging.info("serializing generalization")
s_id_increment, r_id_increment = 0, 0
if groups_sbml:
doc = convert_to_lev3_v1(input_model)
groups_model = doc.getModel()
groups_plugin = groups_model.getPlugin("groups")
if groups_plugin:
logging.info(" saving ubiquitous species annotations")
s_group = groups_plugin.createGroup()
s_group.setId("g_ubiquitous_sps")
s_group.setKind(libsbml.GROUP_KIND_COLLECTION)
s_group.setSBOTerm(SBO_CHEMICAL_MACROMOLECULE)
s_group.setName("ubiquitous species")
for s_id in ub_sps:
member = s_group.createMember()
member.setIdRef(s_id)
add_annotation(s_group, libsbml.BQB_IS_DESCRIBED_BY, GROUP_TYPE_UBIQUITOUS)
if out_sbml:
# generalized model
generalized_doc = convert_to_lev3_v1(input_model)
generalized_model = generalized_doc.getModel()
for _ in range(0, generalized_model.getNumReactions()):
generalized_model.removeReaction(0)
r_id2g_eq, s_id2gr_id = {}, {}
if not clu2s_ids:
logging.info(" nothing to serialize")
else:
clu2r_ids = invert_map(r_id2clu)
logging.info(" creating species groups")
for ((c_id, t), s_ids) in clu2s_ids.items():
comp = input_model.getCompartment(c_id)
if len(s_ids) > 1:
t = onto.get_term(t)
t_name, t_id = (t.get_name(), t.get_id()) if t \
else (' or '.join(input_model.getSpecies(s_id).getName() for s_id in s_ids), None)
if not t_id:
t = t_name
if out_sbml:
new_species = create_species(model=generalized_model, compartment_id=comp.getId(), type_id=None,
name="{0} ({1}) [{2}]".format(t_name, len(s_ids), comp.getName()))
add_annotation(new_species, libsbml.BQB_IS, t_id, CHEBI_PREFIX)
new_s_id = new_species.getId()
else:
s_id_increment += 1
new_s_id = generate_unique_id(input_model, "s_g_", s_id_increment)
for s_id in s_ids:
s_id2gr_id[s_id] = new_s_id, t
if groups_sbml and groups_plugin:
# save as a group
s_group = groups_plugin.createGroup()
s_group.setId(new_s_id)
s_group.setKind(libsbml.GROUP_KIND_CLASSIFICATION)
s_group.setSBOTerm(SBO_CHEMICAL_MACROMOLECULE)
g_name = "{0} [{1}]".format(t_name, comp.getName())
s_group.setName(g_name)
# logging.info("%s: %d" % (g_name, len(s_ids)))
if t_id:
add_annotation(s_group, libsbml.BQB_IS, t_id, CHEBI_PREFIX)
for s_id in s_ids:
member = s_group.createMember()
member.setIdRef(s_id)
add_annotation(s_group, libsbml.BQB_IS_DESCRIBED_BY, GROUP_TYPE_EQUIV)
generalize_species = lambda species_id: s_id2gr_id[species_id][0] if (species_id in s_id2gr_id) else species_id
s_id_to_generalize = set(s_id2gr_id.keys())
logging.info(" creating reaction groups")
for clu, r_ids in clu2r_ids.items():
representative = input_model.getReaction(list(r_ids)[0])
r_name = "generalized %s" % representative.getName()
if out_sbml:
reactants = dict(get_reactants(representative, stoichiometry=True))
products = dict(get_products(representative, stoichiometry=True))
if (len(r_ids) == 1) and \
not ((set(reactants.keys()) | set(products.keys())) & s_id_to_generalize):
create_reaction(generalized_model, reactants, products, name=representative.getName(),
reversible=representative.getReversible(), id_=representative.getId())
continue
r_id2st = {generalize_species(it): st for (it, st) in reactants.items()}
p_id2st = {generalize_species(it): st for (it, st) in products.items()}
reversible = next((False for r_id in r_ids if not input_model.getReaction(r_id).getReversible()), True)
new_r_id = create_reaction(generalized_model, r_id2st, p_id2st, name=r_name, reversible=reversible,
id_=representative.getId() if len(r_ids) == 1 else None).getId()
elif len(r_ids) > 1:
r_id_increment += 1
new_r_id = generate_unique_id(input_model, "r_g_", r_id_increment)
if len(r_ids) > 1:
for r_id in r_ids:
r_id2g_eq[r_id] = new_r_id, r_name
if groups_sbml and groups_plugin:
# save as a group
r_group = groups_plugin.createGroup()
r_group.setId(new_r_id)
r_group.setKind(libsbml.GROUP_KIND_COLLECTION)
r_group.setSBOTerm(SBO_BIOCHEMICAL_REACTION)
r_group.setName(r_name)
for r_id in r_ids:
member = r_group.createMember()
member.setIdRef(r_id)
add_annotation(r_group, libsbml.BQB_IS_DESCRIBED_BY, GROUP_TYPE_EQUIV)
if out_sbml:
remove_unused_elements(generalized_model)
save_as_sbml(generalized_model, out_sbml)
if groups_sbml and groups_model:
save_as_sbml(groups_model, groups_sbml)
logging.info("serialized to " + groups_sbml)
return r_id2g_eq, s_id2gr_id