def add_reaction_data(f, reaction_number, reaction, rev=False):
for (m_id, st) in get_reactants(reaction, True):
if m_id in m_id2i:
f.write("%d,%d,%d\n" % (m_id2i[m_id], reaction_number, st if rev else -st))
for (m_id, st) in get_products(reaction, True):
if m_id in m_id2i:
f.write("%d,%d,%d\n" % (m_id2i[m_id], reaction_number, -st if rev else st))
def get_full_id2mask(r_id2mask, model):
id2mask = defaultdict(lambda: 0)
for r_id, mask in r_id2mask.items():
r = model.getReaction(r_id)
if r:
id2mask[r_id] |= mask
id2mask.update({c_id: id2mask[c_id] | mask for c_id in get_r_comps(r_id, model)})
id2mask.update({s_id: id2mask[s_id] | mask for s_id in get_reactants(r)})
id2mask.update({s_id: id2mask[s_id] | mask for s_id in get_products(r)})
return id2mask
def model2stoichiometric_matrix(model, s_id2i, r_id2i):
rows, cols, data = [], [], []
for r in model.getListOfReactions():
cols.extend([r_id2i[r.getId()]] * (r.getNumReactants() + r.getNumProducts()))
for s_id, st in get_reactants(r, stoichiometry=True):
rows.append(s_id2i[s_id])
data.append(-st)
for s_id, st in get_products(r, stoichiometry=True):
rows.append(s_id2i[s_id])
data.append(st)
return csr_matrix((np.array(data), (np.array(rows), np.array(cols))), shape=(len(s_id2i), len(r_id2i))).toarray()
def get_exchange_reactions(model):
result = []
for r in model.getListOfReactions():
rs, ps = set(get_reactants(r)), set(get_products(r))
boundary_s_ids = {s_id for s_id in rs if model.getSpecies(s_id).getBoundaryCondition()}
if boundary_s_ids or not rs:
result.append(r.getId())
continue
boundary_s_ids = {s_id for s_id in ps if model.getSpecies(s_id).getBoundaryCondition()}
if boundary_s_ids or not ps:
result.append(r.getId())
return result
def is_reactant(model, t_id, r, s_id2clu, s_id2term_id, ubiquitous_chebi_ids):
ubiquitous_reactants, ubiquitous_products, specific_reactant_classes, specific_product_classes = \
get_key_elements(model, r, s_id2clu, s_id2term_id, ubiquitous_chebi_ids)
if r.getReversible() and need_to_reverse(
ubiquitous_reactants, ubiquitous_products,
specific_reactant_classes, specific_product_classes):
return t_id in {
s_id2term_id[s_id] if s_id in s_id2term_id else s_id
for s_id in get_products(r)
}
else:
return t_id in {
s_id2term_id[s_id] if s_id in s_id2term_id else s_id
for s_id in get_reactants(r)
}
def constraint_exchange_reactions(model, forsed_r_id2rev, prohibited_r_id2rev=None, cofactors=None, min_flux=0.01):
logging.info("Constraining input reactions...")
for r in model.getListOfReactions():
r_id = r.getId()
r_l, r_u = get_bounds(r)
if forsed_r_id2rev and r_id in forsed_r_id2rev:
rev = forsed_r_id2rev[r_id]
if rev:
# reverse the reaction and set positive bounds,
# as if both bounds are negative, the glp solver produces an error
reverse_reaction(r)
set_bounds(r, max(min_flux, -r_u), max(min_flux, -r_l))
forsed_r_id2rev[r_id] = not rev
else:
set_bounds(r, max(min_flux, r_l), max(min_flux, r_u))
r.setReversible(False)
continue
if prohibited_r_id2rev and r_id in prohibited_r_id2rev:
rev = prohibited_r_id2rev[r_id]
if not rev:
reverse_reaction(r)
set_bounds(r, 0, max(0, -r_l))
prohibited_r_id2rev[r_id] = not rev
else:
set_bounds(r, 0, max(0, r_u))
r.setReversible(False)
continue
if not cofactors:
continue
rs, ps = set(get_reactants(r)), set(get_products(r))
boundary_s_ids = {s_id for s_id in rs if model.getSpecies(s_id).getBoundaryCondition()}
if boundary_s_ids or not rs:
if ps - cofactors:
reverse_reaction(r)
set_bounds(r, 0, max(-r_l, 0))
r.setReversible(False)
continue
boundary_s_ids = {s_id for s_id in ps if model.getSpecies(s_id).getBoundaryCondition()}
if boundary_s_ids or not ps:
if rs - cofactors:
set_bounds(r, 0, max(r_u, 0))
r.setReversible(False)
def get_key_elements(model,
r,
s_id2clu,
s_id2term_id,
ubiquitous_chebi_ids,
ignore_ch_ids=get_proton_ch_ids()):
"""
Gets elements that compose a reaction key: ubiquitous_reactants, ubiquitous_products,
specific_reactant_classes, specific_product_classes
:param model: libsbml.Model
:param r: libsbml.Reaction reaction of interest
:param s_id2clu: dict {metabolite_id: (compartment_id, cluster)}
:param s_id2term_id: dict {metabolite_id: ChEBI_term_id}
:param ubiquitous_chebi_ids: set of ubiquitous ChEBI_ids
:param ignore_ch_ids: set of ChEBI_ids to be excluded from the result (by default protons)
if there is anything else in the result
:return: tuple (ubiquitous_reactants, ubiquitous_products,
specific_reactant_classes, specific_product_classes)
"""
def classify(s_ids):
specific, ubiquitous, ignored_ubs = [], [], []
for s_id in s_ids:
c_id = model.getSpecies(s_id).getCompartment()
if ubiquitous_chebi_ids and s_id in s_id2term_id and s_id2term_id[
s_id] in ubiquitous_chebi_ids:
if s_id2term_id[s_id] in ignore_ch_ids:
ignored_ubs.append((s_id2term_id[s_id], c_id))
else:
ubiquitous.append((s_id2term_id[s_id], c_id))
else:
specific.append((s_id2clu[s_id][1] if s_id in s_id2clu else ((
s_id2term_id[s_id] if s_id in s_id2term_id else s_id), ),
c_id))
transform = lambda collection: tuple(sorted(collection))
return transform(specific), transform(ubiquitous), transform(
ignored_ubs)
specific_reactant_classes, ubiquitous_reactants, ignored_reactants = classify(
get_reactants(r))
specific_product_classes, ubiquitous_products, ignored_products = classify(
get_products(r))
if not ubiquitous_reactants and not ubiquitous_products \
and not specific_reactant_classes and not specific_product_classes:
ubiquitous_reactants, ubiquitous_products = ignored_reactants, ignored_products
return ubiquitous_reactants, ubiquitous_products, specific_reactant_classes, specific_product_classes
def get_key_elements(model, r, s_id2clu, s_id2term_id, ubiquitous_chebi_ids, ignore_ch_ids=get_proton_ch_ids()):
"""
Gets elements that compose a reaction key: ubiquitous_reactants, ubiquitous_products,
specific_reactant_classes, specific_product_classes
:param model: libsbml.Model
:param r: libsbml.Reaction reaction of interest
:param s_id2clu: dict {metabolite_id: (compartment_id, cluster)}
:param s_id2term_id: dict {metabolite_id: ChEBI_term_id}
:param ubiquitous_chebi_ids: set of ubiquitous ChEBI_ids
:param ignore_ch_ids: set of ChEBI_ids to be excluded from the result (by default protons)
if there is anything else in the result
:return: tuple (ubiquitous_reactants, ubiquitous_products,
specific_reactant_classes, specific_product_classes)
"""
def classify(s_ids):
specific, ubiquitous, ignored_ubs = [], [], []
for s_id in s_ids:
c_id = model.getSpecies(s_id).getCompartment()
if ubiquitous_chebi_ids and s_id in s_id2term_id and s_id2term_id[s_id] in ubiquitous_chebi_ids:
if s_id2term_id[s_id] in ignore_ch_ids:
ignored_ubs.append((s_id2term_id[s_id], c_id))
else:
ubiquitous.append((s_id2term_id[s_id], c_id))
else:
specific.append((s_id2clu[s_id][1] if s_id in s_id2clu
else ((s_id2term_id[s_id] if s_id in s_id2term_id else s_id), ), c_id))
transform = lambda collection: tuple(sorted(collection))
return transform(specific), transform(ubiquitous), transform(ignored_ubs)
specific_reactant_classes, ubiquitous_reactants, ignored_reactants = classify(get_reactants(r))
specific_product_classes, ubiquitous_products, ignored_products = classify(get_products(r))
if not ubiquitous_reactants and not ubiquitous_products \
and not specific_reactant_classes and not specific_product_classes:
ubiquitous_reactants, ubiquitous_products = ignored_reactants, ignored_products
return ubiquitous_reactants, ubiquitous_products, specific_reactant_classes, specific_product_classes
def is_reactant(model, t_id, r, s_id2clu, s_id2term_id, ubiquitous_chebi_ids):
ubiquitous_reactants, ubiquitous_products, specific_reactant_classes, specific_product_classes = \
get_key_elements(model, r, s_id2clu, s_id2term_id, ubiquitous_chebi_ids)
if r.getReversible() and need_to_reverse(
ubiquitous_reactants, ubiquitous_products, specific_reactant_classes, specific_product_classes):
return t_id in {s_id2term_id[s_id] if s_id in s_id2term_id else s_id for s_id in get_products(r)}
else:
return t_id in {s_id2term_id[s_id] if s_id in s_id2term_id else s_id for s_id in get_reactants(r)}
def save_as_cytoscape_json(n2lo, model, out_json, ub_sp_ids):
"""
Converts a model with the given node and edge layout to a json file readable by Cytoscape.
:param ub_sp_ids: collection of ubiquitous species ids
:param n2lo: node layout as a dictionary {node_id: ((x, y), (w, h)) if node is not ubiquitous
else node_id : {r_ids: ((x, y), (w, h)) for r_ids of
reactions using each duplicated metabolite}}
:param model: SBML model
:param out_json: path where to save the resulting json file.
"""
# let's scale the map so that a minimal node has a width == 16 (so that the labels fit)
h_min, (x_shift, y_shift), _ = get_layout_characteristics(n2lo)
scale_factor = MARGIN * 1.0 / h_min if h_min else 1
(x_shift, y_shift) = shift(scale((x_shift, y_shift), scale_factor), MARGIN, MARGIN)
nodes, edges = [], []
for comp in model.getListOfCompartments():
c_id = comp.getId()
c_name = get_name(comp)
if c_id in n2lo:
(x, y), (w, h) = transform(n2lo[c_id], x_shift, y_shift, scale_factor)
nodes.append(get_node(x=x + w / 2, y=y + h / 2, z=-1,
**{NAME: c_name, ID: c_id, WEIGHT: w, HEIGHT: h, TYPE: TYPE_COMPARTMENT}))
for species in model.getListOfSpecies():
s_id = species.getId()
s_name = get_name(species)
glyph_type = UNSPECIFIED_ENTITY
sbo_term = species.getSBOTermID()
if sbo_term:
sbo_term = sbo_term.upper().strip()
if sbo_term in SBO_2_GLYPH_TYPE:
glyph_type = SBO_2_GLYPH_TYPE[sbo_term]
if s_id in n2lo:
if isinstance(n2lo[s_id], dict):
elements = n2lo[s_id].items()
else:
elements = [('', n2lo[s_id])]
for r_ids, coords in elements:
if not r_ids or next((it for it in (model.getReaction(r_id) for r_id in r_ids) if it), False):
(x, y), (w, h) = transform(coords, x_shift, y_shift, scale_factor)
nodes.append(
get_node(x=x + w / 2, y=y + h / 2, z=1,
**{GLYPH_TYPE: glyph_type, ID: "%s_%s" % (s_id, '_'.join(r_ids)) if r_ids else s_id,
COMPARTMENT: species.getCompartment(), NAME: s_name, WEIGHT: w, HEIGHT: h,
TYPE: TYPE_SPECIES, UBIQUITOUS: s_id in ub_sp_ids,
'ChEBI': get_chebi_id(species)}))
def get_sref_id(s_id):
if isinstance(n2lo[s_id], dict):
for r_ids in n2lo[s_id].keys():
if r_id in r_ids:
return "%s_%s" % (s_id, '_'.join(r_ids))
return s_id
for reaction in model.getListOfReactions():
r_id = reaction.getId()
r_name = get_name(reaction)
ga = get_gene_association(reaction)
if r_id in n2lo:
(x, y), (w, h) = transform(n2lo[r_id], x_shift, y_shift, scale_factor)
nodes.append(get_node(x=x + w / 2, y=y + h / 2, z=1,
**{ID: r_id, NAME: ga, WEIGHT: w, HEIGHT: h, TYPE: TYPE_REACTION,
REVERSIBLE: reaction.getReversible(),
UBIQUITOUS:
next((False for s_id in chain(get_reactants(reaction), get_products(reaction))
if s_id not in ub_sp_ids), True),
'genes': ga, 'r_name': r_name}))
for s_id in get_reactants(reaction):
edges.append(get_edge(**{ID: "%s_%s" % (r_id, s_id), SOURCE: r_id, TARGET: get_sref_id(s_id),
NAME: '%s is a substrate of %s' % (get_name(model.getSpecies(s_id)), r_name),
UBIQUITOUS: s_id in ub_sp_ids, INTERACTION: SUBSTRATE}))
for s_id in get_products(reaction):
edges.append(get_edge(**{ID: "%s_%s" % (r_id, s_id), SOURCE: r_id, TARGET: get_sref_id(s_id),
NAME: '%s is a product of %s' % (get_name(model.getSpecies(s_id)), r_name),
UBIQUITOUS: s_id in ub_sp_ids, INTERACTION: PRODUCT}))
save_cyjson(nodes, edges, out_json)
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 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