import argparse
import os
import pandas as pd
import numpy as np
from dexom_python.imat_functions import imat
from dexom_python.model_functions import load_reaction_weights, read_model, check_model_options
from dexom_python.result_functions import read_solution, write_solution
from dexom_python.enum_functions.enumeration import create_enum_variables
from warnings import warn
[docs]class RxnEnumSolution(object):
def __init__(self, all_solutions, unique_solutions, all_binary, unique_binary,
all_reactions=None, unique_reactions=None, objective_value=-1.):
self.all_solutions = all_solutions
self.unique_solutions = unique_solutions
self.all_binary = all_binary
self.unique_binary = unique_binary
self.all_reactions = all_reactions
self.unique_reactions = unique_reactions
self.objective_value = objective_value
[docs]def rxn_enum(model, reaction_weights, prev_sol=None, rxn_list=[], eps=1e-4, thr=1e-4, obj_tol=1e-3, out_path='enum_rxn',
save=False):
"""
Reaction enumeration method
Parameters
----------
model: cobrapy Model
reaction_weights: dict
keys = reactions and values = weights
prev_sol: imat Solution object
an imat solution used as a starting point
eps: float
activation threshold in imat
thr: float
detection threshold of activated reactions
tlim: int
time limit for imat
tol: float
tolerance for imat
obj_tol: float
variance allowed in the objective_values of the solutions
out_path: str
path to which the solutions are saved if save==True
save: bool
if True, every individual solution is saved in the iMAT solution format
Returns
-------
solution: RxnEnumSolution object
"""
if prev_sol is None:
prev_sol = imat(model, reaction_weights, epsilon=eps, threshold=thr, full=False)
else:
model = create_enum_variables(model=model, reaction_weights=reaction_weights, eps=eps, thr=thr, full=False)
tol = model.solver.configuration.tolerances.feasibility
prev_sol_bin = (np.abs(prev_sol.fluxes) >= thr-tol).values.astype(int)
optimal_objective_value = prev_sol.objective_value - prev_sol.objective_value * obj_tol
all_solutions = [prev_sol]
all_solutions_binary = [prev_sol_bin]
unique_solutions = [prev_sol]
unique_solutions_binary = [prev_sol_bin]
all_reactions = [] # for each solution, save which reaction was activated/inactived by the algorithm
unique_reactions = []
if save: # when saving each individual solution, ensure that the out_path is a directory
os.makedirs(out_path, exist_ok=True)
if not rxn_list:
rxns = list(model.reactions)
rxn_list = [r.id for r in rxns]
for rid in rxn_list:
if rid not in model.reactions:
warn('The following reaction ID was not found in the model: %s' % rid)
else:
idx = np.where(prev_sol.fluxes.index == rid)[0][0]
with model as model_temp:
if rid in model.reactions:
rxn = model_temp.reactions.get_by_id(rid)
# for active fluxes, check inactivation
if prev_sol_bin[idx] == 1:
rxn.bounds = (0., 0.)
# for inactive fluxes, check activation
else:
upper_bound_temp = rxn.upper_bound
# for inactive reversible fluxes, check activation in backwards direction
if rxn.lower_bound < 0.:
try:
rxn.upper_bound = -thr
temp_sol = imat(model_temp, reaction_weights, epsilon=eps, threshold=thr)
temp_sol_bin = (np.abs(temp_sol.fluxes) >= thr-tol).values.astype(int)
if temp_sol.objective_value >= optimal_objective_value:
all_solutions.append(temp_sol)
all_solutions_binary.append(temp_sol_bin)
if not np.any(np.all(temp_sol_bin == unique_solutions_binary, axis=1)):
unique_solutions.append(temp_sol)
unique_solutions_binary.append(temp_sol_bin)
unique_reactions.append(rid+'_backwards')
if save:
filename = out_path+'_solution_'+str(len(unique_solutions)-1)+'.csv'
write_solution(model, temp_sol, thr, filename)
except:
print('An error occurred with reaction %s_reverse. '
'Check feasibility of the model when this reaction is irreversible.' % rid)
finally:
rxn.upper_bound = upper_bound_temp
# for all inactive fluxes, check activation in forwards direction
if rxn.upper_bound >= thr:
rxn.lower_bound = thr
else:
rxn.lower_bound = rxn.upper_bound
# for all fluxes: compute solution with new bounds
try:
temp_sol = imat(model_temp, reaction_weights, epsilon=eps, threshold=thr)
temp_sol_bin = (np.abs(temp_sol.fluxes) >= thr-tol).values.astype(int)
if temp_sol.objective_value >= optimal_objective_value:
all_solutions.append(temp_sol)
all_solutions_binary.append(temp_sol_bin)
all_reactions.append(rid)
if not np.any(np.all(temp_sol_bin == unique_solutions_binary, axis=1)):
unique_solutions.append(temp_sol)
unique_solutions_binary.append(temp_sol_bin)
unique_reactions.append(rid)
if save:
filename = out_path+'_solution_'+str(len(unique_solutions)-1)+'.csv'
write_solution(model, temp_sol, thr, filename)
except:
if prev_sol_bin[idx] == 1:
print('An error occurred with reaction %s. '
'Check feasibility of the model when this reaction is blocked' % rid)
else:
print('An error occurred with reaction %s. '
'Check feasibility of the model when this reaction is irreversible' % rid)
solution = RxnEnumSolution(all_solutions, unique_solutions, all_solutions_binary, unique_solutions_binary,
all_reactions, unique_reactions, prev_sol.objective_value)
return solution
if __name__ == '__main__':
description = 'Performs the reaction enumeration algorithm on a specified list of reactions'
parser = argparse.ArgumentParser(description=description, formatter_class=argparse.RawTextHelpFormatter)
parser.add_argument('-m', '--model', help='Metabolic model in sbml, matlab, or json format')
parser.add_argument('-l', '--reaction_list', default=None, help='csv list of reactions to enumerate - if empty, '
'will use all reactions in the model')
parser.add_argument('--range', default='_',
help='range of reactions to use from the list, in the format "integer_integer", 0-indexed')
parser.add_argument('-r', '--reaction_weights', default=None,
help='Reaction weights in csv format (first row: reaction names, second row: weights)')
parser.add_argument('-p', '--prev_sol', default=None, help='initial imat solution in .txt format')
parser.add_argument('-e', '--epsilon', type=float, default=1e-2,
help='Activation threshold for highly expressed reactions')
parser.add_argument('--threshold', type=float, default=1e-5, help='Activation threshold for all reactions')
parser.add_argument('-t', '--timelimit', type=int, default=None, help='Solver time limit')
parser.add_argument('--tol', type=float, default=1e-7, help='Solver feasibility tolerance')
parser.add_argument('--mipgap', type=float, default=1e-3, help='Solver MIP gap tolerance')
parser.add_argument('--obj_tol', type=float, default=1e-3,
help='objective value tolerance, as a fraction of the original value')
parser.add_argument('-o', '--output', default='rxn_enum', help='Path of output files, without format')
parser.add_argument('--save', action='store_true', help='Use this flag to save each solution individually')
args = parser.parse_args()
model = read_model(args.model)
check_model_options(model, timelimit=args.timelimit, feasibility=args.tol, mipgaptol=args.mipgap)
reaction_weights = {}
if args.reaction_weights is not None:
reaction_weights = load_reaction_weights(args.reaction_weights)
rxn_list = []
if args.reaction_list is not None:
df = pd.read_csv(args.reaction_list, header=None)
reactions = [x for x in df.unstack().values]
else:
reactions = [r.id for r in model.reactions]
rxn_range = args.range.split('_')
if rxn_range[0] == '':
start = 0
else:
start = int(rxn_range[0])
if rxn_range[1] == '':
rxn_list = reactions[start:]
elif int(rxn_range[1]) >= len(reactions):
rxn_list = reactions[start:]
else:
rxn_list = reactions[start:int(rxn_range[1])]
if args.prev_sol is not None:
initial_solution, initial_binary = read_solution(args.prev_sol, model)
model = create_enum_variables(model, reaction_weights, eps=args.epsilon, thr=args.threshold, full=False)
else:
initial_solution = imat(model, reaction_weights, epsilon=args.epsilon, threshold=args.threshold)
solution = rxn_enum(model=model, rxn_list=rxn_list, prev_sol=initial_solution, reaction_weights=reaction_weights,
eps=args.epsilon, thr=args.threshold, obj_tol=args.obj_tol, out_path=args.output, save=args.save)
uniques = pd.DataFrame(solution.unique_binary)
uniques.to_csv(args.output+'_solutions.csv')