[edkerk]

geckopy and GECKO now give the same result on the same input in the cases below. The cases come from the scenarios of the parity harness in CompMetEng/raven-gecko-parity (the companion changes are in CompMetEng/GECKO).

Sensitivities and tuning

• get_conc_control_coeffs reads the reduced cost of each usage reaction's forward variable and keeps it where a 0.01 percent rise of the bound raises the optimum by at least half of the predicted rise. An enzyme whose own bound does not bind gets 0 while the protein pool binds, where every used enzyme used to get the pool's price. The reduced cost is read from the variable duals of glpk, glpk_exact and gurobi and from the HiGHS bound marginal of the scipy solver; where no usable dual exists the finite-difference loop runs.
• fit_sigma scans the 0.01 grid of GECKO's sigmaFitter by default (method="bisect" stays available). A target that growth cannot reach gives the first sigma of the growth plateau, the bounds of the objective reaction are set to [0, 1000] and stay so, and bisect gives the first grid sigma for a model that cannot grow.
• sensitivity_tuning follows sensitivityTuning.m: tuned reactions are listed in model order with the kcat and source of the input model, ec.notes prints the pre-tune kcat as num2str does, a prot_to_ignore ID is matched without regard to case and raises when unknown, a tuned kcat has no upper limit, and a gecko-light model is tuned.
• truncate_values rounds values exactly halfway away from zero.

Kcat lists

• save_kcat_source writes the detail text as saveKcatSource does (tabs and line breaks become spaces, no empty parts).
• fill_kcats_from_isozymes and get_kcat_across_isozymes default to apply=False and fill only a kcat of 0.
• apply_kcat_constraints lets every isozyme of a reaction compete in light ecModels and treats an empty update list as all reactions.
• apply_custom_kcats reads a quoted reaction list without its quotes.
• set_kcat_for_reactions follows setKcatForReactions: apply defaults to False, IDs match without regard to case, an unknown ID is ignored, and a light-model base ID matches nothing.

Lookups and building

• find_ec_in_db splits an EC cell on ; and whitespace.
• enzyme_usage gives cap_usage NaN where the usage upper bound is 0.
• flexibilize_enz_concs uses the growth reached after the protein-pool branch as its target.
• make_ec_model refuses reaction IDs ending in _REV or _EXP_<n> or starting with usage_prot_ or prot_pool, metabolite IDs starting with prot_, an enzyme_comp that is not a compartment name, and a model without genes. It splits only reversible reactions that are not exchanges, so [-10, -1] and a multi-metabolite sink stay as they are.

Behaviour that callers may notice: the apply defaults of set_kcat_for_reactions and the fill functions, the refusals of make_ec_model (a compartment without a name), the default of fit_sigma, and the removed kcat cap of sensitivity_tuning.

Not checked: the changes on models other than the unit-test ecModel and the tutorial yeast model, cplex and SCIP (not installed), and get_conc_control_coeffs with scipy releases other than the one installed.

Companion changes: CompMetEng/GECKO#13 (GECKO), CompMetEng/raven-gecko-parity#141 (scenarios and ledger), CompMetEng/raven-toolbox#183 (write_yaml_model writes a single ec-code as a list).