Roadmap
The current scope of EnzymeRates.jl covers the three pillars described in Getting Started: deriving rate equations, fitting them to data, and identifying the best mechanism by cross-validation. The directions below are under consideration for future development.
Support more mechanisms
EnzymeRates derives and identifies a broad space of mechanisms, but several families are not yet covered.
KNF (sequential) allostery. The current MWC model treats all subunits as switching conformation in concert. A Koshland–Némethy–Filmer (sequential) model would let subunits change conformation individually, capturing a complementary class of cooperative behavior.
Iso mechanisms. In an iso mechanism the enzyme takes on a different conformation as it converts from the substrate-bound to the product-bound species, and that conformation relaxes back to the original only through a separate, slow steady-state isomerization step (Iso mechanisms works through an example). The derivation already handles these, but the enumeration does not yet generate iso mechanisms, so identify_rate_equation cannot reach them.
Support more measurement types
Fitting and identification currently work from reaction-rate (turnover) data. A planned extension would support binding and affinity observables as well: the fractional binding of one or more metabolites to the enzyme, an apparent Km, or a metabolite dissociation constant Kd. A mechanism could then be derived against and constrained by binding measurements, not rate alone, and report these quantities directly.
Parameter estimation and identifiability
Beyond selecting the best mechanism, a planned analysis would characterize its parameters. Bootstrap resampling of the data would estimate a distribution — and hence a confidence interval — for each fitted parameter rather than a single point value. An accompanying identifiability analysis would flag parameters, individually or in combination, that the data cannot constrain, giving a clear picture of what the experiment actually resolves.
Plotting
There is no built-in visualization. A planned plotting module would overlay fitted rate equations on data, making it straightforward to inspect fit quality and compare the best and runner-up mechanisms visually.
Outlier dataset identification
The cross-validation framework already computes per-fold losses. A planned extension would use those fold losses to flag measurement groups that the selected mechanism fits poorly — surfacing suspect datasets or experimental conditions that violate the assumed mechanism.