Development of kinetic models of metabolism has often been constrained to small systems. In S. cerevisiae models, each next step forward in the understanding of glycolysis encountered a new limitation due to the inherent complexity of the pathway.

Models studying glycolytic oscillations or single GP experiments led to an in-depth analysis of glycolytic dynamics, but to understand central carbon metabolism performance, more pathways than only glycolysis must be considered. For instance, a significant fraction of glucose-derived carbon is taken up at different points in glycolysis [34]. To account for this, a relatively simple option is to implement branches or sink reactions (developed for Escherichia coli in [59]). This led S. cerevisiae models to reproduce steady state where imbalance had been mistakenly predicted [22]. Still, dynamic regulation of storage metabolism is more complex than a sink reaction [18,37] and later models gradually added complexity to the trehalose cycle kinetics to avoid the imbalance from happening upon dynamic perturbation [17,18]. A similar situation could happen for other closely linked pathways such as the TCA or PPP, which have mostly been lumped into a single reaction, even though a few exceptions exist [19,20,21,33]. Simultaneously, other approaches such as linlog kinetics have aimed at attaining high model complexity but with simplified expressions using less parameters [89,90,91].

Furthermore, factors such as growth rate, compartmentation, or transport of metabolites other than glucose, regulate glycolytic response but have barely been considered. First, the growth rate determines how sink reactions behave [34], but most models focus only on a unique growth rate of 0.1 h⁻¹. Since the effect of this variable has not been explicitly considered, models simulating different growth rates had no other alternative than to fit a different parameter set each time [17]. Second, compartmentation and transport reactions have barely been considered and, for instance, this is relevant in trehalose regulation since it is known to accumulate in compartments other than the cytosol [92,93]. Third, transport of metabolites such as gases oxygen (O2) and carbon dioxide (CO2) could allow models to explain differences between respiratory and fermentative behavior [27,30,50] but neither has been implemented.

On top of this, other variables affect individual enzyme kinetics, and have neither been considered. First, cytosolic pH decays upon extracellular glucose perturbation, affecting multiple intracellular processes, including enzyme kinetics [69,94]. Second, PTMs are a fast response mechanism and multiple target sites have been found throughout CCM [95]. Third, different enzyme isoforms are expressed under different growth regimes. Examples of this are the differential expression of GLK/HXK and Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) genes ([96] and [97], respectively).

Finally, a key challenge is the representation of variables that are not part of the carbon flux, such as cofactors. Most models have kept them constant or adopted moiety conservation cycles [98], such as the sum of intracellular adenine nucleotides ([ATP] + [ADP] + [AMP] = [AXP]) or inorganic phosphate [17]. Nonetheless, under intense glucose perturbations, both variables behave in a dynamic manner [19,26,31,99] and alter glycolytic response. An example of this is the ATP paradox, which occurs when ATP and the sum of adenine nucleotides transiently decay [100]. Understanding cytosolic Pi as a dynamic variable and implementation of import from the vacuole turned out to be central in understanding the glycolytic imbalance [18]. Although the availability of Pi was essential for lower glycolysis progression via GAPDH [18], adenine nucleotides exert allosteric regulation on the important controlling enzyme PFK [101].

Scale-down approaches have been developed to understand long-standing problems in industrial bioreactors. Although this has granted valuable knowledge, essential intracellular properties such as in vivo fluxes and kinetics have been captured with only limited resolution, constraining model development. In fact, this has become one of the main challenges in the development of high quality predictive kinetic models, since often multiple variables, such as transcriptomics, metabolomics and fluxomic data, interact to result in the final response [6].

Early works aimed to understand glycolytic oscillations did so with small datasets, reducing their range of implementation. On most occasions only extracellular data such as growth and nutrient exchange rates was available [76] or a few metabolites at most [75], until in vivo quantification of metabolite concentrations and fluxes became a common practice, where most cofactors, glycolytic intermediates and rates were simultaneously observable [22]. Later, a standardized dynamic glucose perturbation experimental setup with CEN-PK yeast strains was adopted (see Table 2). This consisted of chemostat growth at dilution rate of 0.1 h⁻¹, followed by an external glucose perturbation, where extracellular concentration increased to 1 g L⁻¹. These stimulus response experiments were used to infer more physiological patterns [26] and the use of Nuclear Magnetic Resonance (NMR) and Mass Spectroscopy (MS) techniques made a wide range of intracellular metabolites measurable. From only a few glycolytic concentrations, datasets gradually grew to include most metabolites in glycolysis, the trehalose cycle, the TCA cycle, and the PPP. Adenine nucleotides and NAD:NADH ratio have also been made a standard and other nucleotides and amino acids which are affected by carbon uptake dynamics are quantified in the most recent publications.