2.3.1. Validation of the Model

A medium-scale stoichiometric, central, carbon metabolism model of *C. cohnii* has been developed. The model is organized in three compartments (extracellular, cytosol and mitochondria) and has 398 reactions and 468 metabolites. Out of these 398 reactions, 35 are transport reactions (metabolite uptake, shuttle transport, metabolite output). The model simulates the uptake of the substrates, as well as H2O, O2, H<sup>+</sup> and ammonia, which is available to the *C. cohnii* for uptake in a bioreactor. The model is available in COBRA

format and MS Excel format (Supplementary File S4) and is available in the BioModels database in SBML format (Bomodels ID: MODEL2112300001).

The biomass equation was created by using biomass composition data from Cui et al. [32], determining the amount of each metabolite needed to form 1 gram of biomass [39]. To determine the ratio between the nucleotides that make up the RNA and DNA, *C. cohnii* transcriptome [31] and *Symbiodinium minutum* genome [40] data were used. The unit used for the reaction fluxes in the model is mmol·gDW−1·h<sup>−</sup>1.

The stoichiometric model was validated using published experimental results, as well as experiments performed during this study (Table 2), reaching the specific growth rate when consuming the substrate at the experimentally observed uptake rate.


**Table 2.** Validation data.

The maximal biomass productivity with the given substrate uptake, according to validation data (Table 2), was determined by maximizing biomass production in the stoichiometric model to demonstrate that, in most cases, the μmax of the model is close or higher than the experimentally observed μ (Figure 5), indicating that model predictions are close to the experimentally determined values or above them. Higher model predictions suggest that the growth in the experiment did not reach the maximal rate for unspecified reasons.

**Figure 5.** The stoichiometric model predicted the maximal specific growth rate μmax comparison with experimentally determined μ at experimentally determined values of substrate consumption (Table 2).

2.3.2. Validation of Steady-State Fluxes of the Kinetic Model

The structure of the stoichiometric model includes all reactions of the pathway-scale kinetic model. This enables the feasibility testing of steady-state, pathway-scale kinetic model fluxes within the framework of the medium-scale stoichiometric model, namely, the biomass production at the experimentally determined substrate consumption and intracellular reaction rates of Acetyl-CoA production. This model has been validated by three steady-state flux datasets (Supplementary File S5) of simulations mentioned in Table 1.

The stoichiometric model could simulate the kinetic model steady-state fluxes of glucose consumption, largely due to the fact that the fluxes were based on 13C flux experimental data that covered all relevant branches (Supplementary File S5, Sheet "Glucose").

In the case of glycerol, the kinetic model did not take into account the flux to the pentose phosphate pathway. Therefore, larger flux values for the reactions PGI, PFK and FBA were allowed in the stoichiometric model, and the small kinetic model values in glucose and ethanol uptake were set to zero (Supplementary File S5, Sheet "Glycerol").

The kinetic model steady-state flux set for ethanol consumption also had to be corrected to enable the operation of the pentose phosphate pathway in a similar way, as in the case of glycerol (Supplementary File S5, Sheet "Ethanol"). The transport rates of other substrates were set to zero.

Steady states were reached with the accepted variability of some reactions up to 4% for glucose, 10% for glycerol and 3% for ethanol. This variability was introduced to compensate for potential measurement errors and to meet the full balance pre-condition of constraint-based stoichiometric modeling.
