*4.1. Experimental Materials and Methods*

Culture maintenance and cultivations were performed on a medium with sea salts and yeast extract, as described previously [30]. In brief, *Crypthecodinium cohnii* CCMP 316 was obtained from the National Center for Marine Algae and Microbiota, USA. It was cultivated on a complex medium containing 2 g L−<sup>1</sup> yeast extract, 25 g L−<sup>1</sup> sea salt (Sigma-Aldrich) and various concentrations of glycerol, glucose and/or ethanol, as specified in the Results section. Cultivations were carried out aerobically at 25 ◦C in 0.5 L or 1 L Erlenmeyer shaken flasks with 200 mL of culture on a rotary shaker at 140–180 r.p.m. The concentrations of glucose, ethanol and glycerol in culture media were monitored by HPLC, as described previously [30,48].

FTIR spectra of algal biomass were recorded using Vertex 70 coupled with the microplate reader HTS-XT (Bruker, Germany). Spectra were recorded in the frequency range of 3800–600 cm−1, with a spectral resolution of 4 cm−1, and 64 scans were coadded. Only spectra with absorbance within the absorption limits between 0.25 and 0.80 (where the concentration of a component is proportional to the intensity of the absorption band) were used for data analysis. The FTIR spectra were vector normalized and deconvoluted (second derivative) for more precise evaluation of weak-intensity spectral bands and to resolve the overlapping components, if any [49]. Data were processed using OPUS 7.5 software (Bruker Optics GmbH, Ettlingen, Germany). The baseline of each spectrum was corrected by the rubber band method.

**Figure 7.** Information flow-to and -from pathway-scale kinetic model and central carbon metabolism scale constraint-based stoichiometric model.

#### *4.2. Development of a Pathway-Scale Kinetic Model*

The model was developed in *COPASI* (COmplex PAthway SImulator) simulation software [50,51] version 4.34 (Build 251). The estimation for kinetic equation parameters that were not found in literature or databases was conducted using built-in parameter estimation functionality using global stochastic optimization methods. The model-specific parameter estimation performance of global stochastic optimization methods implemented in *COPASI* was tested using *ConvAn* software [52]. During parameter estimation, multiple parallel optimization runs were applied, using *COPASI* wrapper *SpaceScanner* [53] to select the most efficient global stochastic optimization algorithms, reducing misinterpretation risks of optimization results [54]. The total concentration of used amino acids in the reactions included in the model was limited to avoid unnecessarily high enzyme concentrations that would not be evolutionarily favorable.

Model parameters were either obtained from the literature or inferred from experimental data. An additional parameter *Vm* was added to the reactions used from [36] to change the *V*max of these reactions without changing the *V*<sup>f</sup> to *V*<sup>r</sup> ratios. Kinetic equations of all enzymatic reactions had overexpression coefficients k that could be used for optimizing enzyme concentrations to increase Acetyl-CoA or other molecule production. Currently, all coefficients *k* = 1 so that the model corresponds to wild-type concentrations of enzymes.

Species concentration constraints were applied in the parameter estimation task in *COPASI*. From Park et al. [55], it was implemented as a constraint that the metabolite concentrations in this model should not exceed 12 mmol/L, except for cellular ethanol, which was allowed to reach 32 mmoL/L.

The metabolic flux unit in the kinetic model is mmoL·min−1·L−<sup>1</sup> since it is frequently used in kinetic models, while in the stoichiometric model, the metabolic flux unit is mmoL·gDW−1·h−1. To transition from dry-weight-related measurements to absolute weight, it was assumed that dry weight made up 33% of the absolute weight and the cell density was 1 g·mL−1. Mitochondrial volume was made to be 1% of the cytosol volume [56].

To determine the parameters that were dependent on enzyme concentrations, three sets of experimental data were used. The first set included reaction fluxes adapted from Cui et al. 13C metabolic flux analysis for growth on glucose. The second and third sets included experimentally measured glycerol and ethanol consumption rates (Section 2.1) determined during this study. The experimental data of glucose consumption were not used in parameter estimation due to the high similarity with 13C experimental data. The *Parameter estimation* task was used in *COPASI*; the data sets were added as different experiments.
