Conference Agenda

Sensitivity Analysis (SA) is well-established in systems modelling, aiding the identification and quantification of the impact that parametric uncertainty can have on model outputs (Triantafyllou et al., 2023; Kotidis et al., 2019). Beyond that, Global SA (GSA) is allowing investigation of what is known as “second order interactions”, referring to the investigation of impact that parametric uncertainty can have on the system parameters, instead of outputs (Saltelli et al., 2019; Andalibi et al., 2020). The latter can shed light on underlying, perhaps non-evident, interactions enhancing system understanding. Despite its potential and usefulness, GSA performance is dependent to the model complexity. In this context, large-scale and nonlinear models can render GSA challenging to perform, requiring excessive computational effort. This is further augmented in cases of large sets of parameters. To this end, approaches have been developed, successfully reducing the complexity of the GSA, by segmenting the set to smaller groups of parameters (Sheikholeslami et al., 2019). This, however, can limit the potential of a full-scale GSA as it does not consider the parametric set universally at once. Metamodels have also been used as model surrogates, however they are prone to overfitting in higher dimensions (Becker, 2015).

In this work, we investigate the potential of Machine Learning (ML) to reduce the complexity of Pharmacokinetic/Pharmacodynamic (PK/PD) models. Such models (Abbiati et al., 2018) are suitable for this analysis given their non-linearity and because they involve parameters reflecting patient characteristics. In this context, understanding the parameter space-model output relationship implies linking patient population heterogeneity to the therapeutic outcome variability. Here, we explore how the level of hybridisation can impact the GSA performance, but also, critically, whether the use of surrogates affects the resulting model sensitivity to parametric uncertainty.

Abbiati, R.A., Savoca A., Manca D., (2018) Chapter 2 - An engineering oriented approach to physiologically based pharmacokinetic and pharmacodynamic modeling. Computer Aided Chemical Engineering, 42, 37-63.
https://doi.org/10.1016/B978-0-444-63964-6.00002-7.

Andalibi, M.R., Bowen, P., Carino, A. & Testino, A. (2020) Global uncertainty-sensitivity analysis on mechanistic kinetic models: From model assessment to theory-driven design of nanoparticles. Computers & Chemical Engineering. 140, 106971. doi:10.1016/j.compchemeng.2020.106971.

Kotidis, P., Demis, P., Goey, C.H., Correa, E., McIntosh, C., Trepekli, S., Shah, N., Klymenko, O.V. & Kontoravdi, C. (2019) Constrained global sensitivity analysis for bioprocess design space identification. Computers & Chemical Engineering. 125, 558–568. doi:10.1016/j.compchemeng.2019.01.022.

Saltelli, A., Aleksankina, K., Becker, W., Fennell, P., Ferretti, F., Holst, N., Li, S. & Wu, Q. (2019) Why so many published sensitivity analyses are false: A systematic review of sensitivity analysis practices. Environmental Modelling & Software. 114, 29–39. doi:10.1016/j.envsoft.2019.01.012.

Sheikholeslami, R., Razavi, S., Gupta, H.V., Becker, W. & Haghnegahdar, A. (2019) Global sensitivity analysis for high-dimensional problems: How to objectively group factors and measure robustness and convergence while reducing computational cost. Environmental Modelling & Software. 111, 282–299. doi:10.1016/j.envsoft.2018.09.002.

Triantafyllou, N., Sarkis, M., Shah, N., Kontoravdi, C. & Papathanasiou, M.M. (2023) Integrated Process and Supply Chain Design and Optimization. In: R. Pörtner (ed.). Biopharmaceutical Manufacturing: Progress, Trends and Challenges. Cham, Springer International Publishing. pp. 213–239. doi:10.1007/978-3-031-45669-5_7.

Used cooking oils (UCOs) or waste cooking oils (WCOs) are widely generated urban residues as a result of food preparation at households, restaurants, hosteling sites, and industry. Despite UCOs are highly contaminated with components of different nature, they are mainly composed of triacylglycerols of fatty acids, which can be used as raw materials for the oleochemical industry. However, a significant challenge in the valorization of UCOs is their high content of free fatty acids (FFA), which can lead to equipment corrosion, deactivate alkaline catalysts during transesterification reactions and reduce yields in various processes. To address these issues, it is necessary to reduce the acidity of UCOs. Industrially, this is typically achieved through high-energy and materials-intensive processes such as low vacuum distillation and neutralization (Noriega et al., 2022). For this reason, alternative processes such as alcoholic extraction have proven to be very effective, exhibiting lower energy consumption, reduced residue generation, and the potential to utilize the recovered free fatty acids (FFAs) through further esterification with the solvent.

In recent studies it was verified that FFAs could be removed from acid vegetable oils and/or UCOs by using ethanolic extraction in a high surface area liquid-liquid contactor under continuous operation (Cárdenas et al., 2022; Noriega et al. 2022). This equipment seeks to maximize the contact area between liquid phases using a semi structural packing, which in turn allows the equipment to operate under laminar regime reducing dispersion and facilitating downstream decantation. In a previous experimental exploration (Cárdenas et al., 2022) it was found that in a single stage contactor of 1.07m, it was possible to reduce UCO’s acidity by 51%, whereas using acidified palm oil (Noriega et al., 2022) it was possible to reduce acidity below 0.1 wt.%. In both cases, separation was carried out in a single stage contactor under a fixed ethanol-to-oil ratio.

In this regard, this work was aimed to develop and correlate a mathematical model to describe the operation of a liquid-liquid film contactor in the FFAs extraction from UCOs. In previous research (Noriega et al., 2022), it was established that mass transfer in the oil phase was the primary resistance affecting the overall process. Accordingly, the developed model proposes a rigorous description of the liquid-liquid mass transfer in the oil phase based upon previously validated phase equilibria data and reported physicochemical properties of the mixture. Additionally, by applying subsequent stochastic and deterministic optimization algorithms, the mass transfer parameters were adjusted by using reported data from deacidification experiments carried out under different UCO flowrates, ethanol to UCO mass flow ratios and contactor lengths (Cárdenas et al., 2022). Once the parameters were regressed and validated, the liquid-liquid extraction model was used to determine the best operating conditions and configuration to carry out the deacidification of UCOs. It was found that a multistage configuration was required to reduce UCO’s acidity below the specifications of oleochemical feedstocks (< 0.5 wt.%), and that the contactor enabled to intensify the mass transfer process in comparison with traditional configurations of mixed tank contactors with settlers.

The human digestion is a complex phenomenon that takes places by utilizing multiple resources and processes so as to ensure the longevity and maximal absorption of essential nutrients. Due to the controversial and costly nature of human medical intervention for research purposes, an advanced demand for in vitro model systems is evident throughout the recent years. A major drawback of these systems though is their laborious and expensive operation as well as the contained experimentation extent, as a result of aiming to mimic accurately physiologically based processes. Moreover, their susceptibility towards environmental conditions (e.g. survival of gut microbiota), highlights the need for radical in silico models, thus the importance for creating a FooDSIM digital twin was given birth to.
Main aim of this work is to create a mathematical digital twin that will perform predictive modelling, while upper aim is the future incorporation of gut microbiota’s microbial interactions in the system, under steady state conditions of operation and intake of different input formulas.
For the formulation of the model, processes as hydraulics, biochemical interactions between enzymes and substrates, as well as their dependency on pH are expressed through Modified Ordinary Differential Equations (ODEs), and the simulation is performed at 3 different intervals (6 hours, 24 hours, 7 days). To accomplish this, the digital twin simulates the human gastrointestinal tract (GIT) by employing 4 bioreactors in series, which represent distinct organs of the GIT. The emptying profile of the 1st reactor representing the stomach, follows the Elashoff exponential equation, and influences all subsequent feeding and emptying processes taking place in the system, where the addition of simulated digestion fluids, enzyme solutions and of a food model system as input, creates a standardized model environment. Lastly, the absorption processed is simulated by expressing the function of a hollow fibre membrane, which operates with a cut-off limit and restricts the passage of molecules based on their molecular weight.
During the simulations, enzyme concentration increases initially because they cannot be absorbed during dialysis, thus affecting the kinetic parameters used to simulate interactions between enzymes and substrates, which follow Michaelis-Menten kinetics. The pH’s impact on enzyme activity is also emphasized, as pH levels outside the optimal range cause enzyme inactivation. To model this condition, enzyme activity is fitted to two different cardinal models, and their goodness of fit is assessed. For the shake of simulation, MATLAB’s ode15s solver is used due to the system’s stiffness and complexity, whereas simulations interval progress, enzyme concentration changes, impacting substrate digestion. Initial results show system stabilization after 3-4 feeding cycles, with substrate concentrations plateauing after 24-30 hours. The dialysis phase causes significant enzyme concentration changes, affecting substrate consumption rates. Cardinal models show good performance (MSE < 0.0114 and 0.0134), while only pancreatic trypsin follows the second model. GSA reveals key interactions between substrates and enzyme activity, especially for starch

To reduce the global dependence on fossil fuels, large scale production of biofuels becomes increasingly important. A promising class of potential biofuels are methyl ketones that can be produced via fermentation ^[1]. A big challenge is the downstream purification of methyl ketones as it requires an industrially scalable process for the extraction of a hydrophobic product that is highly diluted in an aqueous fermentation broth. A key factor in ensuring an economically viable process is the selection of a suitable solvent ^[2].

This study focusses on the simultaneous design of a suitable extraction solvent along with the respective extraction distillation process to minimize the expected overall cost of the methyl ketone fermentation process ^[3]. Therefore, this study extends and refines a previously developed framework for the simultaneous design of molecules, processes parameters and equipment size ^[4].

The framework is based on a genetic algorithm that iteratively optimizes solvent structure, process parameters and equipment size to minimize the overall costs including the estimation of equipment size ^[5]. In each generation, the genetic algorithm generates a fixed number of potential solvents from a given set of molecular fragments. All relevant physical properties of the constructed solvents are then predicted via methods of computational chemistry and passed to the process model. The process model optimizes process parameters based on operating and investment cost for all created molecules of a generation and ranks them according to total cost. This information is passed back to the genetic algorithm and used to improve the next generation.

In this study, the framework is refined by three major improvements. First, the integration of the Rectification Body Method ^[6] complemented by a sequencing algorithm for a more accurate simulation of the distillation. Second, the framework is extended by a screening-guided warm start method for CAMD design ^[7]. This method improves the convergence speed of the genetic algorithm and furthermore allows the design of more suitable solvent candidates by improving the library of molecular fragments used to source solvent candidates. Third, we evaluate the potential of predictive methods for considering bio compatibility. We show that the algorithm is able to identify promising solvent candidates outperforming literature solvents, highlighting the potential of molecule design and the importance of early-stage process and equipment design.

References:

[1] Grütering et al. (2024), DOI: 10.1039/D4SE00035H

[2] Zhou et al. (2020), DOI: 10.1016/j.coche.2019.10.007

[3] Ziegler et al.(2023), DOI: 10.1093/jimb/kuad029.

[4] Polte et al. (2023), DOI: 10.1002/cite.202200144

[5] Douguet et al. (2005), DOI: 10.1021/jm0492296

[6] Bausa et al.(1998), DOI: 10.1002/aic.690441008

[7] Wang et al. (?),DOI: in submission.

Approximately 13.2% of the world's food is lost before being harvested (United Nations, 2023). Thus, food waste management has high relevance. In Mexico, the agricultural sector contributes with 3.4% of the national GDP, where the fruits and vegetables sector stand out with 45% of the sector's exportations. Tomatoes contributes to exportation with 8.41%. Mexico is the main supplier of this product worldwide, with a 19% share of the exported volume in the period 2003-2017 (Montaño Méndez et. al., 2021). However, not all tomatoes are used due to their short shelf life, causing approximately 30% of the crop to be wasted (Méndez-Carmona et al., 2022). In addition, waste from tomato sauce production can also be valorized. Tomato waste can be used to isolate high-value compounds such as carotenoids, polyphenols, vitamins, fibers, flavonoids, among others. Among these, lycopene is considered a well-known carotenoid and the most abundant pigment in tomatoes, responsible for giving them their color. It is used as a raw material in the cosmetic, pharmaceutical and food industries (Kuvendziev et al., 2024). Lycopene from tomato waste is traditionally recovered by solvent extraction, but research on this topic has only been carried out through laboratory-scale experimentation (e.g. Poojary et al., 2015; Catalkaya et al., 2019).

This work addresses the systematic design and optimization of an industrial-scale process to obtain lycopene from tomato waste. A comparison is made between the use of conventional solvents, as the mixture acetone: hexane, with an ethanol-based extraction. As results, industrial-scale processes have been proposed and designed to produce lycopene from tomato waste. The process include the drying of the tomato waste, grinding, extraction of lycopene, and recovery of the solvent. Yields of lycopene from dried tomato waste of 21.96 mg/kg and 4.03 mg/kg are obtained for acetone: hexane and ethanol, respectively. The acetone: hexane route has been identified as a promissory route in terms of yield, but the ethanol route is promissory in terms of environmental impact and potential application in the food and pharmaceutical areas.

References

Catalkaya, G., Kahveci, D. (2019). Optimization of enzyme assisted extraction of lycopene from industrial tomato waste. Separation and Purification Technology, 219, 55-63.

Kuvendziev, S., Lisichkov, K., Marinkovski, M., Stojchevski, M., Dimitrovski, D., Andonovikj, V. (2024). Valorization of tomato processing by-products: Predictive modeling and optimization for ultrasound-assisted lycopene extraction. Ultrasonics Sonochemistry, 110, 107055.

Montaño Méndez, I.E., Valenzuela Patrón, I.N., Villavicencio López, K.V. 2021. Competitividad del tomate rojo de México en el mercado internacional: análisis 2003-2017. Revista Mexicana de Ciencias Agrícolas, 12 (7), 1185-1197. (Spanish)

Méndez-Carmona, J.Y., Ascacio-Valdez, J.A, Alvarez-Perez, O.B., Hernández-Almaraz, A-Y., Ramirez-Guzman, N., Sepúlveda, L., Aguilar-González, M.A., Ventura-Sobrevilla, J.M., Aguilar, C.N. 2022. Tomato waste as a bioresource for lycopene extraction using emerging technologies. Food Bioscience, 49, 101966.

United Nations. 2023. The Sustainable Development Goals Report: Special Edition. At https://unstats.un.org/sdgs/report/2023/The-Sustainable-Development-Goals-Report-2023.pdf (accessed september 20, 2024).

It is estimated that 79% of all plastics ever produced have accumulated in the environment, while only 9% have been incinerated and a mere 12% recycled¹. Furthermore, about 95.5% of plastics are petroleum-derived (PlasticsEurope), whose environmental impact is no longer a mystery to anyone. To address the dual challenges of plastic pollution and fossil fuel dependence, fully biodegradable plastics made from sustainable biological resources present a promising and environmentally friendly alternative. In the realm of bioplastics, polyhydroxyalkanoates (PHAs) stand out as a prominent class of naturally occurring intracellular microbial polyesters with poly(3-hydroxybutyrate) (PHB) being the model and most studied member of the family ². Large scale production of PHAs is hindered mainly due to high feedstock costs and low PHA productivities. A promising solution to these challenges is integrating PHB production into biodiesel facilities, using crude glycerol—a byproduct of biodiesel production—as the fermentation substrate ^3,4. This promising bioprocess can be systematically designed and optimized in silico, eliminating the need for time-consuming trial-and-error experiments. Furthermore, to the best of our knowledge, no prior research has specifically examined the individual steps of PHB biodegradation or the factors influencing the degradation rate at each stage.

In continuation of previous work conducted in our research group ^3,4, we are developing a holistic mechanistic kinetic model for PHB production by Cupriavidus necator DSM 545, using glycerol as the sole carbon source and ammonium sulfate (AS) as the nitrogen source. The model is constructed on data from various batch and fed-batch bioreactor fermentations conducted under controlled pH 6.8 and dissolved oxygen (DO) at 30% saturation, at a range of initial concentrations of carbon, nitrogen, and biomass to derive the kinetic constants. Additionally, the obtained PHBs are subjected to biodegradation under different processing conditions to identify the relationship between these parameters and the biodegradation rate.

References

1. Dimante-Deimantovica, I. et al. Downward migrating microplastics in lake sediments are a tricky indicator for the onset of the Anthropocene. Sci. Adv. 10, eadi8136 (2024).

2. Park, H. et al. PHA is not just a bioplastic! Biotechnol. Adv. 71, 108320 (2024).

3. Pérez Rivero, C., Sun, C., Theodoropoulos, C. & Webb, C. Building a predictive model for PHB production from glycerol. Biochem. Eng. J. 116, 113–121 (2016).

4. Sun, C., Webb, C. & Theodoropoulos, C. Dynamic Metabolic Modelling of Cupriavidus necator DSM 545 in PHB Production from Glycerol. in Computer Aided Chemical Engineering vol. 38 2217–2222 (Elsevier B.V., 2016).