Chemical and Bioprocess Engineering Theses

Monoclonal antibodies (mAbs) are the most dominant selling class of biotherapeutics in the global market. These complex biomolecules are produced through mammalian cell culture and are prone to structural heterogeniety. This heterogeniety can have an adverse effect on the overall stability and efficacy of the drug product and must be closely monitored through product characterisation. Characterisation of mAbs can be carried out on the intact, subunit and peptide level. The drawback with traditional characterisation techniques include high sample requirement, high level of expertise needed to produce reporducible data and the methods involved are usually time consuming. The objective of this work was develop characterisation strategies at all three levels using novel instrumentation to overcome these drawbacks and improve our understanding of product characteristics at all stages of the mAb production cycle.

In this thesis, various systems containing interfaces of titanium dioxide (TiO2) and haematite (a-Fe2O3) with water are examined using a number of atomistic simulation methodologies. These systems include large-scale anatase (101) and rutile (110) surface slabs modelled using force-field molecular dynamics (MD); smaller haematite (001) and rutile (110) surface slabs modelled using density functional theory MD; and a large scale anatase nanoparticle and smaller anatase (101) and rutile (110) surface slabs, modelled using density functional tight-binding MD. As part of these studies a variety of analyses are presented, aimed at providing a quantitative understanding of the effects that each surface or nanoparticle has on the properties of water molecules near the interface; and thereby assessing, in a qualitative way, how these effects are manifested using the different methodologies. These analyses include established techniques in the field of atomistic simulations, such as hydrogen bond analysis and electronic density of states calculations. Also employed are techniques novel to the field of atomistic simulation, such as the coherence spectrum. Two points of emphasis are present throughout this thesis: firstly, to improve the understanding of the materials examined towards the development of photoelectrochemical catalysts; and secondly, to explore the current state-of-the-art in atomistic simulations, and "push the boundaries" of the available techniques.

Continuous crystallization has been identified as a means of reducing costs and improving product consistency in the pharmaceutical industry. Withdrawal classification has been identified as a complex topic with broad relevance to the use of continuous stirred-tank crystallizers in the industry and in the laboratory.This thesis begins by summarising recent research into various forms of continuous crystallization, contextualising the main body of work and identifying relevant gaps in the literature. Withdrawal classification constitutes one such gap, with little research data available. The review is used as justification for the research following.Preliminary experiments are detailed, the purpose of which was to compare and assess operational strategies. This work also confirms that controllable classified withdrawal is possible with the experimental setup decided upon. The findings were applied to the main body of work following and used to optimise experimental methods.Two main sets of experiments are detailed in the principal experimental portion of the thesis. The first focuses on withdrawal velocity as a means of directly controlling classified behaviour. This was found to be a significant predictor of certain key system characteristics. The second experimental set focuses on the influence of mixing, by changing agitation rate and withdrawal location. These were also found to significantly affect classification behaviour.Finally, the experimental data was combined to produce a set of empirical equations which allow for qualitative modelling of the system.Two separate approaches to modelling classification behaviour were developed. Experimental data was used to fit kinetics to the models, allowing for simple simulation and comparison. A method for meaningful comparison of model outputs was developed and simulated results compared.Finally, more sophisticated simulation was performed by combining the models with the empirical data gathered. It was demonstrated that particle size data can be predicted based only on system settings and characteristics using this method. Crystallizer responses to independent variables were assessed. The work concludes with a discussion of the possible mechanisms of classification, the applicability of the different models to these, and the relative strengths and weaknesses of the models.

The dissolution processes of active pharmaceutical ingredient (API) crystals have been extensively studied in the pharmaceutical industry, as they have a significant impact upon the bioavailability of drugs within the body. Although much experimental work has been conducted and many models have been established, none of the models comprehensively describe the dissolution process of the API. Moreover, the impact of physiochemical properties on the dissolution of the API is not well understood. A deeper understanding of dissolution behaviour and of the factors affecting the dissolution process is critical to the design, evaluation, control and therapeutic efficacy of solid dosage forms. Hence in this thesis, the dissolution processes of API crystals were investigated using both experimental and modelling approaches. A BCS Class II drug, ibuprofen, characterized by a relatively low solubility but high permeability compared to other drugs was used as a model compound to investigate dissolution kinetics. The effect of physicochemical properties of the particles on the dissolution kinetics was also investigated. Firstly, an ibuprofen preparation protocol for dissolution was established. Different ibuprofen crystals with tailored solid-state characteristics (such as crystal morphology and size distribution) were prepared by cooling crystallization. The crystallization process was monitored in situ by process analytical techniques (PATs) such as FBRM, PVM and ATR-FTIR. The properties of the obtained crystalline products were also characterized by off-line techniques such as high performance liquid chromatography (HPLC), microscope, scanning electron microscope (SEM), powder X-ray diffraction (PXRD), differential scanning calorimetry (DSC), and Malvern Mastersizer, etc. A laboratory methodology for the dissolution processes of the obtained IBU crystals was then developed. UV and ATR-FTIR spectroscopic techniques were employed to measure the solute concentration and a FBRM probe was applied to track the change of the particle size and population profile during dissolution. The influence of the initial undersaturation ratio, agitation rate, crystal morphology and particle size on the dissolution were examined. Variations in the dissolution rate were observed, from which two distinct dissolution mechanisms during the dissolution process were proposed. Eight mathematical models which have been typically employed to quantify the dissolution of immediate and modified release dosage forms, including the zero-order, first-order kinetics, Weibull, Higuchi, Hixson-Crowell, Korsmeyer-Peppas, Baker-Lonsdale and Hopfenberg models, were used to correlate the dissolution profiles of ibuprofen crystals. The dissolution parameters of each model were determined and the simulation accuracies of the different models were evaluated by comparing simulated and experimental results. It was found that the Weibull, Korsmeyer-Peppas and first-order kinetics models provided the most accurate results, suggesting that these models may be successfully applied to the dissolution of API powders in both API processing and drug product performance analysis. A diffusion-based model which can be used to predict non-sink circumstances was next devised to study the dissolution kinetics of ibuprofen. Variations of the model were created to compare the accuracy of simulation results when applied to spherical and cylindrical particle geometries, with and without the inclusion of a size-dependent diffusion layer thickness component in the model. Experimental data was fitted to the model to obtain the diffusion layer thickness and post-dissolution particle size distribution predictions. The comparison between simulated and experimental results demonstrated that both size-dependent and size-independent models can give good simulation results.

Understanding effective energy-conversion systems and dealing with the problem of intermittency through scalable energy-storage systems are the two major difficulties in renewable energy. At the Grid size, relatively little progress has been done, and two considerable issues remain: (i) minimizing environmental harm, and (ii) the issue of ecologically friendly energy conversion. Light-driven photoelectrochemical (PEC) water-splitting can create hydrogen, but it is inefficient; instead, we focus on how electric fields can be applied to metal-oxide/water systems to adjust the interplay with their intrinsic electric fields, and how this can change and increase PEC activity, drawing both on experiment and non-equilibrium molecular simulation. Non-equilibrium molecular-dynamics simulations of liquid water were carried out in the canonical ensemble in the presence of both external static and oscillating electric fields of(r.m.s.) intensities 0.05 V/Å and 0.10 V/Å, with oscillating-field frequencies 50, 100 and 200 GHz. The rigid potential model TIP4P/2005 was used, and NEMD simulations were performed, including in the supercooled region, at temperatures ranging from 200 to 310 K. Significant changes in the percentage dipole alignment and self-diffusion constant were found vis-à-vis zero-field conditions, as well as shifting of the probability distribution of individual molecular self- diffusivities. The application of static fields was typically found to reduce the self-diffusion of liquid water, effectively due to some extent of "dipole-locking", or suppression of rotational motion, whereas diffusivity was found to be enhanced in oscillating fields, especially at high frequencies and outside the supercooled region. Classical molecular-dynamics techniques were used to evaluate the distribution of individual water molecules’ self-diffusivities in adsorbed layers at TiO2 surfaces anatase (101) and rutile (110) at 300 K for inner and outer adsorbed layers. Using local order parameters, the layered-water structure was identified and classed in layers, which proved to be an equally viable way of "self-ordering" molecules in layers. Anatase and rutile differed significantly in disrupting these molecular distributions, particularly in the adsorbed outer layer. Anatase (101) had much greater self-diffusivity values, owing to its "corrugated" structure, which allows for increased hydrogen bonding interaction with adsorbed molecules beyond the initial hydration layer. On the contrary, rutile (110) has more securely "trapped" water molecules in the region between Ob atoms, resulting in less mobile adsorbed layers. Finally, the dynamical properties of physically and chemically adsorbed water molecules on pristine hematite-(001) surfaces were investigated using non-equilibrium ab-initio molecular dynamics (NE -AIMD) in the NV T ensemble at room temperature, in the presence of externally applied, uniform static electric fields of increasing intensity. Significant changes in the dipole moment and self-diffusion constant were observed in comparison to zero-field circumstances, as well as a shift in the probability distribution of individual molecule self-diffusivities. For example, static fields were shown to promote the self-diffusion of water molecules at the a-Fe2O3 surface, owing to some degree of ’dipole-locking’ in the applied direction of the field.

Many countries are working on transitioning towards a carbon-neutral economy by 2050. This has predominantly driven investments towards renewable electricity technologies such as solar and wind generators, driving research to increase energy systems' interconnection. To investigate the role of gas-based vectors in the transition of the energy system, the following sectors are considered, i.e., gas for energy storage, gas for heating, and gas for power generation. Maximizing green hydrogen production in Power to Gas (P2G) systems by understanding 1) stack performance of low and high-temperature electrolyzers and 2) operational modes in response to curtailed renewable electricity could aid potential investors to achieve maximum return on their investment. Progressively increased intermittent operational modes in electrolyzers showed that low-temperature electrolyzers were more resilient to flexible operation (< 7.7% difference in total H2 production) when compared to high-temperature electrolyzers (< 67%). These results indicated that a stack-level understanding of electrolyzers in the integration of the energy system, needs to be incorporated to maximize green hydrogen production. Utilizing the present gas infrastructure for flexible power generation is foreseen as a primary role for natural gas/hydrogen in a low carbon future. A least-cost optimization for systems operational cost in a one-node multi-renewable electricity systems model increased operational costs when large coal generators were progressively replaced with smaller gas generators. The role of reciprocating engines and industrial gas turbines was primarily in providing fast frequency response and reserve regulation. Another pathway for natural gas and hydrogen in a low-carbon energy system is in the heating sector. A detailed investigation of potential decarbonization achievable in Irish building stock that still uses traditional gas boilers for domestic and space heating was analyzed. The study concluded that thermal retrofits and blending of 20% H2 in the gas network alone would contribute to a maximum decarbonization of 34% in Ireland’s heating sector. The studies presented in the thesis provide a novel approach to understanding the use of P2G systems, analyzing the role of gas for heating and flexible gas generators in a future low carbon energy system.

Photoelectrochemical (PEC) water splitting cells, used to create hydrogen from solar energy, are crucial to the implementation of solar-to-fuel technology. Research in science and technology is focused on improving the functionality and efficiency of these devices while also ensuring that they will last for a long time and will be affordable. Machine-learning based computer-aided simulation is one of the methods exploited to design the next generation of photoelectrochemical cells. Such simulation techniques have filled the gap between conventional force fields, which are fast and inaccurate, and electronic structure simulations, which are slow and accurate. In this thesis, we developed and/or improved machine-learning interatomic potentials (MLPs) for all the chemical systems involved in the design of PEC cells. Having established the reliability of such potentials in describing the many-body share of energy for argon clusters, we then utilize neural networks as a powerful machine-learning method to construct potential energy surfaces for the other studied systems. High-dimensional neural-network (HDNN) model developed by Behler et.al (Behler, 2007) was used to obtain the coordinates of a chemical system as input and output its total energy. We used atom-centered symmetry functions (Behler, 2011) to encode the chemical environment into a fixed-length input vector. As part of this thesis, various sampling schemes are examined and improved to construct the training set needed for model development, as well as a new technique to manipulate publicly available data for model development is implemented using machine-learning. In addition, the network's weights and biases are adjusted using different optimization techniques. We also implemented a molecular dynamics code that includes long-range electrostatic energy via Ewald sums and delta-machine-learning, and we proposed and developed a technique based on checkpoint ensembles in machine-learning to improve the accuracy of a neural network model for total energy prediction.

Bacterial adhesion and the subsequent biofilm formation is a complex phenomenon which has many consequences in water filtration. This aggregation of microorganisms can be difficult to remove from nanofiltration and reverse osmosis membrane surfaces, causing damage and eventual replacement of the membrane. In order to elucidate the cause of this biofilm formation, three influential factors were studied: surface topography, nutrient concentration and shear stress. Analysis was performed on the surface topographical heterogeneities in order to examine the influence of surface topography. Image analysis of the adhesion of Pseudomonas fluorescens (Ps. fluorescens) and Staphylococcus epidermidis (S. epidermidis) to the surface topographical heterogeneities was determined for two commercial membranes, NF270 and BW30, using a flow-cell system. Membrane area analysis, using AFM and SEM, showed up to 13% of topographical heterogeneities on the membrane surface with up to 30% of total adhered cells that were discovered within these topographical heterogeneities. For the analysis of the nutrient availability and shear stress on the structural formation of Ps. fluorescens biofilm under two different dynamic conditions, an air-liquid interface biofilm and a flow cell grown biofilm were assessed by confocal scanning laser microscopy (CLSM). The analysis showed a three-fold increase in the EPS biovolume of the high nutrient air-liquid interface grown biofilm. However, the flow cell biofilm increased the biovolume for low nutrient and higher shear stress conditions, suggesting harsher growth conditions of the biofilm results in greater biofilm development. Finally, the adhesive and viscoelastic properties of the Ps. fluorescens air-interface grown biofilm for two different nutrient dilution factors was determined by nanoindentation. The low nutrient availability showed higher adhesion force and work of adhesion with distributed colonies across the surface, while the high nutrient grown biofilm led to a reduction in the adhesive and elastic nature of the biofilm.

Chinese hamster ovary (CHO) cells are the dominant mammalian expression host for recombinant therapeutic protein production. In terms of manufacturing efficiency, much has been accomplished in areas such as optimised transgene design and cell line development. Since the publication of the Chinese hamster genome the field has gained a more refined understanding of the relationship between the CHO biological system and desirable bioprocess traits. Despite the central importance of protein synthesis, few studies to date have focussed on characterising translation in CHO cells. The goal of this thesis is to evaluate the utility of ribosome footprint profiling (Ribo-seq) to further improve our understanding of CHO cell biology and highlight routes towards enhanced biopharmaceutical manufacturing. A key aspect of this work is the combination of multiple translation inhibitors for Ribo-seq to enable the simultaneous analysis of translation initiation and elongation for the first time. The availability of these data enabled the identification of previously uncharacterised open reading frames (ORFs) including those non-AUG start codons. Novel ORFs comprised of N-terminal extensions of canonical proteins, ORFs found in genes previously thought to be non-coding and those found in the 5’ leader sequence of mRNAs (i.e. upstream ORFs). Through the use of Ribo-seq and RNA-seq data, these upstream ORFs were found to have a repressive effect on the translation efficiency of the main ORF. In addition, following comparison of CHO cells at day 4 and day 7 of cell culture as well upon a reduction of cell culture temperature, genes undergoing differential translation were identified. A number of these genes did not have a corresponding change in gene expression, confirming that Ribo-seq can provide an additional dimension compared to using RNA-seq in isolation. Ribosome profiling has further enabled the computation of transcriptome wide decoding times for each codon, and revealed influence of codon context on translational rate. These data provide a potential route towards more efficient codon optimised transgene sequences. Perhaps the most striking finding of this work is the identification of thousands of novel small open reading frames (sORFs) predicted to encode microproteins (i.e. proteins < 100aa). Host cell protein analysis, revealed that 8 microproteins were present in adalimumab, confirming that microproteins are a novel class of potential process related impurity. In summary, ribosome footprint profiling is a powerful analytical method for improving the annotation of the CHO cell genome, understanding CHO cell biology and identification of routes to improve not only the upstream process but also enhance the characterisation of the final drug product.

The viability of mammalian cells is primarily tested by dye exclusion assays to examine the integrity of the outer membrane. Precursor events to the onset of cell death are detectable using a combination of online and offline technologies. This work explores the use of dielectric spectroscopy and impedance flow cytometry to characterize changes in the biophysical properties of cells as they progress through batch cultures. At-line single cell imaging was examined in tandem with these methods to prove further insight into the identification of morphological changes in the cell culture. This information was collated to better understand at what point cells can no longer be classified as recoverable prior to the loss in membrane integrity. Autophagic activity such as the increased presence of lysosomes was identified using digital holographic imaging. An earlier decline in the online capacitance signal relative to offline counts occurred in tandem with the onset of autophagy due the shifting dynamics of the cell population. Schwan modelling gave insight on the changes in the bulk membrane capacitance and intracellular conductivity of the cells during this period. Single cell impedance measurements were used to examine the population dynamics with greater accuracy. Opacity and phase parameters were derived at suitable frequencies and compared to the online models. Multifrequency data from the capacitance probe proved useful in the identification of apoptotic activity which followed autophagy. The Cole-Cole a and critical frequency of the changing ß-dispersion curve properties were examined relative to these starvation events. A feeding strategy was employed to delay the onset of autophagy in batch cultures, through the introduction of amino acids. Controlled refeeding experiments were shown to affect both the presence of lysosomes and shifts in opacity trends, suggesting that cells could be recovered during autophagy. The effects of such a feed on the online modelling data was examined to see if a real time parameter from the multifrequency trends could be used as an indicator for culture refeeding.

The volatility of renewable energies poses challenges to power system reliability and calls for more flexible electricity resources, both on the supply and the demand side. Energy-intensive water services such as wastewater treatment offer great demand flexibility potential in that regard. However, current demand response modelling approaches are insufficient for assessing this potential accurately. This study aims to fill the knowledge gap in industrial demand response modelling by introducing an integrated energy-water system model, which takes into account the constraints of the wastewater treatment process on power system scheduling in a joint system dispatch problem. The model is applied to a case study of the Irish wastewater treatment sector and power system. The objective of this study is to identify the benefits of energy demand and supply flexibility of wastewater treatment plants for power system operation, wastewater treatment operators and electricity consumers. The findings indicate that the wastewater treatment sector can be a valuable demand response resource for the power system. Wastewater treatment operators, electricity consumers and power system operators benefit from more flexible electricity demand from wastewater treatment plants, even in the presence of other flexibility measures in the system. Furthermore, it decreases the carbon intensity of domestic power generation. There is also a benefit for the power system operator in harnessing the flexibility of demand response and biogas production simultaneously. However, this can result in temporarily high electricity prices in the model, leading to increased electricity costs for consumer and wastewater treatment plants. Two main conclusions can be drawn from the findings of this study. First, wastewater treatment plants have untapped potential for demand response and utilising it for power system flexibility benefits wastewater treatment operators, electricity consumers and power system operators. The results inform policy makers on how to evaluate and support the electricity demand and supply flexibility of wastewater treatment plants. Given the benefits and minimal capital costs, policy makers should incentivise WWTP operators to tap into this readily available flexibility potential. Further, policy makers should carefully select the appropriate support schemes. In particular, smart demand response schemes should take into account possible interactions with electricity supply flexibility from biogas generation. Second, including wastewater treatment constraints in the system dispatch problem is crucial in order to estimate the flexibility potential accurately and uncover bottlenecks, which would probably be concealed by a black-box approach. Thus, this study provides a valuable case study for investigating the demand response potential of highly complex industrial processes, such as wastewater treatment.