The brain coordinates a continuous coupling between perception and action in the presence of uncertainty and incomplete knowledge about the world. This mapping is enabled by control policies and motor learning can be perceived as the update of such policies on the basis of improving performance given some task objectives. Despite substantial progress in computational sensorimotor control and empirical approaches to motor adaptation, to date it remains unclear how the brain learns motor control policies while updating its internal model of the world. In light of this challenge, we propose here a computational framework, which employs error-based learning and exploits the brain’s inherent link between forward models and feedback control to compute dynamically updated policies. The framework merges optimal feedback control (OFC) policy learning with a steady system identification of task dynamics so as to explain behavior in complex object manipulation tasks. Its formalization encompasses our empirical findings that action is learned and generalised both with regard to a body-based and an object-based frame of reference. Importantly, our approach predicts successfully how the brain makes continuous decisions for the generation of complex trajectories in an experimental paradigm of unfamiliar task conditions. A complementary method proposes an expansion of the motor learning perspective at the level of policy optimisation to the level of policy exploration. It employs computational analysis to reverse engineer and subsequently assess the control process in a whole body manipulation paradigm. Another contribution of this thesis is to associate motor psychophysics and computational motor control to their underlying neural foundation; a link which calls for further advancement in motor neuroscience and can inform our theoretical insight to sensorimotor processes in a context of physiological constraints. To this end, we design, build and test an fMRI-compatible haptic object manipulation system to relate closed-loop motor control studies to neurophysiology. The system is clinically adjusted and employed to host a naturalistic object manipulation paradigm on healthy human subjects and Friedreich’s ataxia patients. We present methodology that elicits neuroimaging correlates of sensorimotor control and learning and extracts longitudinal neurobehavioral markers of disease progression (i.e. neurodegeneration). Our findings enhance the understanding of sensorimotor control and learning mechanisms that underlie complex motor tasks. They furthermore provide a unified methodological platform to bridge the divide between behavior, computation and neural implementation with promising clinical and technological implications (e.g. diagnostics, robotics, BMI).

The project focused on the Ultra Scale-Down (USD) modelling of harvest and early purification for viral vaccine and vector production processes. USD models seek to mimic the commercial scale process environment using only millilitres of material. Many commercial processes include a disc-stack centrifuge, micro-filtration, tangential flow ultrafiltration and density gradient ultracentrifugation steps, which are investigated in this work. A key parameter of USD, in particular for centrifugation, has been shown to be shear, which may cause significant damage to biological products and increase the difficulty of subsequent processing steps. To quantify shear rate therefore the PSC-5 disk-stack centrifuge and PKII density gradient ultra-centrifuge have been modelled using Comsol Multiphysics®. This EngD project extended existing Computational Fluid Dynamic (CFD) analysis by treating the centrifuge as a 3 dimensional system, drawing comparison to the rotating disc shear device based on a simple population balance analysis (accounting for different shear rate levels and time of exposure), and identifying the nature of the shear (simple versus extensional). This modified approach produced a good prediction of shear rate as validated against experimental results, as well as intuitively representing a closer match to the physically important points of comparison between the two systems. In order to overcoming the challenge of experimentally validating CFD predictions in downstream process units a novel approach was adopted: use of synthetic particles that mimic mammalian cells. Adju-Phos was applied successfully for the prediction of shear rate in a number of pilot scale centrifuges of both tubular bowl and diskstack design, by correlating changes in the particle size distribution against that seen in the USD shear device. The significant advantages of using Adju-Phos include reduced cost, pre-work and time. For the PKII density gradient ultracentrifuge, a smaller virus shear mimic, consisting of an off-the-shelf polystyrene nanobead coated with streptavidin, was used for experimental validation of CFD. This work also represents the first steps in developing an USD model for the continuous feed industrial ultracentrifuge, consisting of the rotating disc shear device and laboratory ultracentrifuge matched to the large scale on the basis or  2 t and head at the product layer. The fact that these nanobeads showed an impact from hydrodynamic forces despite their small size, raised the possibility that viruses, which had been thought to be too small for damage by hydrodynamic or other forms of shear, might also be susceptible. There is limited if any literature on the impact of shear rate on cells used to produce viral products, or on the viral products themselves. The existing USD tool was used to provide some insight into the performance of the harvest step for influenza and adenovirus production processes in continuous cell-lines. The results indicated that virus infected cells do not actually show any increase in sensitivity to shear forces, and may indeed become less shear sensitive, in a similar manner to that previously observed in old or dead cell cultures. Clarification may be most significantly dependent on the virus release mechanism, with the budding influenza virus producing a much greater decrease in clarification than the lytic, nonenveloped adenovirus. A good match was also demonstrated to the industrial scale performance in terms of clarification, protein release and impurity profile. The impact on the influenza and adenoviruses themselves was measured using infectivity assays, TEM and particle size distribution and protein contents of post shear purified viral particles. Whilst no impact was recorded based on the infectivity assays or TEM, a small but significant change in particle size distribution was shown for adenovirus, and is indicated for influenza. The influenza virus also showed a possible change in protein make-up of purified viral particles, indicating possibly a loss of those proteins from the virus structure. The significance of these morphological changes, if any, remains to be confirmed.

This doctoral thesis was in collaboration with a research group from the Institute of Ophthalmology at Moorfields Eye Hospital and the London Project to Cure Blindness. The group was in late stage pre-clinical development creating a pluripotent stem cell (PSC) derived therapy to produce retinal pigment epithelium cells (RPE) for the treatment of Age Related Macular Degeneration (AMD). Creating effective treatments for AMD is vital given that it is the leading cause of blindness in the developed world with around 500,000 people in the UK afflicted with the disease, half of which are registered as visually impaired. In addition, for the dry form of the disease, which represents 85-90% of total cases, there is no effective treatment. The current lab based manufacturing protocols intended for use during Phase 1 trials are not feasible for subsequent larger scale clinical trials or for the commercial manufacture of an affordable mass-produced therapy. The aim of this thesis was to explore the feasibility of applying methodologies and technologies from traditional biotechnology production to cell therapy manufacturing. Specifically use of bioreactors to scale up cell culture (Chapter 2), the application of online monitoring and control to PSC culture (Chapter 3), the use of a design of experiments (DoE) approach to PSC differentiation (Chapter 4) and cell-cell purification (Chapter 5). Taken as a whole it is clear that the application of traditional biomanufacturing technologies and approaches have much potential when applied to cell therapies, and specifically a PSC derived cell therapy to treat blindness. However much more work is needed to reduce variability in the process to better understand the impact of process parameters on critical quality attributes and how best to approach these in an environment where there is little clinical experience to define the target product profile.

Adsorption isotherms in chromatography are critical in determining the separation of solutes during column separations. Multicomponent protein adsorption isotherms, which are relevant during the downstream processing of biopharmaceutical products, have received limited study historically. The studies and methodologies which have been assessed have mainly focused on small, simple, chromophore containing proteins which have limited applicability to industrially relevant bio-therapeutics. The reasons why this area of study has received limited attention include the experimental effort associated with generating such large data sets as well as the difficulty in obtaining data of good enough quality. The work explored here presents and optimises the deployment of highthroughput chromatography formats as well as automated liquid handling systems in order to elucidate adsorption isotherms of proteins. Additionally, alternative rapid analytical methods involving the collection of protein UV spectra in conjunction with multivariate data analysis have been applied to quantify protein mixtures. These rapid high-throughput methods decrease the experimental effort associated with multicomponent isotherm studies. 3 binary isotherms and 1 ternary isotherm have been studied for larger, non-chromophore containing model proteins. The propagation of error in single component and multicomponent isotherms has been investigated to understand what drives the propensity for error as well as methods to mitigate problematic regions of investigation. The fitting of the multicomponent ion exchange isotherms across multiple salt levels to isotherm formalisms proved elusive which precluded their application for in silico modelling of column separation. Short of that a heuristic optimisation of a binary mixture was achieved quantifying eluted fractions using the UV spectra multivariate method.

The increased use of antibodies for human therapy has driven rational approaches to accelerate bioprocess development in producing cost effective and highly productive antibodies. The potential of microwell based systems and miniature bioreactors (MBR) to mimic the scalability and operations of conventional bench reactors are seen as an alternative. This study has investigated the microtitre plate (MTP), microMatrix and MBR (HEL-BioXplore) as scale-down mimic for rapid and accurate reproduction of Chinese hamster ovary (CHO) cell growth and product yields in bench scale stirred tank reactors. A microtitre plate with sandwich lid CR1524a (for slow growing animal cells) was found to be suitable for CHO cell cultivation. An evaluation of feeding approaches in MTP showed that bolus addition resulted in 9.19 x 106 cell mL-1 and 38 % higher IgG titres compared to addition of FeedBeads. In order to enable scale translation, the engineering parameters for the MBR were characterised with regard to mixing time, volumetric oxygen transfer coefficient and power input. The MBR system was fitted with either direct driven impeller or magnetically driven impeller with singular hole impeller or horseshoe sparger. The combination of the direct driven impeller and horseshoe type sparger with bolus addition was selected as the best configuration and produced 8.89 x 106 cell mL-1 and 0.84 gL-1 IgG titres. Additionally, a prototype micro-Matrix system was characterised for its performance in a cell culture process. The micro-Matrix with controlled aeration and continuous feeding supported a cell concentration of 8.67 x 106 cell mL-1 and viability >90 % after 264 hours. Furthermore, scale translations of the studied systems were evaluated at the matched mixing time of 6 s with conventional lab scale 5L stirred tank reactors (STR). The scale-up studies demonstrated that the miniature systems were able to mimic the performance of the conventional bench reactors. Results from the scale-up studies between the MTP, MBR and STR with bolus feeding addition showed a comparable viable cell concentration of 9.30 x 106 cell mL-1 , 9.56 x 106 cell mL-1 and 10.04 x 106 cell mL-1 and IgG titres of 0.92, 0.69 and 0.83 gL-1 respectively. Whereas, scale translation studies between micro-Matrix and MBR with continuous feeding gave equivalent viable cell concentration with 11.1 x 106 cell mL-1 and 9.76 x 106 cell mL-1 and IgG titres of 0.50 gL-1 and 0.64 gL-1 respectively. Overall, the miniature bioreactors evaluated have the potential for cell screening and optimisation studies which could generate early data for bioprocess development.

Affinity chromatography is an indispensable method of protein purification used during the manufacture of many therapeutic antibodies. The protein A ligand is a popular choice for selective purification of immunoglobulin G (IgG) molecules. Aggregated product is often found in protein A elution pools; this is generally attributed to the effect of low pH (elution) on protein structure. However, there is evidence that other facets of the chromatography process influence aggregation phenomena. Physical and chemical transitions, such as concentration of protein on the adsorbent surface and change in buffer composition, may challenge the structural integrity of proteins, increasing their propensity for aggregation. The influence of elution from protein A on the aggregation rate of an IgG4 during a subsequent low pH hold was investigated. IgG4 was incubated in elution buffer after protein A chromatography and the monomer concentration in neutralised samples was quantified by size exclusion chromatography. Rate constants for monomer decay over time were determined by fitting exponential decay functions to the data. Similar low pH experiments were implemented in the absence of a chromatography step. The IgG4 demonstrated highly pH-dependent and apparently concentration-independent aggregation behaviour. The findings suggested that aggregation was governed predominantly by a pH-dependent unfolding/ re-folding equilibrium. Elution from protein A was found to increase aggregation rates by half an order of magnitude, while other aspects of the aggregation kinetics appeared un-affected. In order to advance understanding of how adsorption processes might impact antibody stability, neutron reflectivity was used to characterise the structure of adsorbed IgG on model surfaces. In the first model system IgG was adsorbed directly to silica and demonstrated a side-on orientation with high surface contact. A maximum dimension of 60Å in the surface normal direction and high density surface coverage were observed under acidic pH conditions. In the second model system protein A was attached to a silica surface to produce a configuration representative of a glass chromatography resin. Interfacial structure was probed during sequential stages from ligand attachment, through IgG binding and elution. Adsorbed IgG structures extended up to 230Å away from the surface and showed dependence on surface blocking strategies. The data was suggestive of two IgG molecules bound to protein A with a somewhat skewed orientation and close proximity to the silica surface. The findings provide insight into the orientation of adsorbed antibody structures under conditions encountered during chromatographic separations. The outcomes of this work provide a broad scope for future investigations. Based on preliminary work, using neutron measurement techniques to monitor aggregate formation inside glass chromatography resins is suggested as an interesting direction.

Biorefineries allow for the sustainable production of higher value products from biomass. In addition to bioethanol, they can produce added value chemicals and pharmaceutical intermediates from isolated component compounds such as sugars. Sugar beet pulp (SBP) is a high volume, low value by-product from sugar beet processing with a low lignin and a high carbohydrate content, making it an attractive biomass feedstock for biorefinery processing. The pectin fraction of SBP can be isolated via steam explosion, which, after complete acid hydrolysis, gives a hydrolysate rich in monosaccharides: primarily L-arabinose (Ara) and D-galacturonic acid (GA), with some D-galactose (Gal) and L-rhamnose (Rha). Isolation of these sugars is therefore a critical step in realising an integrated, whole crop biorefinery. Currently, little work has been reported on the separation and utilisation of SBP hydrolysates. The aim of this thesis is to establish novel, scalable separation processes for the isolation of the component monosaccharides from crude hydrolysed sugar beet pulp pectin. Centrifugal partition chromatography (CPC) is a liquid-liquid separation technique with no solid stationary phase and offers an alternative to traditional resin-based chromatographic techniques. As such it can more easily cope with crude feedstreams such as hydrolysates. Hydrophilic ethanol : ammonium sulphate two-phase systems were examined based on monosaccharide partition coefficients and phase settling times. An ethanol : aqueous ammonium sulphate (300 g L-1 ) (0.8:1.8 v:v) system was chosen for CPC separations of the crude SBP hydrolysate and was shown to be capable of removing the coloured contaminants and isolating three sugar fractions in a single step: Rha, Ara and Gal, and GA. The separation was optimised and the throughput was increased by maximising the sample loading. Operation in an elution-extrusion mode allowed for reproducible separations in 100 min without additional column regeneration. The process was scaled up from a 250 to a 950 mL column providing a final throughput of 1.9 gmonosaccharides L -1 column h -1 using the crude SBP. The following purities and recoveries of the three main fractions were achieved: Rha at 92% purityand 93% recovery; Ara at 84% purity and 97% recovery; and GA at 96% purity and 95% recovery. Simulated moving bed (SMB) allows for continuous chromatographic separations using multiple columns, improving separation performance and throughputs. Isolation of Ara from the neutral sugars Gal and Rha was performed with resins and conditions screened on single columns leading to the selection of a Dowex 50W X8 resin in the Ca2+ form. SMB separation using 8 columns was performed in the 4-zone and 3-zone setups and achieved 94% purity with 99% recovery at a throughput of 4.6 gmonosaccharides L -1 column h -1 with a synthetic mixture of the neutral sugars (Ara, Gal and Rha). However, equivalent separations could not be achieved using the crude SBP hydrolysate which needed pretreatment before SMB. Decolourisation with activated carbon was able to remove 97% of the coloured contaminants with sugar losses of 15% (w/w) in a batch process demonstrated to 50 mL scale. Anion exchange chromatography using a Dowex 1x8 resin was then found to be capable of isolating GA from a synthetic crude mixture of GA and neutral sugars with a dynamic binding capacity of 1.31 mmol mL-1 resin. However, further work is needed to enable this anion exchange step to achieve satisfactory separations with the decolourised crude hydrolysate. The isolated neutral sugars, after GA removal, can be processed on the SMB with comparable separation performance and throughput to a mixture of neutral sugars prepared without GA. In summary, this thesis presents two possible process paths each with their own benefits and drawbacks. CPC is capable of processing the crude SBP hydrolysate directly, isolating the sugars and removing the coloured contaminants in a single step. However, Ara co-elutes with Gal providing a stream that is only 84% pure. In SMB, the potential throughputs and separation performance are higher, however, this could only be experimentally demonstrated with synthetic crude mixtures of sugars and not with the crude SBP hydrolysate. Further pretreatment or SMB method development would be required in order to process the crude hydrolysate, and the resulting multistep processes may reduce the overall viability. Overall this thesis demonstrates two feasible approaches to the preparative scale separation of SBP pectin hydrolysates and supports development of an integrated SBP biorefinery.

The thesis describes an experimental investigation of the flow, mixing and suspension dynamics in cylindrical orbitally shaken bioreactors (OSRs). Amongst the plethora of bioreactor types and geometries available for cell culture, the OSR is ubiquitous in bioprocess research and development. Offering a well defined liquid-gas interface, high throughput potential and experimental flexibility, it is the vessel of choice in early bioprocess research, either as microtiter-plates, Erlenmeyer flasks or other geometries. Despite recent advances in the field, an accurate and exhaustive engineering characterisation of OSRs from this point of view is lacking. In the present study a mixing time estimation methodology is developed and employed to assess the effect of operational parameters on the mixing of cylindrical OSRs. Particle Image Velocimetry measurements are carried out to evaluate the effect of vessel geometry modifications on the flow. Laser Induced Fluorescence is used to produce accurate description of the micro-mixing. Solid suspension studies are also undertaken to assess potential strategies to improve microcarrier culture in OSRs. Accurate determination of the mixing time in OSRs is essential for the optimisation of mixing processes and minimization of concentration gradients that can be deleterious to cell cultures. The Dual Indicator System for Mixing Time (DISMT) is employed, together with a purposely built image processing code to objectively measure mixing times in cylindrical and Erlenmeyer flask bioreactors. Relevant data acquisition aspects to optimise the accuracy of DISMT measurements are discussed in detail, with direct comparison of different mixing time measurement methodologies, including DISMT, pH probe and iodine thiosulfate decolourisation results obtained in two types of stirred reactors. The DISMT is employed to determine mixing characteristics of OSRs at different flow conditions and develop an effective feeding strategy, by evaluating the effect of the position of the feed at different radial locations in the vessel. At low Fr the flow presents a toroidal vortex below the free surface, which controls mass transport process across the entire vessel and defines two distinct regions in and outside of the vortex exhibiting different mixing rate. At higher Fr the axial flow enhances the mixing of the fluid located next to the wall as the mean flow transition to axial flow coincides with a regime flow transition and onset of turbulence fluctuations. By controlling the locations of eed addition, the flow characteristics can be exploited to enhance initial distribution of the added liquid and decrease the time required to reach homogeneity. The mixing number is highly dependent on the position of the feeding pipe. Insertion close to the vessel walls, and in the periphery of the toroidal vortex, where local shear stresses and deformation rates are highest, were found to significantly enhance mixing. In order to provide an effective scaling methodology, the results obtained in OSRs are compared with data previously reported in the literature for both cylindrical reactors and Erlenmeyer flasks. The employment of a critical Froude number shows promise for the establishment of a scaling law for mixing time across various types and sizes of shaken bioreactors. The flow dynamics in cylindrical shaken bioreactors of different conical bottom geometries (inward facing) is investigated by means of phase-resolved Particle Image Velocimetry. The cylindrical bioreactor with a conical bottom geometry is selected to assess its potential application in three-dimensional cultures, and improve solid suspension in shaken systems. The effects of conical shaped bottoms of different heights on the fluid flow are evaluated for different operating conditions with water being the working fluid. The results provide evidence that the presence of the conical bottom affects the transition from laminar to turbulent flow, increases the vorticity and generates shear stresses at well defined locations. The increased kinetic energy content measured with PIV in cylindrical OSRs with a conical bottom is found to effectively enhance solid suspension in microcarrier or embryod body cell culture. The dynamics of solid suspension is studied using commercially available Cytodex-3, stained with trypan blue for improved visual contrast and image acquisition. The presence of the conical bottom improves solid suspension by requiring lower agitation rates for the microcarriers to lift from the bottom completely. The critical Froude, which determines the flow type controlling the bioreactor, can also be used to scale the suspension of microcarriers in OSRs. Full characterisation of macro- and micro- mixing scales in OSRs for highly viscous is obtained by DIMST and pLIF, respectively. This data also provides an effective visualisation of the flow structures controlling the bioreactor transport processes. The mixing characteristics of high viscosity fluids in OSRs are investigated by means of DISMT and pLIF, for the macro- and microscales of mixing, respectively. Fluids of viscosities 2-14 times that of water, exhibit flow characteristics different to those observed at ν=10−6 m^2s^−1. A toroidal vortex, similar to that observed in water, is present at low Fr. Small increments of agitation rate provide transition to other flow structures, never reported in the literature. The pLIF measurements allow to characterise the small scale features of the flow not observable from phase averaged PIV, and visualise well defined elongation and striation dynamics for different regimes. Although extensively used, OSRs are yet to be fully characterised. Further research is required on the hydrodynamic phenomena dominating orbitally shaking vessels, to enable the development of scale-up platform to simplify and speed progress from cell line development to industrial production of bio-products. The development of the scaleup platform must be made considering the advantages and requirements of single-use technology, to provide the industry with robust and reproducible scale-up model.

Recombinant human Acid Alpha Glucosidase (GAA) is the therapeutic enzyme used for the treatment of Pompe disease, a rare genetic disorder characterised by GAA deficiency in the cell lysosomes. The manufacturing process for GAA can be challenging, in part due to protease degradation. The overall goal of this project was to understand the effects of GAA overexpression on cell lysosomal phenotype and host cell protein (HCP) release, and any resultant consequences for protease levels and ease of manufacture. To do this we first generated a human recombinant GAA producing stable CHO clonal cell line and then developed a two-step bioprocess based on capture chromatographic step anion exchange (IEX) and intermediate hydrophobic interaction (HIC). The purity of GAA after HIC was determined via LC/MSMS to be above 80%. We then collected images of cell lysosomes via transmission electron microscopy (TEM) and compared the resulting data with that from a Null CHO cell line. TEM imaging revealed 72% of all lysosomes in the GAA cell line were engorged indicating extensive cell stress; by comparison, only 8% of lysosomes in the Null CHO had a similar phenotype. Furthermore, comparison of the HCP profile among cell lines [GAA, mAb and Null] capture eluates, showed that while most HCPs released were common across them, some were unique to the GAA producer, implying that cell stress caused by overexpression of GAA has a molecule specific effect on HCP release. Protease analysis via zymograms showed an overall reduction in proteolytic activity after the capture step but also revealed the presence of co-eluting proteases at approximately 80 KDa, which MS analysis putatively identified as dipeptidyl peptidase 3 and prolyl endopeptidase.

Antimicrobial resistance is a challenge facing the world in the twenty-first century with an estimated 10 million deaths by 2050 if no actions are taken. Microbial resistance to drugs is a natural consequence when bacteria develop and adapt genetically to face new challenges including antibiotics. Currently, this development occurs at a higher rate than drug discovery. Hence there is a need for a new generation of antibiotics that kill pathogenic bacteria. Nature itself provides inspiration for such new antibiotics. For example, our immune system secretes antimicrobial peptides (AMPs), which have been successful agents in killing pathogens with no reported bacterial resistance. Compared with conventional antibiotics, these peptides are larger and more sophisticated biological molecules, which disturb the bacterial membrane, leading to cell lysis. It is currently costly to extract AMPs from natural resources to be used for fighting infections. Alternatively, synthetic AMPs that mimic natural ones could provide a sustainable cheap weapon against such thread. This also provides a unique opportunity to understand the structure–function relationships of such molecules to optimise these effective, non-toxic antimicrobial properties. Our collaborators at National Physical Laboratory have designed and synthesised new AMPs from their essential building blocks (amino acids). This thesis describes the use of atomic force microscopy (AFM) as a nanoscale imaging technique for characterising and imaging membrane poration mechanisms of four new AMP systems. Two of these systems are helical peptides, explained in chapter 3. The third system, explained chapter 4, is a triskelion with three arms of antimicrobial β-sheet peptide that co-assemble to form a hollow antimicrobial capsules. The latter has two possible functions: gene delivery and bactericidal effects. The fourth system, explained in chapter 5, contains two peptide monomers that are designed to co-assemble and form antimicrobial hollow capsids, inspired by the natural viral capsids. Finally, chapter 6 is a plan for taking these AMPs a step closer to commercialisation, including a business plan for one potential application.