Biopharmaceutical drug development is risky, lengthy, and expensive. Decisions in this delicate process are complicated by constraints on resources such as available capacity and uncertainties that include the risk of clinical failure. Hence, the impact of making sub-optimal decisions in this environment can be severe. Accordingly, this work explores the development of algorithms to support strategic drug development decisions and contains four results sections. Firstly, a decision-support framework based on multi-criteria decision making (MCDM) is presented for assessing options when acquiring biopharmaceutical manufacturing capacity. An example case illustrates the use of this framework where a biopharmaceutical company is faced with options for acquiring commercial manufacturing capacity. The development portfolio consists of three monoclonal antibody drugs at varying stages of clinical development with varying levels of demand. Capacity acquisition options include building in-house capacity, outsourcing, and partnering in addition to some hybrids of these. Deterministic and stochastic analyses showed that building manufacturing capacity ranked highest for the scenario considered when accounting for both financial and operational metrics. Secondly, the development of a stochastic combinatorial multi-objective optimisation framework is presented which confronts the problem of handling the multitude of decisions and trade-offs when designing portfolio management strategies, which results in extremely large decision spaces. The framework is considerate of strategic decisions that include the portfolio composition, the scheduling of critical development and manufacturing activities, and the involvement of third parties for these activities. The framework simulates development and manufacturing alongside the wider commercial environment. Machine learning and evolutionary computation techniques are also harnessed to characterise the conditional and probabilistic structure of superior decisions and evolve strategies to multi-objective optimality. A case study is constructed to derive insight from the framework where results demonstrate that a variety of options exist for formulating nondominated strategies in the objective space considered, giving the manufacturer a range of pursuable options. The most preferred means for development across the set of optimised strategies is to fully integrate development and commercial activities in-house, however, alternatives include partnering during early stages of portfolio development and then coordinating outsourced and in-house activities for remaining drugs. Popular scheduling strategies tend to develop two drugs in close succession while spacing out the remaining drug development activities into longer time frames. Thirdly, this framework is expanded to explore the impact of the size of biopharmaceutical drug development portfolio and cash flow constraints on algorithmically formulated strategies. Illustrative examples suggest that naively applying strategies optimal for a particular size of portfolio to a portfolio of another size is inappropriate. Also, the size of the portfolio appears to have a larger impact on strategy than the magnitude of cash flow constraint. Fourthly and finally, the economics of biopharmaceutical manufacture are explored with the aim of developing equations that can estimate the cost of manufacturing for both monoclonal antibodies and antibody fragments using mammalian cell culture and bacterial fermentation respectively. The correlations, derived using multiple linear regression, allow the cost of goods to be estimated given the following inputs: the required annual output, fermentation titre, whole process yield, and the probability of achieving a successful batch.

There are numerous potential advantages of producing significant quantities of a monoclonal antibody (MAb) via transgenic tobacco plants over other heterologous production systems, thus paving the way for new prophylactic and therapeutic applications within global human and animal health. However, current information on the key processing factors for large scale production of antibodies from transgenic plants is limited. This thesis presents the issues involved in the production of monoclonal antibodies in transgenic tobacco plants with a specific focus on initial extraction and aids the design and characterisation of an optimal small-scale extraction process using ultrascale down and micromanipulation techniques based on large-scale principals, in addition to offering different harvesting and extraction strategies dependent upon the specific target subcellular or tissue compartment. One of the preliminary objectives of this project was to examine methods for the initial extraction of recombinant IgGl antibodies from the leaf tissue of transgenic tobacco. Three different transgenic plant lines were investigated with the intention of establishing the parameters for optimal extraction of MAbs that reside in the apoplasm (IgG), endoplasmic reticulum (IgG-HDEL), or are bound to the plasma membrane (mlgG). For each transgenic line, seven techniques for physical extraction were evaluated. For IgG that is secreted and accumulated in the apoplasm, dry freeze-thaw (the freezing of leaf discs at -20 °C followed by room temperature thawing before buffer addition) was an appropriate technique for extraction of a high yield and a low release of native plant proteins from leaves in comparison to the other techniques investigated. In addition to lowering the downstream purification burden, the large-scale equipment involved in this step is likely to have a lower operating cost than a mechanical, energy-intensive grinding device. IgG-HDEL-expressing transgenic plants demonstrated an increase in IgG-HDEL yield with technique severity, demonstrating that harsher techniques such as dry-freeze-thaw followed by grinding were optimal. Conversely, the membrane-bound IgG required the leaf tissue to be ground in buffer that included a non-ionic detergent (Triton X-100), the optimal concentration of which was 0.1% (v/v). Grinding samples on ice or at room temperature was found to have no effect on IgG yield for all three MAbs. This indication of plant-derived IgG stability at room temperature is an obvious cost benefit at industrial scale. For all forms of the IgG, there was a wide variety of usable pHs (pH 5 to 7) with the exception of very low pHs (pH 3 and 4). Overall, an important finding of this study was that determining factors of optimal antibody extraction from plants had a direct influence on the initial choice of expression strategy, and thus it was essential to be addressed from the outset. In addition, a principally important consideration was the use of small scale techniques that were applicable to large scale purification. Another important factor of recombinant protein production in transgenic plants that is often overlooked is the initial bioprocesing step of harvesting. The major harvesting factors that need to be addressed are when to harvest, which part of the plant to harvest and how to harvest. Here some of these factors for the production of a secreted IgG and an intracellular retained form of this IgG in transgenic tobacco were addressed. Data analysis resulted in an interesting observation of plant wound response and its consequences for time-response IgG levels. The same monoclonal antibody (MAb), (Guy's 13 that acts against Streptococcus mutans, the main agent of tooth decay in the mouth) targeted to two different subcellular compartments, showed varying IgG response levels after wounding. In addition, there was a significantly different type of wound response and the subsequent IgG levels for young and old plants expressing the secreted form of IgG with a negative effect (IgG reduction) on young growing plants and a positive effect (IgG boost) in older plants. Additionally, for secreted IgG expressing plants, IgG strongly depended on plant age with the highest amount of IgG being found in young leaves of old plants and or young plants, but with a marked reduction in older tissue that was most likely due to senescence. In contrast, intracellularly retained IgG that accumulated in the endoplasmic reticulum was not significantly affected by mechanical wounding and its frequency or by overall senescence. In addition to transgenic tobacco leaves, tobacco roots were investigated as a potential source of the MAb. It was found that despite the focus of current related literature on recombinant protein recovery being from the leaves of whole transgenic tobacco plants, roots offer a promising alternative.

The potential design space for biological systems is complex, vast and multidimensional. Therefore, effective large-scale synthetic biology requires computational design and simulation. By constraining this design space, the time- and cost-efficient design of biological systems can be facilitated. One way in which a tractable design space can be achieved is to use the extensive and growing amount of biological data available to inform the design process. By using existing knowledge design efforts can be focused on biologically plausible areas of design space. However, biological data is large, incomplete, heterogeneous, and noisy. Data must be integrated in a systematic fashion in order to maximise its benefit. To date, data integration has not been widely applied to design in synthetic biology. The aim of this project is to apply data integration techniques to facilitate the efficient design of novel biological systems. The specific focus is on the development and application of integration techniques for the design of genetic regulatory networks in the model bacterium Bacillus subtilis. A dataset was constructed by integrating data from a range of sources in order to capture existing knowledge about B. subtilis 168. The dataset is represented as a computationally-accessible, semantically-rich network which includes information concerning biological entities and their relationships. Also included are sequence-based features mined from the B. subtilis genome, which are a useful source of parts for synthetic biology. In addition, information about the interactions of these parts has been captured, in order to facilitate the construction of circuits with desired behaviours. This dataset was also modelled in the form of an ontology, providing a formal specification of parts and their interactions. The ontology is a major step towards the unification of the data required for modelling with a range of part catalogues specifically designed for synthetic biology. The data from the ontology is available to existing reasoners for implicit knowledge extraction. The ontology was applied to the automated identification of promoters, operators and coding sequences. Information from the ontology was also used to generate dynamic models of parts. The work described here contributed to the development of a formalism called Standard Virtual Parts (SVPs), which aims to represent models of biological parts in a standardised manner. SVPs comprise a mapping between biological parts and modular computational models. A genetic circuit designed at a part-level abstraction can be investigated in detail by analysing a circuit model composed of SVPs. The ontology was used to construct SVPs in the form of standard Systems Biology Markup Language models. These models are publicly available from a computationally-accessible repository, and include metadata which facilitates the computational composition of SVPs in order to create models of larger biological systems. To test a genetic circuit in vitro or in vivo, the genetics elements necessary to encode the enitites in the in silico model, and their associated behaviour, must be derived. Ultimately, this process results in the specification for synthesisable DNA sequence. For large models, particularly those that are produced computationally, the transformation process is challenging. To automate this process, a model-to-sequence conversion algorithm was developed. The algorithm was implemented as a Java application called MoSeC. Using MoSeC, both CellML and SBML models built with SVPs can be converted into DNA sequences ready to synthesise. Selection of the host bacterial cell for a synthetic genetic circuit is very important. In order not to interfere with the existing cellular machinery, orthogonal parts from other species are used since these parts are less likely to have undesired interactions with the host. In order to find orthogonal transcription factors (OTFs), and their target binding sequences, a subset of the data from the integrated B. subtilis dataset was used. B. subtilis gene regulatory networks were used to re-construct regulatory networks in closely related Bacillus species. The system, called BacillusRegNet, stores both experimental data for B. subtilis and homology predictions in other species. BacillusRegNet was mined to extract OTFs and their binding sequences, in order to facilitate the engineering of novel regulatory networks in other Bacillus species. Although the techniques presented here were demonstrated using B. subtilis, they can be applied to any other organism. The approaches and tools developed as part of this project demonstrate the utility of this novel integrated approach to synthetic biology.

The focus of the research is on the analysis of complex bioprocess datasets with the ultimate goal of forming a link between the data and its underlying biological patterns. The challenges associated with investigating complex bioprocess data include the high dimensionality of the underlying measurements, the limited number of “observations”, and the complexity of selecting meaningful features to characterise the data. Contained within these data is a wealth of information that can contribute to inferring process outcomes and providing insight into improving productivity and process efficiency. To address these challenges, there is a real need for techniques to analyse and extract knowledge from the data. This thesis investigates an integrated discrete wavelet transform (DWT) and multiway principal components analysis (MPCA) approach to extract meaningful information from different types of bioprocess data. The integrated methodology is demonstrated by application to two types of bioprocess data: a near infrared (NIR) dataset collected from an industrial monoclonal antibodies (MAb) process, and an electrospray ionisation mass spectrometry (ESI-MS) dataset generated during the development of recombinant mammalian cell lines. The objective of the thesis was to develop a methodology that enabled the extraction of information from these two data sets. For the industrial NIR dataset, the genealogy or parent-child relationship of batch process from monoclonal antibodies (MAb) manufacturing was investigated whilst for the ESI-MS dataset goal was to identify characteristics that would enable the differentiation between high and low cell producers. The main challenges of the NIR and ESI-MS data sets lay in the complexity of the spectra. The NIR spectra usually have broad overlapping peaks and baseline shifts. Furthermore, as the NIR spectra used in this thesis were collected from batch process, there is an extra dimension in the data that of batch. On the one hand, the extra dimension provides extra information but on the other, it presents a further challenge as the data now is three-dimensional and requires additional pre-processing, including data matrix unfolding and batch alignment. Similar to the NIR spectra, the ESI-MS dataset also faces the problem of baseline shifts along with other complexities including high noise to signal ratio, shifts in the mass-to-charge ratio, and differences in signal intensities. These challenges lead to difficulties in extracting relevant information about the feature of interest. The proposed methodology was proven effective in extracting meaningful information from both data sets. In summary, the proposed method which utilised the integration of discrete wavelet transform and multiway principal component analysis was able to differentiate the distinguished characteristics of the spectra in the datasets thereby providing understanding of the relationships between spectral data and the underlying behaviour of the process.

Halophilic and halotolerant microorganisms are known to have numerous potential biotechnological applications, but to date, they have been largely underexploited in comparison to other extremophiles. The overall aim of the research presented within this thesis was to further investigate some of the biotechnologically useful products and processes that halophilic and halotolerant microorganisms are known to possess. Firstly, halotolerant bacterial isolates from a polluted marine environment were shown to be extremely tolerant to l-alkyl-3-methylimidazolium chloride ionic liquids, with much greater levels of tolerance to these compounds than has previously been reported for other microorganisms, and in addition, some of the isolates had the ability to biodegrade these compounds. This indicates that bacteria from the marine environment, due to their adaptation to salinity and the presence of hydrocarbons within this environment, would be highly suited to biological processes involving exposure to ionic liquids. Kilroot salt mine was investigated as a source of halophilic microorganisms as its culturable microbiome has never before been profiled; exploration of this environment indicated a great culturable biodiversity of both bacteria and archaea. The haloarchaeal isolates were shown to form biofilms, which enhanced the tolerance of the haloarchaea to an antimicrobial challenge. This is the first time this protective function ofhaloarchaeal biofilms has been demonstrated. Screening the isolates against a panel of antibiotics revealed an unexpectedly high level of natural resistance, indicating the presence of antimicrobial-producing microorganisms in the salt mine environment. To examine this further, organic extracts of each isolate from the mine were tested against a range of pathogenic bacteria, with approximately 40% displaying antimicrobial activities. One activity of particular interest was that of a haloarchaeal isolate of the genus Halorubrum, which exhibited both in vitro and in vivo antimicrobial and antibiofilm activity against P. aeruginosa. Whole genome sequence analysis of this isolate revealed further biotechnologically-important functions which provides numerous opportunities for additional studies, and reinforces the biotechnological potential of these organisms that is waiting to be exploited.

Bioprocesses for therapeutic protein production typically require significant resources to be invested in their development. This development could be improved with technologies that can elucidate the physicochemical properties of process stream components with small sample volumes in a rapid and readily performed manner. This is especially true in early phase development when material and established analytical methods are limiting. This thesis has investigated various process materials but was focussed mainly on those existing in an ApolipoproteinA-IM (ApoA-IM) process, produced using an Eschericia coli (E. coli) host. Using a mass spectrometric technique this project began by monitoring the product and contaminant during the ApoA-IM process and how this analytical approach compares to traditional analytical methods such as high performance liquid chromatography (HPLC). Results showed that, unlike many other analytical methods, surface enhanced laser desorption ionisation mass spectrometry (SELDI-MS) can handle early process samples that contain complex mixtures of biological molecules with limited sample pre-treatment and thereby provide meaningful process-relevant information. The change in material during the flocculation/centrifugation stage of the process was then examined. When only a change in cellular debris was observed an existing methodology developed at University College London (UCL) was implemented to maximise cellular debris removal. The predictive scale down methodology enabled rapid optimization of the operating conditions for a flocculation followed with a centrifugation step using only small volumes (20mL) of a high solids (~20% w/w) E. coli heat extract. These experiments suggested that adding a higher level of a cationic polymer could substantially increase the strength of the flocculated particles produced, thereby enhancing overall clarification performance in a large scale centrifuge. This was subsequently validated at pilot scale.The proteins remaining from this flocculation/centrifugation stage were then compared using the mass spectrometric technique to calculate the difficulty of removing each protein contaminant from the ApoA-IM product and suggested conditions for future sorbent scouting runs.

In order to examine the metabolism of bacteria in the genus Enterobacteriaceae tools for gene complement comparison and stoichiometric model building have been developed to take advantage of both the number of complete bacterial genome sequences currently available and the relationship between genes and metabolism. A functional genomic approach to improving knowledge of the metabolism of Escherichia coli CFT073 (a uropathogen) has been undertaken taking into account not only its genome sequence, but its close relationship to E. coli MG1655. A fresh comparison of E. coli CFT073 has been done with E. coli MG1655 to identify all those genes in CFT073 that are not present in MG1655 and may have metabolic characteristics. These genes have further been bioinformatically assessed to determine whether they might encode enzymes for the metabolism of chemicals commonly found in human urine, and one set of such genes has been experimentally confirmed to encode an L-sorbose utilisation pathway. Little experimental work has been done as yet to elucidate how bacteria adaptively respond to the introduction of heterologous metabolic genes. To investigate how bacteria respond to such DNA, genes encoding the L-sorbose utilisation and uptake operon from CFT073 have been cloned and transformed into DH5 and a selective pressure (minimal medium with L-sorbose as sole carbon source) has been applied over 100 generations of growth of this strain in serial passage to investigate the change in its behaviour. The availability of large numbers of completely sequenced genomes, along with the development of a stoichiometric metabolic model with very high coverage of E. coli metabolism (iAF1260 [1]) have made possible the analysis of the core metabolism of large numbers of bacteria to investigate gene essentiality in these bacteria. A novel way of assessing gene complement has been developed using BLAST and DiagHunter to improve reliability of gene synteny comparisons with contextual information about the genes and to extend work by others to cover all E. coli and Shigella genome sequences with available sequences on GanBank (as of 1st June 2009) in order to bioinformatically investigate essential genes in these bacteria and the heterogeneity of their metabolic networks. Further to this a metabolic model has been constructed for DH5 with an added L-sorbose pathway and for CFT073 and these models have been used to investigate behavioural changes during adaptation of bacteria to novel heterologous genes.

One by-product of the large-scale manufacture of biological products is the generation of significant quantities of process data. Typically this data is catalogued and stored in accordance with regulatory requirements, but rarely is it used to enhance subsequent production. A large amount of useful information is inherent in this data; the problems lie in the lack of appropriate methods to apply in order to extract it. The identification and/or development of tools capable of providing access to this valuable, untapped resource are therefore an important area for research. The main objective of this research is to investigate whether it is possible to attain knowledge from the information inherent within process data. The approach adopted in this thesis is to utilise the tools and techniques prevalent in the areas of data mining and pattern recognition. Through the application of these techniques, it is hypothesised that useful information can be acquired. Specifically the industrial sponsors of the research, Avecia Biologics, are interested in looking at methods for comparing new proteins to those they have previously worked on, with the intention of inferring information pertaining to the large scale manufacturing route for different processes. It is hypothesised that by comparing proteins and looking for similarities at the molecular level, it could be possible to identify potential pit-falls and bottlenecks in the recovery process before they occur. This would then allow Avecia to highlight areas of process development that may require specific attention. Two main techniques are the primary focus of the study; the Self-Organising Map (SOM) and the Support Vector Machine (SVM). Through a detailed investigation of these techniques, from benchmarking studies to applications with real-world problems, it is shown that these methods have the potential to become a useful tool for extracting information from biological process data.

Chromatography is an integral part of downstream processes to enable highly selective separation of biopharmaceutical products. A relatively high portion of total production costs are associated with the chromatography stage, hence optimisation of this unit operation is desirable. In this thesis column hydrodynamics were investigated under different operating conditions to provide an understanding of the effect of the column and matrix properties on the maximum flow velocity that the column is capable of being operate1at, the so called critical velocity, ucrit. Two automated methods for locating the column critical velocity were developed and compared to the conventional manual method. These automated methods utilised proportional-integral controllers and set either flow or pressure drop as the independent variable. Both methods proved to be relatively more efficient and accurate than the conventional manual inspection method, with the level of improvement depended on parameter settings such as the initial flow/pressure set point, the equilibration period, the step size, and the PID settings, etc. Of the three methods tested, the automatic pressure step (APS) method showed a better overall level of performance by providing better efficiency and accuracy. During chromatographic separations the column resin is commonly subjected to varying, and sometimes extreme, physiochemical conditions as part of the column cleaning procedures. Techniques were developed in this thesis to investigate the effects of pH changes on the ucrit. Results show that by utilising rigid resin materials, the significant reduction in the ucrit that was observed when changing the pH of the mobile phase from 7 to 13 can be avoided. This study demonstrates the importance of considering the impact of mobile phase pH on the column stability, especially when operating over long campaigns in which the column may be subjected to many cycles of widely different physiochemical conditions. To improve the ucrit, the idea of utilising various designs of inserts within the column to provide additional wall support was developed and tested. Results showed that despite leading to a reduction in the column efficiency as measured by HETP, inserts provide a positive impact on the column bed stability with an ucrit improvement of up to 48 % recorded. The utilisation of insert may be useful for the capture stage of a downstream process, where the benefits of increasing the permissible operation flow velocity often outweigh any drawbacks of decreased column efficiency.

The study of aggregation in proteins in this work demonstrates the need to understand its mechanism and moderate it as it is a bottleneck in bioprocessing and formulation where it reduces product yield. It explores the analytical techniques used to detect and analyze aggregation, puts forward thermodynamic parameters that can help predict precipitation in small scale bioprocess and determines aggregation prone regions and regions that may increase stability via site-directed mutagenesis. Thermodynamic parameters such as delta G and m values for precipitation were determined from global fit data of hen egg white lysozyme and alcohol dehydrogenase. The free energy was used to predict when a protein is likely to precipitate and the correlating m values were used to measure when a protein is likely to precipitate according to the dependence on ammonium sulphate. In addition, the curve fitting would allow the packing and nucleation of precipitated molecules to be determined which can be used to further correlate with aggregation mechanisms. The impact of sodium citrate and polyethylene glycol precipitation on fragment antibody binding was also examined. This identified a complex protein-polymer interaction between polyethylene glycol and fragment antibody binding whereby the interaction blocked sodium citrate fragmentation but promoted fragment antibody binding oligomerisation. The proportion of species of fragment antibody binding was further confirmed with analytical ultracentrifugation and small angle x-ray scattering analysis where results indicated that dimer was mostly present, which may act as the reactive species leading to aggregation. Site-directed mutagenesis of fragment antibody binding was also carried out, where an increase in hydrophobicity correlated with increased aggregation rates. The leucine to lysine mutant differed as this mutation was at the interface of the heavy and light chain, leading to fragmentation. The serine to lysine mutant on the other hand was the most stable, but did begin to aggregate after one month due to peptide hydrolysis and non-specific interactions. Further work with nuclear magnetic resonance indicated the pseudo wild type fragment antibody binding was correctly folded but had some dynamic conformational rearrangement indicative of some unfolding. This work highlights the need to understand the aggregation mechanisms under different conditions in order to moderate it in bioprocess. It also discusses the different models of aggregation where these are applied to the case study and whether aggregation can be engineered out of proteins.