The work described in this thesis was carried out at the Sir Peter Mansfield Magnetic Resonance Centre at the University of Nottingham between March 2006 and December 2009. All work described in this thesis was performed by the author, except where indicated. This thesis aims to develop and implement ASL techniques to measure haemodynamic responses to neural activity. The development of a new technique Double Acquisition Background Suppression (DABS) is presented as a remedy for a newly discovered artefact affecting Philips Achieva 7 T scanners and other sources of variation in baseline signals such as physiological noise. The new technique (DABS) was developed for simultaneous acquisition of ASL (with suppressed static tissue signal) and BOLD data using the FAIR scheme. This method not only provided a solution to obtaining ASL data at 7 T, despite the Roman Artefact, but also proved to reduce the contribution of physiological noise to ASL images, which is problematic, especially at ultra-high magnetic field strengths. The statistical verification was carried out based on the neural activation induced by a finger-tapping stimulus. A simplified model for quantifying CBVa.with the Look-Locker sampling method is proposed in this thesis to overcome the need for the Step-wise Compartmental Model (SCM). The Look-Locker sampling scheme acquires multiple readout pulses following the labelling and provides an estimation of transit time as well as CBVa. Here the simplified model is used to assess changes due to visual stimulation and validated against the SCM model. The application of LL-FAIR to form CBF and CBVa weighted data with improved SNR compared to traditional single TI FAIR technique is then shown. This method uses a summation over LL-EPI readout pulses and is used to asses the temporal characteristics and absolute changes in CBF and CBVa haemodynamic responses to a short (4.8 s) and long (9.6 s) visual stimulus. LL-FAIR methods are then used to appraise the neural coupling of haemodynamic parameters and assess Grubb's relationship. CBF and CBVa. Data were collected together with CBVtot data from a bolus injection of contrast agent. Assessing Grubb's power-law (CBVtot = CBFCI:)for neuronal activation, which was originally derived in primates during a steady state response of hypercapnia, a was found in this human study to be between 0.22 ± 0.08 and 0.29, dependent on the analysis method. In addition, the power-law relationship between CBVtot and CBVa.was assessed, and resulted in a similar relation, yielding aTA = 0.42 ± 0.14 and 0.40. Since CBF is thought to be driven by CBVa.the power-law between these parameters was also tested with a value of aFA = 1.35 ± 0.64 and 1.21, found in close agreement with earlier animal work.

This thesis concerns two areas of bioinformatics related by their role in protein structure and function: protein structure prediction and post translational modification of proteins. The dihedral angles Ψ and Φ are predicted using support vector regression. For the prediction of Ψ dihedral angles the addition of structural information is examined and the normalisation of Ψ and Φ dihedral angles is examined. An application of the dihedral angles is investigated. The relationship between dihedral angles and three bond J couplings determined from NMR experiments is described by the Karplus equation. We investigate the determination of the correct solution of the Karplus equation using predicted Φ dihedral angles. Glycosylation is an important post translational modification of proteins involved in many different facets of biology. The work here investigates the prediction of N-linked and O-linked glycosylation sites using the random forest machine learning algorithm and pairwise patterns in the data. This methodology produces more accurate results when compared to state of the art prediction methods. The black box nature of random forest is addressed by using the trepan algorithm to generate a decision tree with comprehensible rules that represents the decision making process of random forest. The prediction of our program GPP does not distinguish between glycans at a given glycosylation site. We use farthest first clustering, with the idea of classifying each glycosylation site by the sugar linking the glycan to protein. This thesis demonstrates the prediction of protein backbone torsion angles and improves the current state of the art for the prediction of glycosylation sites. It also investigates potential applications and the interpretation of these methods.

The Notch signalling pathway is pivotal in ensuring that the processes of arterial specification, angiogenic sprouting and haematopoietic stem cell (HSC) specification are correctly carried out in the dorsal aorta (DA), a primary arterial blood vessel in developing vertebrate embryos. Using the zebrafish as a model organism, and additional experimental observations from mouse and cell line models to guide mathematical modelling, this thesis aims to better understand the mechanisms involved in the establishment of a healthy vasculature in the growing embryo. We begin by studying arterial and HSC specification in the zebrafish DA. Mathematical models are used to analyse the dose response of arterial and HSC genes to an input Notch signal. The models determine how distinct levels of Notch signalling may be required to establish arterial and HSC identity. Furthermore, we explore how Delta-Notch coupling, which generates salt-and-pepper patterns, may drive the average gene expression levels higher than their homogeneous levels. The models considered here can qualitatively reproduce experimental observations. Using laboratory experiments, I was able to isolate DA cells from transgenic zebrafish embryos and generate temporal gene expression data using qPCR. We show that it is possible to fit ODE models to such data but more reliable data and a greater number of replicates at each time point is required to make further progress. The same VEGF-Delta-Notch signalling pathway is involved in tip cell selection in angiogenic sprouting. Using an ODE model, we rigourously study the dynamics of a VEGF-Delta-Notch feedback loop which is capable of amplifying differences betwen cells to form period-2 spatial patterns of alternating tip and stalk cells. The analysis predicts that the feeback strengths of Delta ligand and VEGFR-2 production dictate the onset of patterning in the same way, irrespective of the parameter values used. This model is extended to incorporate feedback from filopodia, growing in a gradient of extracellular VEGF, which are capable of facilitating tip cell selection by amplifying the resulting patterns. Lastly, we develop a PDE model which is able to properly account for VEGF receptor distributions in the cell membrane and filopodia. Receptors can diffuse and be advected due to domain growth, defined by a constitutive law, in this model. Our analysis and simulations predict that when receptor diffusivity is large, the ODE model for filopodia growth is an excellent approximation to the PDE model, but that for smaller diffusivity, the PDE model provides valuable insight into the pattern forming potential of the system.

Protein Structure Comparison (PSC) is at the core of many important structural biology problems. PSC is used to infer the evolutionary history of distantly related proteins; it can also help in the identification of the biological function of a new protein by comparing it with other proteins whose function has already been annotated; PSC is also a key step in protein structure prediction, because one needs to reliably and efficiently compare tens or hundreds of thousands of decoys (predicted structures) in evaluation of 'native-like' candidates (e.g. Critical Assessment of Techniques for Protein Structure Prediction (CASP) experiment). Each of these applications, as well as many others where molecular comparison plays an important role, requires a different notion of similarity, which naturally lead to the Multi-Criteria Protein Structure Comparison (MC-PSC) problem. ProCKSI (www.procksi.org), was the first publicly available server to provide algorithmic solutions for the MC-PSC problem by means of an enhanced structural comparison that relies on the principled application of information fusion to similarity assessments derived from multiple comparison methods (e.g. USM, FAST, MaxCMO, DaliLite, CE and TMAlign). Current MC-PSC works well for moderately sized data sets and it is time consuming as it provides public service to multiple users. Many of the structural bioinformatics applications mentioned above would benefit from the ability to perform, for a dedicated user, thousands or tens of thousands of comparisons through multiple methods in real-time, a capacity beyond our current technology. This research is aimed at the investigation of Grid-styled distributed computing strategies for the solution of the enormous computational challenge inherent in MC-PSC. To this aim a novel distributed algorithm has been designed, implemented and evaluated with different load balancing strategies and selection and configuration of a variety of software tools, services and technologies on different levels of infrastructures ranging from local testbeds to production level eScience infrastructures such as the National Grid Service (NGS). Empirical results of different experiments reporting on the scalability, speedup and efficiency of the overall system are presented and discussed along with the software engineering aspects behind the implementation of a distributed solution to the MC-PSC problem based on a local computer cluster as well as with a GRID implementation. The results lead us to conclude that the combination of better and faster parallel and distributed algorithms with more similarity comparison methods provides an unprecedented advance on protein structure comparison and analysis technology. These advances might facilitate both directed and fortuitous discovery of protein similarities, families, super-families, domains, etc, and also help pave the way to faster and better protein function inference, annotation and protein structure prediction and assessment thus empowering the structural biologist to do a science that he/she would not have done otherwise.

This thesis argues that bioethics emerged in the late 1960s and early 1970s not as a novel way to engage new technological or social ethical questions of life (bios), but rather as a late, post-Enlightenment secular phenomenon. In particular, bioethics seeks to adopt a methodology of theorizing on morality that is prominent in modern science, and this is a strategy that I contest by following Wittgenstein’s critique of scientific theorizing. Wittgenstein’s later exercises with language present a critical and clarifying way to identify the immanent and self-referential schema of principlism in bioethics. Additionally, I show how Wittgenstein’s approach to philosophy as a skillful and therapeutic activity rather than a cognitive content is informative for bioethics. Hence, I suggest that in pre-modern, traditional eras—or even in many contemporary non-Western global sectors—bioethics largely would be indistinct from religious and theological dogma and practices. I argue that the modern mind prioritizes material causality, leading to a moral techne that divides spirit from matter, vios from bios. Within such a schema, nature—and especially the medicalized human body—is managed, produced, and constructed. Furthermore, I argue that Wittgenstein gestures towards an ancient transcendent way beyond the modern division of vios and bios, and that a full vision of seeing life may be glimpsed through an apophatic epistemology that guides one towards an understanding of ethics itself as a form of apophatic and embodied knowledge.

In this thesis, a multi-scale approach is provided to a pharmacokinetic and a pharmacodynamic problem. The first part of this research provides a realistic mathematical physiological model of the liver to predict drug drug interactions (DDIs). The model describes the geometry of a lobule (liver unit) and integrates the exchange processes, diffusion and active transport, between the hepatocytes and the blood and possible drug-drug interactions such as; reversible inhibition, mechanistic based inhibition (MBI) and enzyme induction. The liver model is subsequently integrated into a PBPK model with 7 compartments (artery blood, venous blood, gut, liver, kidney, lung, rest of the body). To assess the efficiency of the model to predict DDIs, 77 clinical DDI studies were compared to the model. These 77 clinical studies represent 5 victim drugs (midazolam, simvastatin, triazolam, cerivastatin and nifedipine) and 30 perpetrator drugs. The reversible inhibition, MBI and induction parameters for the majority of the perpetrators were estimated with in vitro experiments and adjusted for the human liver size. The PK parameters, such as clearance and absorption rate, and the physiological parameters were obtained from the literature. The DDIs were measured as the ratio of the AUC (Area Under the Curve of the blood concentration) or the ratio of the maximum concentration Cmax of the victim drug administered with and without the perpetrator drug. The predicted ratios were compared with the clinical observation by calculating the geometric fold error GMFE. The GMFE for AUCratio and Cmax,ratio were calculated to be 1.54 and 1.34, respectively. Moreover, the PBPK model excluding the gut compartment under-predicts both inhibition (lower AUCratio) and induction (higher AUCratio) which strongly suggests that the gut DDI component can not be neglected for accurate clinical prediction. However, the static combined model by Fahmi et al. [1, 2] without the gut component fortuitously predicts the clinical AUCratio better than inclusion with the gut component. To conclude, the model predicts DDIs relatively well as it is in the lower range of errors reported in the literature (1.47-2.00 [1, 2]). Moreover, the model is able to predict the pharmacokinetics of drugs and provides a dynamic description of the DDIs, such as the enzyme level and spatial distribution within a lobule. Furthermore, the perpetrator dose regimen can be changed or the error in the in vitro parameters can be integrated to observe their influences on the AUC ratio. The second part of this research explored the Warburg effect in a avascular tumour growth model incorporating a cell shedding term to account for tumour shrinkage. The tumour model was based on an extension of the Ward and King model [3], where two sub-populations; living cells and dead cells are considered. Three diffusion equations for glucose, lactate and the drug are considered and included into the model for growth rate, natural death rate and a death rate due to the drug. The simulation of the model without a drug shows similar behaviour to the original model by Ward and King despite the presence of the shedding term and predicts an extracellular pH of 6.8. However, when a drug treatment is added, the model is able to simulate the shrinkage of the tumour unlike the original model. Moreover, two scenarios with a basic, neutral and acidic drug were explored, assuming similar efficiency at physiological pH to assess the effect of changes in the extracellular pH. Acidic or weak base drugs seem to be more efficient in low pH environment as the fraction of neutral form is greater and therefore more drug is available to cross the cell membrane to reach its target.

Planarian flatworms are gaining popularity in regenerative medicine research due to the fact that they have unparalleled regeneration capacity. Their tissue recovery abilities are dependent on a pool of adult stem cells (neoblasts). Studies in the recent years have shown that epigenetic mechanisms have an important role in neoblasts’ self-renewal and differentiation properties. This thesis focuses on the study of trithorax-related genes and their function in neoblast regulation. Despite the fact that mammalian trithorax-related genes Mll3 and Mll4 are among the most frequently mutated genes in cancer, trithorax-related genes are the least well-studies members of the trithorax gene group (TrxG) of histone modifiers. The current study traced the evolutionary history of trithorax-related genes and concluded that they have undergone a number of independent gene fission events across phyla. In planarians, there are three partial orthologue of the mammalian Mll3 and Mll4 genes – Smed-LPT (corresponding to the N-terminus of Mll3/4), Smed-trr-1 and Smed-trr-2 (both corresponding to the C-terminus of Mll3/4). The three planarian trithorax-related genes are expressed in stem cells and control neoblast differentiation down certain lineages (brain, gut, eyes, pharynx, epidermis). Down-regulation of Smed-LPT results in hyperproliferation of stem cells, leading to tumour-like outgrowth formation. It was shown that trithorax-related genes’ function in stem cell regulation correlates with histone modification changes, specifically alterations in H3K4me1, H3K4me3 and H3K27me3. Future studies will focus on examining this correlation further via Next-Generation sequencing techniques.

This thesis considers the statistical modelling and analysis of proteomic mass spectrometry data. Proteomics is a relatively new field of study and tried and tested methods of analysis do not yet exist. Mass spectrometry output is high-dimensional and so we firstly develop an algorithm to identify peaks in the spectra in order to reduce the dimensionality of the datasets. We use the results along with a variety of classification methods to examine the classification of new spectra based on a training set. Another method to reduce the complexity of the problem is to fit a parametric model to the data. We model the data as a mixture of Gaussian peaks with parameters representing the peak locations, heights and variances, and apply a Bayesian Markov chain Monte Carlo (MCMC) algorithm to obtain their estimates. These resulting estimates are used to identify m/z values where differences are apparent between groups, where the m/z value of an ion is its mass divided by its charge. A multilevel modelling framework is also considered to incorporate the structure in the data and locations exhibiting differences are again obtained. We consider two mass spectrometry datasets in detail. The first consists of mass spectra from breast cancer cells which either have or have not been treated with the chemotherapeutic agent Taxol. The second consists of mass spectra from melanoma cells classified as stage I or stage IV using the TNM system. Using the MCMC and multilevel techniques described above we show that, in both datasets, small subsets of the available m/z values can be identified which exhibit significant differences in protein expression between groups. Also we see that good classification of new data can also be achieved using a small number of m/z values and that the classification rate does not fall greatly when compared with results from the complete spectra. For both datasets we compare our results with those in the literature which use other techniques on the same data. We conclude by discussing potential areas for further research.

Motivation: A rapid technological development in the biosciences and in computer science in the last decade has enabled the analysis of high-dimensional biological datasets on standard desktop computers. However, in spite of these technical advances, common properties of the new high-throughput experimental data, like small sample sizes in relation to the number of features, high noise levels and outliers, also pose novel challenges. Ensemble and consensus machine learning techniques and data integration methods can alleviate these issues, but often provide overly complex models which lack generalization capability and interpretability. The goal of this thesis was therefore to develop new approaches to combine algorithms and large-scale biological datasets, including novel approaches to integrate analysis types from different domains (e.g. statistics, topological network analysis, machine learning and text mining), to exploit their synergies in a manner that provides compact and interpretable models for inferring new biological knowledge. Main results: The main contributions of the doctoral project are new ensemble, consensus and cross-domain bioinformatics algorithms, and new analysis pipelines combining these techniques within a general framework. This framework is designed to enable the integrative analysis of both large- scale gene and protein expression data (including the tools ArrayMining, Top-scoring pathway pairs and RNAnalyze) and general gene and protein sets (including the tools TopoGSA , EnrichNet and PathExpand), by combining algorithms for different statistical learning tasks (feature selection, classification and clustering) in a modular fashion. Ensemble and consensus analysis techniques employed within the modules are redesigned such that the compactness and interpretability of the resulting models is optimized in addition to the predictive accuracy and robustness. The framework was applied to real-word biomedical problems, with a focus on cancer biology, providing the following main results: (1) The identification of a novel tumour marker gene in collaboration with the Nottingham Queens Medical Centre, facilitating the distinction between two clinically important breast cancer subtypes (framework tool: ArrayMining) (2) The prediction of novel candidate disease genes for Alzheimer’s disease and pancreatic cancer using an integrative analysis of cellular pathway definitions and protein interaction data (framework tool: PathExpand, collaboration with the Spanish National Cancer Centre) (3) The prioritization of associations between disease-related processes and other cellular pathways using a new rule-based classification method integrating gene expression data and pathway definitions (framework tool: Top-scoring pathway pairs) (4) The discovery of topological similarities between differentially expressed genes in cancers and cellular pathway definitions mapped to a molecular interaction network (framework tool: TopoGSA, collaboration with the Spanish National Cancer Centre) In summary, the framework combines the synergies of multiple cross-domain analysis techniques within a single easy-to-use software and has provided new biological insights in a wide variety of practical settings.