Thesis (M. Phil.)--University of Hong Kong, 2005. / Title proper from title frame. Also available in printed format.

München, Techn. Univ., Diss., 2004.

Th. 3e cycle--Math. pures--Grenoble 1, 1982. N°: 5.

Th. 3e cycle--Électronique, électrotech., autom.--Toulouse 3, 1977. N°: 1971.

The skin is an important barrier against a range of different environmental challenges. The skin associated immune system is able to detect breaks in the barrier to initiate a protective immune response. Langerhans cells express a high density of CD1a, which presents lipid antigens to T-cells. However little is known about CD1a-restricted lipid antigens and the role of CD1a-restricted T-cells in inflammatory skin disease. In order to investigate the role of T-cells that react to CD1a presenting lipid in skin disease, wasp and bee venom were studied as an antigen source. Venoms are able to cause allergic hypersensitivity reactions, associated with skin T-cell infiltration, a venom protein-specific T-cell response in allergic individuals and are independent of filaggrin. Using primary antigen presenting cells and target cells lacking surface MHC expression (K562 cells) transfected with CD1a, allowed investigation of polyclonal T-cells responses from unrelated donors. Bee and wasp venoms were shown to induce CD1a-restricted T-cell responses in both peripheral blood and skin. Surprisingly this activity was not contained within the lipid fraction of the venoms, but instead was mediated through the generation of a lipid ligand by venom phospholipase. Furthermore, wasp venom delivery results in the production of phospholipase products in the skin of humans. A significantly increased frequency of IFNγ-, GM-CSF- and IL-13-producing venom specific CD1a-restricted T-cells was observed in allergic individuals compared to healthy controls. During subcutaneous immunotherapy, frequencies of CD1a-reactive T cells were initially induced, peaking by weeks 5, but then reduced despite escalation of antigen dose. CD1a-reactive T cells were further investigated to characterise their physiological role by generating T cell lines which produced a range of different cytokines, including IL-22 on stimulation with phospholipase and other phospholipase containing allergens. In summary, we identified a novel pathway of skin inflammation where lipids generated by allergen-derived phospholipase can be recognised by CD1a-restricted T cells which produce type 1 and type 2 cytokines and associate with allergic reactivity. These findings have implications for novel therapeutic strategies for allergic disease.

This practice-based PhD consists of a portfolio of creative work and a supporting commentary. The portfolio illustrates the design decisions relating to my digital music performance system, and focuses upon the visibility and the fluidity of digital music performance. The goal of the design is to enhance the visibility without violating the audience’s auditory imagination unnecessarily, and to enhance the fluidity without relinquishing the unique fixed nature of digital music. The performance system consists of an audio engine, control mapping engine and visual engine. The audio engine and the control mapping engine were programmed with Cycling ‘74 Max. They let the performer deconstruct and reconstruct pre-recorded audio files with her/his hands via MIDI controllers during performance. The visual engine was programmed with Derivative TouchDesigner. In various ways, it visualises and exaggerates the performer’s actions which cause sonic changes, and filters out the rest. The works are presented as videos and the supporting commentary, deals with the contexts and thinking processes which determined the current performance system. By exploring theories of electronic music performance, audiovisual, visual music, acousmatic and medium specificity, I aim to explain the reasoning behind the performance system.

Computational studies on light‐tissue interactions in medical treatment and diagnosis have offered deeper insights in the processes underlying laser treatments and fluorescence measurements. I apply this approach in the study of fluorescence detection and of laser therapy. First, I investigate three methods of fluorescence detection and the reported contrast between healthy skin and malignant tissue. I varied the concentration of haemoglobin in the target, the concentration of melanin in the epidermis, the scattering of light in the skin, the depth at which the target is located in the skin, the width of the target, the thickness of the target, the concentration of photosensitizer in the target, and the concentration of photosensitizer in the skin. My findings confirm previous clinical studies in that the auto‐fluorescence corrected fluorescence detection method generally shows a higher contrast than the other methods. The results support earlier clinical studies and are in accordance with expert experience. Second, I study laser therapy for psoriasis. In a series of simulations, I analyse three types of pulsed dye laser systems and one IPL system. The investigated biological effects are heat shock proteins, hyperthermic tissue damage and vasoconstriction of the microvasculature. The changes in the skin concern blood volume, blood oxygenation and scattering in the epidermis. The calculations show that there are some notable differences in the effect changes in the composition of psoriatic tissue has on the efficacy of laser and IPL therapy. Still, Inter‐device variance was more prominent than intra‐geometry variance. My study adds to the understanding of fluorescence detection of keratinocyte skin cancers, as well as that of laser therapy for psoriasis. Additionally, it offers potential avenues for increasing the efficacy and efficiency of these therapies.

Production of recombinant FVIII, the protein that is missing or dysfunctional in haemophilia A patients, is highly inefficient compared to other recombinant clotting factors such as FIX. This is predominantly due to complex intracellular trafficking, short half-life and protein instability. This study aimed to increase the amount of functional FVIII produced in mammalian cells by co-expression with anti-FVIII Camelid antibody fragments (VHH). Three VHH ligands were supplied by BAC BV (as DNA constructs), two of which when expressed in yeast are known to bind recombinant FVIII (ligands 2 and 7) and are used commercially as FVIII purification tools. From these three constructs, nine new VHH plasmids constructs were designed and transiently expressed in a stable BHK-human FVIII-expressing cell line. Of the nine VHH fragments that were co-expressed in the BHK FVIII cell line, four of these had a statistically significant impact on the ‘clotting time’ of the cell media as demonstrated by the activated partial thromboplastin time assay (aPTT). Two ligand 2 constructs (L2C1 and L2C2) prolonged the coagulation time by 4 seconds (P-value 0.0001, 95% confidence intervals 38.5-43.5), and 3.4 seconds (P-value 0.0072, 95% CI 36.5-40.3) respectively, indicating a decrease in functional FVIII activity versus media from the untransfected and null transfected BHK-FVIII cell line. Two ligand 7 constructs (L7C1 and L7C3) caused a decrease in coagulation time of 3.2 seconds (P=0.0057, 95% CI 30.5-33.3), and 4 seconds (P=0.0002, 95% CI 29.1-32.9) respectively, indicating an increase in functional FVIII activity versus media from the untransfected and null transfected BHK-FVIII cell line. Ligand 7 and ligand 2 both bind to the FVIII light chain, albeit in different regions and with different affinities (data confidential to BAC BV). BAC studies showed that ligand 7 competes with vWF on the FVIII light chain, which is known to increase stability of FVIII in vivo, whereas ligand 2 does not compete for this binding site. The opposing effects of ligand 7 and ligand 2 on FVIII clotting times seen in this study could be due to their differences in FVIII binding properties, since it is known that binding location of FVIII ligands can have an impact on FVIII clotting activity.

Pattern recognition has become an accessible tool in developing advanced adaptive products. The need for such products is not diminishing but on the contrary, requirements for systems that are more and more aware of their environmental circumstances are constantly growing. Feed-forward neural networks are used to learn patterns in their training data without the need to discover by hand the relationships present in the data. However, the problem of estimating the required size of the neural network is still not solved. If we choose a neural network that is too small for a particular given task, the network is unable to "comprehend" the intricacies of the data. On the other hand if we choose a network size that is too big for the given task, we will observe that there are too many parameters to be tuned for the network, or we can fall in the "Curse of dimensionality" or even worse, the training algorithm can easily be trapped in local minima of the error surface. Therefore, we choose to investigate possible ways to find the 'Goldilocks' size for a feed-forward neural network (which is just right in some sense), being given a training set. Furthermore, we used a common paradigm used by the Roman Empire and employed on a wide scale in computer programming, which is the "Divide-et-Impera" approach, to divide a given dataset in multiple sub-datasets, solve the problem for each of the sub-dataset and fuse the results of all the sub-problems to form the result for the initial problem as a whole. To this effect we investigated modular neural networks and their performance.

The very large dimensionality of real world datasets is a challenging problem for classification algorithms, since often many features are redundant or irrelevant for classification. In addition, a very large number of features leads to a high computational time for classification algorithms. Feature selection methods are used to deal with the large dimensionality of data by selecting a relevant feature subset according to an evaluation criterion. The vast majority of research on feature selection involves conventional single-label classification problems, where each instance is assigned a single class label; but there has been growing research on more complex multi-label classification problems, where each instance can be assigned multiple class labels. This thesis proposes three types of new Multi-Label Correlation-based Feature Selection (ML-CFS) methods, namely: (a) methods based on hill-climbing search, (b) methods that exploit biological knowledge (still using hill-climbing search), and (c) methods based on genetic algorithms as the search method. Firstly, we proposed three versions of ML-CFS methods based on hill climbing search. In essence, these ML-CFS versions extend the original CFS method by extending the merit function (which evaluates candidate feature subsets) to the multi-label classification scenario, as well as modifying the merit function in other ways. A conventional search strategy, hill-climbing, was used to explore the space of candidate solutions (candidate feature subsets) for those three versions of ML-CFS. These ML-CFS versions are described in detail in Chapter 4. Secondly, in order to try to improve the performance of ML-CFS in cancer-related microarray gene expression datasets, we proposed three versions of the ML-CFS method that exploit biological knowledge. These ML-CFS versions are also based on hill-climbing search, but the merit function was modified in a way that favours the selection of genes (features) involved in pre-defined cancer-related pathways, as discussed in detail in Chapter 5. Lastly, we proposed two more sophisticated versions of ML-CFS based on Genetic Algorithms (rather than hill-climbing) as the search method. The first version of GA-based ML-CFS is based on a conventional single-objective GA, where there is only one objective to be optimized; while the second version of GA-based ML-CFS performs lexicographic multi-objective optimization, where there are two objectives to be optimized, as discussed in detail in Chapter 6. In this thesis, all proposed ML-CFS methods for multi-label classification problems were evaluated by measuring the predictive accuracies obtained by two well-known multi-label classification algorithms when using the selected featuresม namely: the Multi-Label K-Nearest neighbours (ML-kNN) algorithm and the Multi-Label Back Propagation Multi-Label Learning Neural Network (BPMLL) algorithm. In general, the results obtained by the best version of the proposed ML-CFS methods, namely a GA-based ML-CFS method, were competitive with the results of other multi-label feature selection methods and baseline approaches. More precisely, one of our GA-based methods achieved the second best predictive accuracy out of all methods being compared (both with ML-kNN and BPMLL used as classifiers), but there was no statistically significant difference between that GA-based ML-CFS and the best method in terms of predictive accuracy. In addition, in the experiment with ML-kNN (the most accurate) method selects about twice as many features as our GA-based ML-CFS; whilst in the experiments with BPMLL the most accurate method was a baseline method that does not perform any feature selection, and runs the classifier once (with all original features) for each of the many class labels, which is a very computationally expensive baseline approach. In summary, one of the proposed GA-based ML-CFS methods managed to achieve substantial data reduction, (selecting a smaller subset of relevant features) without a significant decrease in predictive accuracy with respect to the most accurate method.