Doctoral theses
Collections
- Faculty of Agriculture and Forestry [0]
- Faculty of Arts [0]
- Faculty of Biological and Environmental Sciences [0]
- Faculty of Educational Sciences [1]
- Faculty of Law [0]
- Faculty of Medicine [0]
- Faculty of Pharmacy [1]
- Faculty of Science [2]
- Faculty of Social Sciences [0]
- Faculty of Theology [0]
- Faculty of Veterinary Medicine [0]
Recent Submissions
-
(2023)A wide variety of nitrogen-containing compounds present in the air can contribute to air pollution, which in turn affects both human health and the climate. In this thesis, the applicability of two miniaturized air sampling techniques, solid-phase microextraction (SPME) Arrow and in-tube extraction (ITEX) was studied, for the selective collection of nitrogen-containing compounds in air samples. Different types of sorbent materials, including Mobil Composition of Matter No. 41 (MCM-41), titanium hydrogen phosphate-modified MCM-41 (MCM-41-TP), and zinc oxide-modified mesoporous silica microspheres, were used as sorbent materials in the ITEX sampling system. The adsorption and desorption behavior of gaseous nitrogen-containing compounds in passive SPME-Arrow and active ITEX sampling systems, coated and packed with different sorbent materials, was investigated. In addition, saturation vapor pressures of atmospheric trace gases were experimentally and theoretically estimated. The sampling systems with selected sorbent materials were applied to the determination of nitrogen-containing compounds in boreal forest SMEAR II station, indoor air, and cigarette smoke. Adsorbent and adsorbate properties, such as hydrophobicity and basicity, were the major factors that affected sorbent selectivity towards nitrogen-containing compounds. Moreover, the pore volume and pore sizes of the sorbents were essential parameters for the adsorption performance, especially in the SPME Arrow system. The ITEX packing and the SPME Arrow coatings were reproducible and reusable. Due to the active sampling principle, the ITEX sampler with higher adsorption and desorption rates provided better results for the analysis, especially when quick injection was needed in gas chromatography. The selectivity of the ITEX sampling system was increased with the trap accessory, but further study is needed to prevent the loss of the targeted compounds. Whereas the ITEX’s filter accessory was successfully employed to remove particles, enabling ITEX to collect only gas-phase samples. Vapor pressure results were achieved by laboratory experiments (by retention index approach) and by the COSMO-RS model. An aerial drone was successfully employed as a platform to study vertical profiles of VOCs at high altitudes, from 50 to 400 m, for miniaturized SPME Arrow and ITEX atmospheric air sampling systems, along with portable devices for the real-time measurement of black carbon (BC) and total particle numbers. There was a clear distribution of the nitrogen-containing compounds collected at different altitudes at SMEAR II station, Finland, depending on their sources. In addition, other VOCs demonstrated the same trend.
-
(2023)The vast majority of the world's languages are low-resource, lacking the data resources required in advanced natural language processing (NLP) based on data-intensive deep learning. Furthermore, annotated training data can be insufficient in some domains even within resource-rich languages. Low-resource NLP is crucial for both the inclusion of language communities in the NLP sphere and the extension of applications over a wider range of domains. The objective of this thesis is to contribute to this long-term goal especially with regard to truly low-resource languages and domains. We address truly low-resource NLP in the context of two tasks. First, we consider the low-level task of cognate identification, since cognates are useful for the cross-lingual transfer of many lower-level tasks into new languages. Second, we examine the high-level task of document planning, a fundamental task in data-to-text natural language generation (NLG), where many domains are low-resource. Thus, domain-independent document planning supports the transfer of NLG across domains. Following recent encouraging results, we propose neural network models to these tasks, using transfer learning methods in three low-resource scenarios. We divide our high-level objective into three research tasks characterised by different resource conditions. In our first research task, we address cognate identification in endangered Sami languages of the Uralic family, given scarce labelled training data. We propose a Siamese convolutional neural network (S-CNN) and a support vector machine (SVM), which we pre-train on unrelated Indo-European data, lacking high-resource close relatives. We find that S-CNN performs best at direct transfer to Sami, and adapts fast when fine-tuned on a small amount of Sami data. In our second research task, we address a scenario with only unlabelled data to adapt S-CNN from Indo-European to Uralic data. We propose both discriminative adversarial networks and pre-trained symbol embeddings, finding that adversarial adaptation outperforms an unadapted model, while symbol embeddings are beneficial when languages have disparate orthographies. In our third research task, we address document planning in data-to-text generation of news, in a domain with no annotated training data whatsoever. We propose distant supervision, automatically constructing labelled data from a news corpus, and train a neural model for sentence ordering, a task related to document planning. We examine Siamese, positional, and pointer networks, and find that a variant of S-CNN results in generation with higher human-perceived quality than heuristic baselines. The contributions of this thesis include addressing novel low-resource scenarios considering two NLP tasks, at which the potential of deep learning has not been fully explored. We propose novel approaches to these tasks using neural models in combination with transfer learning, and our experiments indicate their performance in comparison with baselines. Finally, although we acknowledge that rule-based methods and heuristics might still be superior to deep learning in truly low-resource scenarios, our approaches are more language- and domain-independent, supporting a wider coverage of NLP across languages and domains.
-
(2023)Checkpoint inhibitors have been regarded as a milestone in cancer therapy due to the imminent success in the clinics. Among such checkpoint inhibitors, PD-L1 antibodies have shown outstanding clinical efficacy and have been approved for the treatment of more than 14 different type of cancers. Despite the success, only 14% of patients are eligible for PD-L1 antibody therapy and from them only 44% respond. Therefore, a clear improvement of such therapy is required. Such antibodies are not able to elicit effector mechanisms due to point mutations in the IgG Fc-region removing binding to Fc- receptors. This is done because of safety concerns, since PD-L1 expression is not solely limited to the tumor but widely expressed on different types of healthy cells. Nevertheless, in vivo studies have shown that arming PD-L1 checkpoint inhibitors with Fc-regions able to elicit effector mechanism leads to increased tumor killing. In need of enhanced PD-L1 checkpoint inhibitors, in this thesis we developed two powerful PD-L1 checkpoint inhibitors able to elicit Fc-effectors mechanisms of both an IgG1 and IgA1. Moreover, to limit side effects we used oncolytic adenoviruses as biological carries to express and limit the secretion of the PD-L1 checkpoint inhibitors to the tumor. Oncolytic adenoviruses have a specific tumor-tropism that can be utilized to express any desired gene of interest to the tumor microenvironment. Nevertheless, the current methods using homologous recombination to clone oncolytic adenoviruses are time-consuming and inefficient. In Study I, we designed and tested a novel cloning method, called GAMER-Ad, which utilizes the Gibson assembly method rather than homologous recombination for cloning. To test GAMER-Ad, we designed three oncolytic adenoviruses to express CXCL9, CXCL10 or IL-15. GAMER-Ad was shown to be a viable strategy to clone oncolytic adenoviruses in the period of 2-3 days. Also, the cloning method did not affect the oncolytic/replication fitness of the viruses and yielded functioning viruses able to express the corresponding gene. GAMER-Ad was then used in the following studies to clone all oncolytic adenoviruses. In Study II, we developed an oncolytic adenovirus (Ad-Cab) able to secrete a PD-L1 checkpoint inhibitor able to elicit Fc-effector mechanisms of an IgG1 and IgA1. The expressed checkpoint inhibitor consisted of PD-1 ectodomain (able to bind to murine and human PD-L1) connected to a cross-hybrid IgGA Fc-region (contains heavy chain regions of an IgG1 and IgA1). The virally released cross-hybrid Fc-fusion peptide was able to activate PBMCs, PMNs, complement proteins and macrophages not usually done by either IgG1 or IgA antibodies solely. The engagement of multiple effector mechanisms did lead to an enhanced tumor killing by Ad-Cab compared to PD-L1 antibodies with an IgG1 or IgA Fc-region when all immune components were present. This enhancement was also translated in vivo since Ad-Cab outperformed conventional PD-L1 antibodies with various tumor models (4T1, CT26 and A549). This enhancement was attributed to an increased activation of NK cells and reduction of myeloid derived suppressor cells in vivo. Moreover, Ad-Cab was shown not to require CD8+ T cells for in vivo efficacy unlike the conventional PD-L1 antibodies used. As expected, no signs of toxicity were observed since no reduction in weight was observed in mice and the Fc-fusion peptide was limited to the tumor. Therefore, arming PD-L1 antibodies with Fc-effector mechanisms of an IgG1 and IgA1 leads to higher tumor-killing and safety concerns can be circumvented using oncolytic adenoviruses Further building on Ad-Cab, in Study III we designed Ad-Cab FT which had the same Fc-fusion peptide designed in Study II but with four-point mutations in the IgG region. These point mutations increased the affinity towards activating Fc- receptors leading to higher NK cell activation. At higher concentrations, Ad-Cab FT had similar levels of tumor lysis as Ad-Cab when PBMCs were added. However, at lower concentrations Ad-Cab FT induced higher tumor killing than Ad-Cab with PBMCs. This enhancement was not shown with PMNs or complement activation. Due to the high activation of PBMCs at lower concentration, Ad-Cab FT outperformed Ad-Cab in vivo when low doses and reduced administrations of the virus was given. With Ad-Cab FT treated mice, a higher activation of NK cells in the tumor microenvironment was observed compared to Ad-Cab treated mice. Hence, Ad-Cab FT represents a potentiated therapy with potential use in the clinic. Taken together, this thesis has highlighted the importance of eliciting multiple immune populations to enhanc tumor killing with PD-L1 checkpoint inhibitors and potentially with other therapies. The IgGA Fc-region may be used in other antibody-based therapies to further increase tumor killing and subsequently clinical efficacy. Finally, oncolytic adenoviruses have demonstrated in this thesis to be excellent biological carriers, limiting the toxicity of dangerous anti-tumor agents.