Models and Simulations of Diamond-like Carbon for large-area high voltage power diodes

Silicon-based discrete high-power devices need to be designed with optimal performance up to several thousand volts and amperes to reach power ratings ranging from few kWs to beyond the 1 GW mark. To this purpose, a key element is the improvement of the junction termination (JT) since it allows to drastically reduce surface electric field peaks which may lead to an earlier device failure. This thesis will be mostly focused on the negative bevel termination which from several years constitutes a standard processing step in bipolar production lines. A simple methodology to realize its counterpart, a planar JT with variation of the lateral doping concentration (VLD) will be also described. On the JT a thin layer of a semi insulating material is usually deposited, which acts as passivation layer reducing the interface defects and contributing to increase the device reliability. A thorough understanding of how the passivation layer properties affect the breakdown voltage and the leakage current of a fast-recovery diode is fundamental to preserve the ideal termination effect and provide a stable blocking capability. More recently, amorphous carbon, also called diamond-like carbon (DLC), has been used as a robust surface passivation material. By using a commercial TCAD tool, a detailed physical explanation of DLC electrostatic and transport properties has been provided. The proposed approach is able to predict the breakdown voltage and the leakage current of a negative beveled power diode passivated with DLC as confirmed by the successfully validation against the available experiments. In addition, the VLD JT proposed to overcome the limitation of the negative bevel architecture has been simulated showing a breakdown voltage very close to the ideal one with a much smaller area consumption. Finally, the effect of a low junction depth on the formation of current filaments has been analyzed by performing reverse-recovery simulations.

Structural and biophysical characterization of RAD52 as novel pharmacological target for synthetic lethality

RAD52 is a protein involved in various DNA reparation mechanisms. In the last few years, RAD52 has been proposed as a novel pharmacological target for cancer synthetic lethality strategies. Hence, this work has the purpose to investigate RAD52 protein, with biophysical and structural tools to shed light on proteins features and mechanistic details that are, up to now poorly described, and to design novel strategies for its inhibition. My PhD work had two goals: the structural and functional characterization of RAD52 and the identification of novel RAD52 inhibitors. For the first part, RAD52 was characterized both for its DNA interaction and oligomerization state together with its propensity to form high molecular weight superstructures. Moreover, using EM and Cryo-EM techniques, additional RAD52 structural hallmarks were obtained, valuable both for understanding protein mechanism of action and for drug discovery purpose. The second part of my PhD project focused on the design and characterization of novel RAD52 inhibitors to be potentially used in combination therapies with PARPi to achieve cancer cells synthetic lethality, avoiding resistance occurrence and side effects. With this aim we selected and characterized promising RAD52 inhibitors through three different approaches: 19F NMR fragment-based screening; virtual screening campaign; aptamers computational design. Selected hits (fragments, molecules and aptamers) were investigated for their binding to RAD52 and for their mechanism of inhibition. Collected data highlighted the identification of hits worthy to be developed into more potent and selective RAD52 inhibitors. Finally, a side project carried out during my PhD is reported. GSK-3β protein, an already validated pharmacological target was investigated using biophysical and structural biology tools. Here, an innovative and adaptable drug discovery screening pipeline able to directly identify selective compounds with binding affinities not higher than a reference binder was developed.

Molecular changes induced by a centrally-induced state of synthetic torpor in multiple organs:a new strategy for radioprotection.

Torpor is a successful survival strategy displayed by several mammalian species to cope with harsh environmental conditions. A complex interplay of ambient, genetic and circadian stimuli acts centrally to induce a severe suppression of metabolic rate, usually followed by an apparently undefended reduction of body temperature. Some animals, such as marmots, are able to maintain this physiological state for months (hibernation), during which torpor bouts are periodically interrupted by short interbouts of normothermia (arousals). Interestingly, torpor adaptations have been shown to be associated with a large resistance towards stressors, such as radiation: indeed, if irradiated during torpor, hibernators can tolerate higher doses of radiation, showing an increased survival rate. New insights for radiotherapy and long-term space exploration could arise from the induction of torpor in non-hibernators, like humans. The present research project is centered on synthetic torpor (ST), a hypometabolic/hypothermic condition induced in a non-hibernator, the rat, through the pharmacological inhibition of the Raphe Pallidus, a key brainstem area controlling thermogenic effectors. By exploiting this procedure, this thesis aimed at: i) providing a multiorgan description of the functional cellular adaptations to ST; ii) exploring the possibility, and the underpinning molecular mechanisms, of enhanced radioresistance induced by ST. To achieve these aims, transcriptional and histological analysis have been performed in multiple organs of synthetic torpid rats and normothermic rats, either exposed or not exposed to 3 Gy total body of X-rays. The results showed that: i) similarly to natural torpor, ST induction leads to the activation of survival and stress resistance responses, which allow the organs to successfully adapt to the new homeostasis; ii) ST provides tissue protection against radiation damage, probably mainly through the cellular adaptations constitutively induced by ST, even though the triggering of specific responses when the animal is irradiated during hypothermia might play a role.

Study of the urinary biomarkers of synthetic cannabinoid receptor agonists (SCRAs) for the forensic unequivocal proof of consumption: medico-legal implications for drug abuse prevention

Introduction. Synthetic cannabinoid receptor agonists (SCRAs) represent the widest group of New Psychoactive Substances (NPS) and, around 2021-2022, new compounds emerged on the market. The aims of the present research were to identify suitable urinary markers of Cumyl-CB-MEGACLONE, Cumyl-NB-MEGACLONE, Cumyl-NB-MINACA, 5F-EDMB-PICA, EDMB-PINACA and ADB-HEXINACA, to present data on their prevalence and to adapt the methodology from the University of Freiburg to the University of Bologna. Materials and methods. Human phase-I metabolites detected in 46 authentic urine samples were confirmed in vitro with pooled human liver microsomes (pHLM) assays, analyzed by liquid chromatography-quadrupole time-of-flight mass spectrometry (LC-qToF-MS). Prevalence data were obtained from urines collected for abstinence control programs. The method to study SCRAs metabolism in use at the University of Freiburg was adapted to the local facilities, tested in vitro with 5F-EDMB-PICA and applied to the study of ADB-HEXINACA metabolism. Results. Metabolites built by mono, di- and tri-hydroxylation were recommended as specific urinary biomarkers to monitor the consumption of SCRAs bearing a cumyl moiety. Monohydroxylated and defluorinated metabolites were suitable proof of 5F-EDMB-PICA consumption. Products of monohydroxylation and amide or ester hydrolysis, coupled to monohydroxylation or ketone formation, were recognized as specific markers for EDMB-PINACA and ADB-HEXINACA. The LC-qToF-MS method was successfully adapted to the University of Bologna, as tested with 5F-EDMB-PICA in vitro metabolites. Prevalence data showed that 5F-EDMB-PINACA and EDMB-PINACA were more prevalent than ADB-HEXINACA, but for a limited period. Conclusion. Due to undetectability of parent compounds in urines and to shared metabolites among structurally related compounds, the identification of specific urinary biomarkers as unequivocal proofs of SCRAs consumption remains challenging for forensic laboratories. Urinary biomarkers are necessary to monitor SCRAs abuse and prevalence data could help in establishing tailored strategies to prevent their spreading, highlighting the role for legal medicine as a service to public health.

Production Lot Detection through Synthetic Generated Data

Correctness of information gathered in production environments is an essential part of quality assurance processes in many industries, this task is often performed by human resources who visually take annotations in various steps of the production flow. Depending on the performed task the correlation between where exactly the information is gathered and what it represents is more than often lost in the process. The lack of labeled data places a great boundary on the application of deep neural networks aimed at object detection tasks, moreover supervised training of deep models requires a great amount of data to be available. Reaching an adequate large collection of labeled images through classic techniques of data annotations is an exhausting and costly task to perform, not always suitable for every scenario. A possible solution is to generate synthetic data that replicates the real one and use it to fine-tune a deep neural network trained on one or more source domains to a different target domain. The purpose of this thesis is to show a real case scenario where the provided data were both in great scarcity and missing the required annotations. Sequentially a possible approach is presented where synthetic data has been generated to address those issues while standing as a training base of deep neural networks for object detection, capable of working on images taken in production-like environments. Lastly, it compares performance on different types of synthetic data and convolutional neural networks used as backbones for the model.

Synthetic data generation for the assessment of antimicrobial resistance through machine learning

As a consequence of the diffusion of next generation sequencing techniques, metagenomics databases have become one of the most promising repositories of information about features and behavior of microorganisms. One of the subjects that can be studied from those data are bacteria populations. Next generation sequencing techniques allow to study the bacteria population within an environment by sampling genetic material directly from it, without the needing of culturing a similar population in vitro and observing its behavior. As a drawback, it is quite complex to extract information from those data and usually there is more than one way to do that; AMR is no exception. In this study we will discuss how the quantified AMR, which regards the genotype of the bacteria, can be related to the bacteria phenotype and its actual level of resistance against the specific substance. In order to have a quantitative information about bacteria genotype, we will evaluate the resistome from the read libraries, aligning them against CARD database. With those data, we will test various machine learning algorithms for predicting the bacteria phenotype. The samples that we exploit should resemble those that could be obtained from a natural context, but are actually produced by a read libraries simulation tool. In this way we are able to design the populations with bacteria of known genotype, so that we can relay on a secure ground truth for training and testing our algorithms.

Synthetic generation of amyloid PET images by non-linear dimensionality reduction inversion

Privacy issues and data scarcity in PET field call for efficient methods to expand datasets via synthetic generation of new data that cannot be traced back to real patients and that are also realistic. In this thesis, machine learning techniques were applied to 1001 amyloid-beta PET images, which had undergone a diagnosis of Alzheimer’s disease: the evaluations were 540 positive, 457 negative and 4 unknown. Isomap algorithm was used as a manifold learning method to reduce the dimensions of the PET dataset; a numerical scale-free interpolation method was applied to invert the dimensionality reduction map. The interpolant was tested on the PET images via LOOCV, where the removed images were compared with the reconstructed ones with the mean SSIM index (MSSIM = 0.76 ± 0.06). The effectiveness of this measure is questioned, since it indicated slightly higher performance for a method of comparison using PCA (MSSIM = 0.79 ± 0.06), which gave clearly poor quality reconstructed images with respect to those recovered by the numerical inverse mapping. Ten synthetic PET images were generated and, after having been mixed with ten originals, were sent to a team of clinicians for the visual assessment of their realism; no significant agreements were found either between clinicians and the true image labels or among the clinicians, meaning that original and synthetic images were indistinguishable. The future perspective of this thesis points to the improvement of the amyloid-beta PET research field by increasing available data, overcoming the constraints of data acquisition and privacy issues. Potential improvements can be achieved via refinements of the manifold learning and the inverse mapping stages during the PET image analysis, by exploring different combinations in the choice of algorithm parameters and by applying other non-linear dimensionality reduction algorithms. A final prospect of this work is the search for new methods to assess image reconstruction quality.