22455 research outputs found
Sort by
Development and optimization of novel cryogenic calorimeters for COSINUS
Get PDFUnderstanding the particle nature of dark matter remains one of the central challenges in modern physics. Despite extensive astrophysical evidence supporting its existence, direct experimental detection of such particles remains elusive. This thesis presents the systematic development, optimization, and characterization of cryogenic sodium iodide (NaI) calorimeters tailored for the COSINUS experiment, a direct dark matter detection experiment designed to independently verify the DAMA/LIBRA findings of a potential dark matter-induced modulation signal.
Central to this research is the remoTES design - a novel approach for cryogenic detectors, in which an absorber crystal is connected to a Transition Edge Sensor (TES) via a dedicated thermal gold link. This decoupling allows mass fabrication of TES devices on separate wafers, preserving the radiopurity of the absorber by eliminating additional surface treatments.
The first part of the thesis addresses some of the crucial challenges faced by scintillation-only NaI-based readout experiments currently operational or under planning and presents a detailed study of the energy-dependent quenching factors of NaI crystals - essential for accurately interpreting experimental results as electron recoils and nuclear recoils have dissimilar scintillation light yields. COSINUS utilizes a two-channel readout system based on TESs that allows for particle discrimination and in-situ determination of these quenching factors. It consists of ultrapure scintillating NaI crystals, read out using the remoTES design to measure the phonon signal of a particle interaction. A silicon detector surrounding the crystal is used to measure the light signal from the same particle interaction.
Subsequent chapters document the construction and commissioning of the low-background experimental facility at Laboratori Nazionali del Gran Sasso (LNGS), Italy, featuring a muon veto system, dedicated clean room, and a specialized dry dilution refrigerator tailored for COSINUS commissioned in the context of this work.
Initial demonstrations with standard absorbers, namely silicon and tellurium dioxide, validated the feasibility of remoTES technology. Performance bottlenecks introduced due to the introduction of the gold link were addressed in subsequent optimization studies.
In parallel to these optimizations, experimental tests on NaI were performed, beginning with above-ground measurements on NaI crystals, demonstrating successful event-by-event particle discrimination. These results were followed up with an underground measurement at LNGS, achieving dark matter exclusion limits close to the benchmark set by the DAMA/LIBRA experiment with only a limited exposure (11.6 g d).
Additional refinements included dedicated studies on the silicon-based light detector, feasibility investigations into evaporated gold films on NaI crystals, and systematic standardization of fabrication techniques, culminating in the optimized 4pi module design prepared for COSINUS Run 1
Advances in deep active learning and synergies with semi-supervision
Get PDFSupervised deep learning models have successfully enabled the automation of processes and the discovery of valuable insights within large datasets, exceeding the capabilities of humans in analyzing and managing the constantly growing volumes of data. However, the effectiveness of these models largely depends on the availability of sufficient high-quality annotated training data. While collecting unlabeled data is often possible with comparably low effort, labeling it is laborious, time-consuming, and costly. In many domains, such as medical or industrial applications, providing accurate annotations requires specialized expertise, which is both scarce and expensive. It is, therefore, essential to reduce the necessity for manual labeling wherever possible.
In this thesis, we address the challenge of insufficient and costly annotations. In particular, we contribute to the field of deep active learning. Unlike traditional approaches that passively rely on pre-labeled data, active learning employs an iterative process alternating between training and labeling. By utilizing the model to decide which instances are most useful for its learning process, the performance is enhanced with a smaller amount of labeled data. Semi-supervised learning is a related field dealing with limited labeled data, which aims to improve models by leveraging both labeled and unlabeled data. Our contributions include new methods and insights into active learning as well as its combination with semi-supervised learning to exploit the strength of both.
Modern deep active learning strategies typically combine model uncertainty with sample diversity to avoid labeling data with redundant information. However, ensuring diversity by calculating distances in learned representations is computationally expensive, particularly for complex, high-dimensional neural networks. Our first contributions address this limitation. We propose using the prediction probabilities to simultaneously select diverse and uncertain instances, substantially accelerating query selection and returning a qualitative query set. Our method proves effective for both tabular and image classification, being superior to competitors in label and time efficiency.
Our next contribution focuses on active learning for node classification. The edges in a graph provide valuable insights into both the importance of individual nodes and the overall graph structure. Hence, it is essential to consider them when actively selecting the most useful instances for labeling. In our work, we introduce a novel active learning method for node classification that leverages diffusion-based graph heuristics in multiple ways for graph learning as well as actively querying nodes for labeling. In contrast to existing methods, our approach demonstrates robust performance across diverse datasets and consistently surpasses random sampling. Moreover, due to pre-computations, it is faster than competitors.
Finally, we turn our attention to the task of image classification with a particular focus on the combination of techniques from semi-supervised learning and active learning. Our first contribution in this domain proposes a novel active pseudo-labeling approach. We show that false pseudo-labels often occur during the initial iterations where label information is particularly sparse, resulting in long-term negative effects due to confirmation bias. To mitigate this, we propose a solution to refine the pseudo-labels produced by a model based on their consistency with predictions of a second model, considerably improving prediction accuracy.
In our last contribution, we analyze the effects of confirmation bias in semi-supervised learning when faced with datasets comprising challenging characteristics as they appear frequently in real-world data. In particular, we consider a high imbalance within and between classes as well as a high similarity between classes. We demonstrate the limitations of semi-supervised methods in overcoming confirmation bias when the data is randomly and passively labeled. By choosing better data samples through active learning, we discuss how confirmation bias can be mitigated, showcasing the potential of combining semi-supervised learning and active learning in the presence of common real-world data challenges
CRISPR/Cas9-mediated gene editing to analyze the impact of NOTCH1 mutations and their potential as a therapeutic target in mantle cell lymphoma
Get PDFMantle cell lymphoma (MCL) is a rare subtype of Non-Hodgkin Lymphoma representing about 5-7% of all Non-Hodgkin lymphomas in Western Europe. Although high initial response rates can be achieved with current standard therapy, early relapses and rapid disease progression determine the clinical course of most MCL patients and prognosis is still poor with an overall survival of only 3-5 years.
NOTCH1 gene mutations occur in 5-10% of mantle cell lymphoma (MCL) and are associated with significantly lower survival rates. The majority of these mutations lead to a truncation of the PEST domain, resulting in overactivity of the Notch1-signalling pathway. However, functional relevance of NOTCH1 mutations and its potential as a specific therapeutic target is not fully elucidated.
In this study, the CRISPR/Cas9 method was used to modify the PEST domain of the NOTCH1 gene in the MCL cell lines Mino and Jeko-1. The aim was to establish genetically identical cell clones that differ by a single mutation in the PEST domain of their NOTCH1 gene. These clones could further be used to perform experiments analyzing
NOTCH1 specific functions that are not confounded by intercellular differences. Additionally, the efficiency of drugs and antibodies targeting NOTCH1 signaling could be tested on the generated cell clones.
In the Mino cell line, we attempted to repair the point mutation in the PEST domain of the gene with a homology directed repair (HDR) template, whereas in Jeko-1 cells, harboring the wildtype sequence of the gene, we aimed to introduce a point mutation in the PEST domain through non-homologous end joining (NHEJ).
For both cell lines, the introduction of the guide RNA into the CRISPR/Cas9 backbone was successful and an electroporation program was established. However, in the Mino cell line, the repair of the mutation could not be achieved as the CRISPR/Cas9 construct disrupted the growth of the cell clones. Nevertheless, a mutation was successfully introduced into the NOTCH1 gene of the Jeko-1 cells and genetically stable Jeko-1 cell clones harboring a point mutation in the PEST domain of the NOTCH1 gene were created. As assessed by Western Blot analysis, NOTCH1-mutated clones expressed a shorter Notch1 protein due to the mutation in the PEST domain leading to a truncated protein with enhanced stability upon stimulation with DLL4.Das Mantelzelllymphom ist eine seltene Untergruppe der Non-Hodgkin Lymphome, die etwa 5-7% aller Non-Hodgkin Lymphome in Westeuropa darstellt. Inzwischen kann durch eine Behandlung nach den aktuellen Therapiestandards initial ein hohes Behandlungsansprechen erzielt werden, jedoch ist der klinische Verlauf ebenfalls durch eine hohe Frührezidivrate und einen aggressiven Verlauf geprägt. Die Prognose von Patienten mit Mantelzelllymphom ist immer noch schlecht bei einem Gesamtüberleben von nur 3-5 Jahren.
NOTCH1 Genmutationen kommen in 5-10% der Mantelzelllymphome vor und sind mit einem signifikant reduzierten Gesamtüberleben assoziiert. Diese Mutationen führen meistens zu einer Verkürzung der PEST-Domäne, was zu einer Überaktivierung des Notch1-Signalweges führt. Die funktionale Bedeutung von Notch1 Mutationen und ihre Bedeutung als Therapieansatz ist noch nicht vollständig erforscht.
In dieser Arbeit wurde mithilfe der CRISPR/Cas9 Methode die PEST-Domäne des NOTCH1 Gens in den Mantelzelllymphomzelllinien Mino und Jeko-1 verändert. Ziel war es, genetisch identische Zellreihen zu etablieren, die sich ausschließlich durch eine Veränderung in der PEST-Domäne des NOTCH1 Gens unterscheiden. Diese CRISPR/Cas9 veränderten Zellen können eingesetzt werden, um die Auswirkungen einer NOTCH1 Mutation durch Experimente zu charakterisieren, ohne die Ergebnisse durch interzelluläre Differenzen zu verfälschen. Zusätzlich kann die Wirksamkeit von NOTCH1 Inhibitoren und monoklonalen NOTCH1 Antikörpern an den hergestellten Zellklonen getestet werden.
In der Mino Zelllinie war das Ziel, CRISPR/Cas9 basiert die Punktmutation in der PEST-Domäne des NOTCH1 Gens mithilfe homologer Rekombinationsmechanismen (HDR) zu reparieren, wohingegen das Ziel in der Jeko-1 Zelllinie darin bestand, eine Mutation mithilfe nicht-homologer Endverknüpfung (NHEJ) in das wildtypische NOTCH1 Gen einzufügen.
Für beide Zelllinien ist es in dieser Arbeit gelungen, passende Guides in ein CRISPR/Cas9 Grundgerüst einzubringen. Außerdem konnte für die anschließende Transfektion mit dem CRISPR/Cas9 Konstrukt erfolgreiche Transfektionsprogramme für diese Zelllinien etabliert werden. Leider konnte in der Mino Zelllinie keine Reparatur der Mutation erzielt werden, da das CRISPR/Cas9 Konstrukt das Zellwachstum negativ beeinflusst. In den Jeko-1 Zelllinien konnte allerdings erfolgreich eine Mutation in die PEST-Domäne des NOTCH1 Gens eingefügt werden, sodass genetisch stabile Jeko-1 Klone mit einer potenten NOTCH1 Mutation hergestellt werden konnten. In der Westernblotanalyse zeigten die NOTCH1-mutierten Klone ein verkürztes Notch1 Protein aufgrund der verkürzten PEST-Domäne mit erhöhter Stabilität nach DLL4 Stimulation
Analyse des durch Glukose organisierten Genexpressionsnetzwerks in NPYR-/MC4R-positiven, hypothalamischen Neuronen
Get PDFCharacterization of BeO optically stimulated luminescence dosimeters for external and medical applications
Get PDFtDCS modulation of neurochemical and cognitive functions in healthy and depressive individuals
Get PDFBispecific CD33-TIM3 CAR T cells enhance specificity while maintaining efficacy against AML
Get PDFBackground: CD33, a specific antigen prevalent on myeloid cells, is found in 88% of samples from patients with acute myeloid leukemia (AML), highlighting its potential as an immunotherapeutic target. Despite numerous ongoing clinical trials targeting AML, challenges related to CAR T-cell specificity—termed "on-target off-leukemia"—continue to present significant hurdles. In response, this research focuses on the development of dual CAR T cells that concurrently target CD33 and TIM3. TIM3 has been identified on leukemic stem cells (LSCs) and is absent on normal hematopoietic cells. Additionally, the role of TIM3 in impeding immune regulation underscores its suitability as a secondary target in AML immunotherapy strategies. The objective of this investigation is to construct and evaluate dual CAR T cells targeting CD33 and TIM3 to enhance the specificity and maintain the anti-leukemic efficacy of these cells.
Methods: This study initiated by producing anti-TIM3 antibodies using hybridoma technology. These antibodies were then validated for their specificity through enzyme-linked immunosorbent assay (ELISA) and fluorescence-activated cell sorting (FACS). DNA sequencing was performed of anti-TIM3 single-chain variable fragments (scFv) derived from these hybridomas. The structural interactions of CD33 and TIM3 scFvs with each antigen were modeled using AlphaFold2, referencing the TIM3 structure (PDB ID: 5F71) and CD33 structure (PDB ID: 6D48). The CD33 scFv was derived from gemtuzumab ozogamicin (clone hP67.6) and both incorporated along with costimulatory domains (CD28 or 4-1BB) into a pMP71 vector. A retroviral production system utilizing 293Vec-GALV and RD114 cells facilitated the production of the retroviruses. In vitro, the cytotoxicity of the CAR-T cells was tested against both wild-type and TIM3-transduced AML cell lines (THP-1 and OCI-AML3) through co-culture assays by multiparametric flow cytometry (MPFC). Cytokine release assays (CBA) were used to measure IFN-γ and IL-2 secretion, and cell-target avidity was analyzed using a Z-Movie analyzer. Off-target effect was monitored through colony-forming unit (CFU) assay on CD34+ cells from healthy donors after 14 days. For long-term efficacy assessments, CAR-T cells were periodically re-stimulated every four days by co-culturing with irradiated TIM3 expressing SKM-1 cells at a 1:1 effector-to-target (E:T) ratio over 24 days. During these intervals, analyses of CAR-T cell proliferation, checkpoint marker expression, and T cell subset differentiation were conducted via MPFC.
Conclusion: This study successfully generated dual-target CAR T cells utilizing both "AND" and "OR" gating strategies to target CD33 and TIM3, enhancing their efficacy to AML cells in vitro. The findings demonstrated that these CAR T cells exhibit superior binding avidity and cytotoxic capabilities towards cells expressing both CD33 and TIM3 antigens, as opposed to targeting a single antigen. Notably, the split CAR T cell approach effectively eradicated CD33+TIM3+ cell lines and primary AML cells, while minimizing impact on healthy hematopoietic cells. The implications of these results suggest that dual CAR T cell configurations might serve as potential bridging therapies before hematopoietic stem cell transplantation. The split CAR T cell particularly offered a potential safer alternative, possibly obviating the need for allogeneic stem cell transplantation due to on-target-off-leukemia toxicity. These advancements have the potential to markedly alter the therapeutic landscape for AML, offering more precise, effective, and safer options for treatment
