Die chronische Nierenerkrankung im Kindes- und Jugendalter ist meist Folge angeborener Fehlbildungen oder seltener, autoimmunologischer Erkrankungen. Demzufolge sind Inzidenz und Prävalenz deutlich geringer als bei erwachsenen Patient:innen, bei denen die Häufigkeit der CKD als Folge jahrelanger Exposition gegenüber chronischen Erkrankungen und Umweltfaktoren global zunimmt. Dennoch hat die Diagnose und die häufig rasche Progression der CKD bei Kindern und Jugendlichen erhebliche Auswirkungen insbesondere auf das kardiovaskuläre System, das Wachstum sowie den Knochen- und Mineralstoffwechsel, was zu einem erheblichen Verlust an Lebensqualität und einer bis zu 1000-fach erhöhten Mortalität führt. Neue Behandlungsoptionen im Erwachsenenbereich, wie beispielsweise SGLT2-I, adressieren primär die Pathophysiologie multimorbider, internistischer Patient:innen, sind jedoch nicht systematisch bei Kindern und Jugendlichen untersucht. Daher ist es wichtig, spezifische therapeutische Ansätze für Kinder mit CKD zu entwickeln, die auf die besonderen phänotypischen Merkmale dieser Patientengruppe abzielen. Hierbei rückt das Darmmikrobiom in den Fokus der Wissenschaft, insbesondere als Schnittstelle zwischen Umwelt und Wirt und geprägt durch eine facettenreiche Interaktion und Modulation zahlreicher, vorwiegend immun-metabolischer Mechanismen. Im Rahmen dieser Arbeit wurden die mikrobiell-produzierten Urämietoxine Indoxylsulfat und p-Cresylsulfat in der multinationalen 4C Studie im Serum von Kindern und Jugendlichen gemessen. Es zeigte sich eine signifikante und unabhängige Assoziation mit frühen kardiovaskulären Veränderungen sowie der Progression der CKD, was die Bedeutung dieser Toxine als Risikofaktor für CKD-assoziierte Komorbiditäten unterstreicht. Hierauf aufbauend wurden in einer neu-rekrutierten, prospektiven Kohorte Veränderungen im Darmmikrobiom bei Kindern mit CKD beschrieben und mit metabolischen und pro-inflammatorischen Phänotypen verknüpft. Hierdurch wurde erstmalig in einer pädiatrischen Kohorte gezeigt, dass Veränderungen im Darmmikrobiom inflammatorische Signalwege aktivieren und somit relevant zum erhöhten Risiko kardiovaskulärer Erkrankungen beitragen können. Die Akkumulation mikrobiell-produzierter Toxine ist zum Teil Folge der verminderten renalen Elimination bei CKD. Durch die Entwicklung und Anwendung stuhlbasierter Tests konnte zudem eine vermehrte Häufigkeit der Schlüsselenzyme für die bakterielle Synthese dieser Urämietoxine im Darm nachgewiesen werden, was die additive und kausale Rolle einer vermehrten bakteriellen Produktion dieser Toxine unterstreicht. Als wichtigen Mediator kardiovaskulärer Erkrankungen konnten wir eine erhöhte Freisetzung von extrazellulären Vesikeln bei CKD Patient:innen nachweisen, was zu einer direkten Schädigung endothelialer Zellen führt und damit ein möglicher Angriffspunkt für neue Therapien darstellt. Schließlich wurde in einer Kohorte nierentransplantierter Patient:innen nachgewiesen, dass die Persistenz CKD-typischer, mikrobieller Veränderungen im Darm mit einem erhöhten Risiko für Abstoßungsreaktionen nach Nierentransplantation einhergeht. Zusammengefasst beschreiben diese Arbeiten die Rolle des Darmmikrobioms und dessen Interaktion mit dem Immunsystem des Wirts, vorwiegend vermittelt durch mikrobiell-produzierte Metabolite, welche bei CKD akkumulieren. Diese Veränderungen sind insbesondere relevant in Bezug auf die Progression der CKD und das frühe Auftreten kardiovaskulärer Erkrankungen. Unsere Ergebnisse sind ein erster Anhaltspunkt für die Entwicklung Mikrobiom-gerichteter Therapien, die im Rahmen einer personalisierten Anwendung das Auftreten und die Progression CKD-assoziierter Komorbiditäten insbesondere bei Kindern und Jugendlichen abschwächen und damit das Outcome dieser vulnerablen Population verbessern könnten.

Spin-crossover molecules (SCMs) are coordination compounds based on transition metals which exhibit bistable electronic configurations (high-spin and low-spin states) that can be tuned through external stimuli, such as temperature, pressure or light. This property makes them highly attractive for applications in molecular electronics, spintronics, and energy conversion. In this work, the empirical values that describe the thermodynamical as well as photochemical interactions of a set of different tridentate-ligand SCMs deposited as thin films on highly oriented pyrolitic graphite (HOPG) (and also in the bulk phase) were investigated. For this, two different characterization methods, namely x-ray absorption spectroscopy (XAS) and differential reflectance spectroscopy (DRS) were used. The parameters obtained from the results can be used to describe the different arrangements of SCM in both their high-spin and low-spin state configurations.

A fundamental understanding of catalyst dynamics is critical to improve the efficiency of catalytic processes, on which modern society relies heavily. Their application ranges from the food industry to energy conversion and storage processes. Catalysts play a key role in the transition from fossil resources to renewable energy sources, as they significantly reduce the activation energy to drive reactions such as the conversion of CO2 into energy-dense feedstock chemicals or fuels, but are highly dynamic systems that evolve under operating conditions. To follow their transformations, suitable characterization methods must be employed and further developed to explore the catalyst properties on different length- and time-scales. In this regard, X-ray absorption spectroscopy (XAS) is an invaluable tool to characterize the atomistic structure and bulk electronic properties of a catalyst. This element-sensitive method allows one to track time-resolved changes in the catalyst’s local structure in complex reaction environments. Synchrotron radiation facilities offer increasing availability of operando XAS with sub-second time resolution, however, this leads to large amounts of data and renders inefficient the conventional XAS data analysis workflows. Furthermore, the existing access modes for synchrotron-based experiments are not compatible with requirements for such long-term studies, which play an important role in catalysis research, since the catalyst stability is key for their practical application. Laboratory-based operando XAS (labXAS) experiments are an attractive solution for this issue, but the methodology for such experiments has not been fully developed so far. Further, the analysis of synchrotron- and laboratory-based operando XAS data, remains challenging, especially when tracking unique heterogeneous structural motifs in multi-element systems or low amounts of minority species. Data analysis methods need to be advanced to disentangle these contributions under different conditions. In this work, I have developed methodologies for advanced operando XAS data acquisition and analysis on the example of mono- and bimetallic Cu-based materials employed for the electrocatalytic conversion of CO2 to value-added products under industrially relevant conditions, including high current densities or pulsed potentials. By employing time-resolved operando XAS, combined with supervised and unsupervised machine learning methods, I studied the evolution of the catalyst’s chemical state and atomistic structures under these conditions. For bimetallic catalysts, I probed the influence of the secondary metal on the Cu redox properties. Finally, I developed a new setup and methodology combining electrocatalysis with labXAS, allowing time-resolved tracking of the catalyst transformations and stability over several days. To conclude, this thesis aims to advance the operando XAS methodology in the field of (electro-)catalysis and contributes to the progress in unraveling the dynamics of catalysts on all time scales, from catalyst restructuring processes occurring within seconds, to sluggish changes that require several hours or days to take place.

As a minor ingredient, dried culinary spices and herbs are often not the focus of foodborne outbreak investigations. However, studies assessing their microbial quality have shown that they can be highly contaminated with microorganisms, including pathogenic species. Among these, Salmonella spp., Bacillus cereus, and Clostridium perfringens are the foodborne pathogens of particular concern with respect to dried spices and herbs. Staphylococcus aureus is a prominent foodborne pathogen known for its ability to survive for extended periods under low–water–activity conditions; nevertheless, it has only rarely been detected in spices and herbs. This thesis therefore investigated the tenacity of S. aureus in dried spices and herbs by applying artificial contamination methods to evaluate both methodological effects – representing different contamination routes (Publication 1) – and matrix effects (Publication 2) on the survival of S. aureus. The aim was to determine whether S. aureus is inherently unable to survive in dried spices and herbs, or whether contamination events simply occur infrequently. Furthermore, the potential antimicrobial impact of spices and herbs on the recovery of S. aureus by culture-dependent methods was assessed to rule out the possibility of underestimating contamination levels. The results showed that antimicrobial compounds present in spices and herbs did not interfere with culture-based enumeration when plating was performed immediately after sample dilution. Moreover, S. aureus was able to survive for several months in dried spices and herbs, depending on the specific matrix and the mode of contamination. These findings indicate that the low prevalence of S. aureus in spices and herbs is primarily due to infrequent contamination events resulting from poor hygiene standards and improper handling during cultivation, harvesting, and processing. All raw ingredients of a food product – from major components such as meat to minor ones such as spices – typically undergo several processing steps before reaching the final product. As a result, accompanying microorganisms are exposed to changing environmental conditions. Unfavorable growth conditions can induce a viable but non-culturable state (VBNC), in which bacteria remain metabolically active but are no longer able to grow under standard laboratory conditions. Consequently, pathogen detection using selective media, as in conventional culturedependent methods, may underestimate the number of viable cells and thus pose a potential risk to consumer health. viii Polymerase chain reaction (PCR) is a highly specific and sensitive molecular method; however, it cannot distinguish between DNA from live and dead cells. To overcome this limitation, viability dyes have been introduced as a pre-treatment prior to PCR to allow selective detection of viable cells. Despite the widespread use and ongoing optimization of the viability PCR (vPCR) methodology, avoiding false-positive results remains a major challenge. This thesis therefore aimed to further optimize vPCR – building on previous methodological improvements – to exclude DNA from dead cells to the greatest extent possible while minimizing the impact on live cells. The optimized approaches were evaluated using Salmonella and S. aureus as model organisms representing Gram-negative and Gram-positive bacteria, respectively. By applying PEMAX – a combination of propidium monoazide (PMA) and ethidium monoazide (EMA) – along with double photoactivation and double tube-change steps, DNA from up to 5.0 × 10⁷ heat-killed Salmonella Typhimurium cells per 200 μl reaction volume could be suppressed below the detection limit, enabling accurate quantification of serial dilutions of live Salmonella cells in the presence of high dead-cell counts (Publication 3). However, this protocol could not be directly transferred to the detection of S. aureus due to the uptake of EMA by viable S. aureus cells. Therefore, PMA alone was applied in a doubletreatment procedure at low concentration, combined with a tube-change step, to achieve a comparable effect (Publication 4). Both optimized protocols were successfully applied to food samples for detecting viable cells after enrichment in the presence of approximately 5 × 10⁶ dead cells/ml. At higher dead-cell loads, however, neither protocol could reliably suppress the dead-cell signal across all tested matrices. Overall, vPCR represents a valuable extension of conventional PCR and can be applied as a rapid pre-screening method for the selective detection of viable pathogens in food samples.

Bipolar disorders (BD) are characterized by recurring episodes that, in addition to affective fluctuations, are also associated with cognitive impairments. While cognitive deficits in acute episodes are well-documented, it remains unclear to what extent they persist in remitted phases (rBD) and how they align with subjective perceptions of cognitive functioning. Furthermore, little is known about the influence of autobiographical memory processes on cognitive performance in BD. The aim of this dissertation is to systematically analyze these two aspects—subjectively perceived cognitive functioning and autobiographical memory—in relation to objectively measured cognitive functions in rBD. In the first study, the cognitive performance of individuals with rBD (n = 26) was compared to that of a healthy control group (HC; n = 24). In addition to a neuropsychological test battery assessing objective cognitive parameters, a self-report questionnaire was used to evaluate subjective perceptions of cognitive deficits. Contrary to the initial hypothesis, no significant differences were found between the groups in executive functions, attention, or verbal memory. However, rBD participants reported greater subjective cognitive impairments and lower self-esteem compared to the HC. In the rBD group, subjective impairments were particularly associated with the number of past depressive episodes and subclinical depressive symptoms, whereas self-esteem was the primary predictor in the HC. These findings suggest that subjective cognitive deficits are not so much influenced by objective deficits but rather by psychological factors such as self-esteem and depressive symptoms. The second study examined autobiographical memory (AM) and its relationship to cognitive performance. A group of individuals with rBD (n = 20) was compared to HC (n = 22) regarding executive functions, verbal memory, and attention. An extended version of the Autobiographical Memory Test (AMT) was used, which included positive, negative, and rejection-related cue words. The results showed no significant differences between the rBD and HC groups in the number of specific memories. However, both groups generated fewer specific memories in response to rejection-related and positive cue words compared to negative cue words. Additionally, a strong correlation was found between higher AM specificity and better executive function as well as verbal memory performance. Regression analyses revealed that these cognitive variables accounted for approximately 50% of the variance in AM specificity. I In summary, these studies highlight that cognitive deficits in rBD do not occur uniformly and must be considered in a differentiated manner. While objective deficits appear to be less pronounced, psychological factors such as self-esteem and subdepressive symptoms significantly influence the subjective perception of cognitive deficiency. Furthermore, AM specificity is closely related to cognitive functions, particularly executive functions and verbal memory. These findings have clinical relevance: subjective cognitive complaints should not be interpreted in isolation but rather in the context of psychological influencing factors. Longitudinal studies could help clarify how these mechanisms change across different phases of the disorder.

The interplay between the charge and spin degrees of freedom of electrons in solids is of fundamental interest in understanding transport in magnetic materials and has useful applications in spintronic devices. Due to its potential in fast information processing, the focus of research has shifted towards ultrafast timescales. This requires describing and understanding charge, spin, and heat transport — all of which are closely linked — in normal metals, topological insulators, magnets, and their interfaces in heterostructures, up to high driving frequencies. This thesis focuses on nonlocal charge transport mediated by magnons in heterostructures with magnetic insulators. The results in this thesis are time-dependent, and the current response at high frequencies is often strikingly different from the dc limit. The first system we consider is an N|F|N trilayer comprised of two normal metals (N) and a ferromagnetic insulator (F). When a time-dependent electric field is applied to one of the normal metals, a charge current and temperature change are measured in the other metal layer. This results from charge-to-spin conversion (and the inverse effect) in both metal layers, as well as coupled spin and heat transport in the ferromagnet and through the interfaces. We calculate the nonlocal conductivity and its dependence on the ferromagnet’s thickness and its magnetic state for driving frequencies up to the terahertz (THz) regime. Furthermore, we find that magnon cooling of electrons in the normal metal is significantly more efficient in a nonlocal setup. We also calculate and compare several nonlinear, unidirectional magnetoelectric effects, e.g. Joule heating or the spin-torque diode effect. While these effects have been measured at low frequencies, it has been difficult to differentiate them experimentally due to their similar experimental signatures. The theoretical model presented here provides a framework for characterizing these effects based on how they scale with ferromagnet thickness, magnetic field, and frequency. Antiferromagnets are another promising platform for spintronic devices operating at high frequencies. Inspired by THz spectroscopy measurements of magnon transport through nickel oxide, we present a ballistic transport model that accounts for the boundary conditions at metal interfaces and facilitates the calculation of a time trace of the current driven by an initial temperature increase induced by a femtosecond pump pulse. The final system we consider is a magnet coupled to a quantum spin-Hall insulator. This topological insulator hosts gapless helical edge states that are spin-momentum locked. Local coupling to a magnet opens a gap in the edge spectrum, leading to the backscattering of electrons accompanied by a spin flip. This transfer of angular momentum to the easy-plane magnet lifts its magnetization out of the plane, leading to precession — and spin pumping — of the magnetization. While the pumped current compensates the reflected electrons for a dc bias voltage and the edge becomes perfectly conducting despite the energy gap, we find that the finite response time of the magnetization to an ac perturbation results in a nontrivial admittance and a frequency filtering of the noise.

This dissertation explores the association between a relationship bank and the earnings quality of private and public firms. The literature so far suggests that there is less need for a borrowing firm to disclose extensive or high-quality information if the firm engages in relationship banking since the bank can rely on private (soft) information to assess the true financial situation of the firm. However, this argument fails to differentiate between the different disclosure incentives of private and public firms and neglects the fact that the value of private information for relationship banks can change. There is mixed empirical evidence with regard to the association of relationship banking and earnings quality. Some literature finds that a closer relationship between a bank and a borrowing firm leads to lower earnings quality. Low earnings quality can be in the best interest of a borrowing firm if it tries to conceal its true performance by obfuscating their financial reports to keep competitors from entering profitable market segments. The relationship bank knows the true creditworthiness of the firm due to access to private information and may be able to extract information rents from the borrowing firm. Thus, it can keep a competitive advantage towards other banks. Other studies, however, suggest that relationship banks are able to improve the borrowing firm’s earnings quality because of comparative advantages in monitoring. In my analyses, I choose heterogeneous settings (private vs. public firms) as well as two different countries (Germany and Japan) with differences in national legislation and bank competition that may help to explain mixed results in the empirical literature. However, the reasons for these varying results may run deeper yet. It is the aim of this dissertation to further explore this gap in the empirical literature and to determine additional factors that influence the association between relationship banking and earnings quality of private and public firms. One important factor that is scarcely addressed in the literature is how much value private information holds for a relationship bank. An underlying assumption for relationship banking is that the bank is able to collect private information and use it to assess the true creditworthiness of the borrowing firm. However, there are different scenarios where (1) banks are legally not allowed to act upon private information (e.g. in case of insider-trading regulation), (2) the different bank types in the heterogeneous group of relationship banks have different objectives, organizational and client structures that obstruct the collection of private information or decrease extractable information rents, or (3) the financial situation of the borrowing firm changes their disclosure incentives so that opaque financial reporting is not in their best interest anymore. Therefore, this dissertation consists of three empirical analyses that each address one of the scenarios above.

Catalysis is a cornerstone of modern chemistry, influencing processes from biological enzyme reactions to industrial chemical production. Given its significant contribution to the global economy, improving catalytic systems, even marginally, holds immense value. This thesis explores advanced computational techniques to address critical challenges in heterogeneous catalysis, focusing on rate-determining steps and uncertainty quantification. We introduce a novel sensitivity analysis framework based on the Cramér-von Mises (CvM) distance, coupled with kinetic Monte Carlo (kMC) simulations, to identify and analyze influ- ential reactions in catalytic networks. The proposed methodology overcomes some of the major limitations of established variance- and gradient-based approaches, offering robustness and computational efficiency in handling uncertainties in input parameters such as reaction energies and barriers. The numerical implementation enables the use of classical quadrature rules, allowing for a broad class of quadrature schemes to be employed. This approach can be applied to both deterministic and stochastic models, with theoretical derivations establishing the conditions under which the quadrature-based method converges to the proper value for stochastic models using just a single sample per quadrature point. The effectiveness of the CvM-based sensitivity analysis is validated through analytical models and two catalytic case studies: CO oxidation on RuO2 and CO2 reduction on a copper-based surface. These studies demonstrate fast convergence and resilience to noise in sampled data, paving the way for reliable predictions of turnover frequencies and surface coverages. Despite inherent uncertainties in Density Functional Theory (DFT) calculations, this thesis highlights the potential to derive actionable insights through advanced sensitivity methods. The work provides a foundation for future investigations into the robustness and theoretical properties of the CvM-based approach under various computational conditions. This work contributes significantly to the computational study of catalysis, providing a robust framework for sensitivity analysis, optimizing catalytic systems, and bridging the gap between atomic-scale modeling and experimental validation.

With the emergence of the Internet of Things (IoT), embedded systemsare more connected than ever. Software reconfiguration, whileubiquitous on desktop and server systems, is fledgling on these IoTdevices. The challenges associated with the constrained nature of thedevices and network links hamper the design of solutions.Any communication with networked devices for delivering adjustmentsto the device must at least be authenticated with digital signatures. Ipresent benchmarks on both post-quantum and pre-quantum digitalsignatures and show the trade-offs required when deploying one ofthese primitives on the devices. I show that while post-quantum-security is indeed doable on these devices, large differences existbetween the different cryptographic primitives and their implementa-tion.Within the heterogeneous space that is the Internet of Things, a genericmechanism to deliver firmware updates to a large number of devices isrequired. I survey open standards, which provide generic buildingblocks for secure firmware updates on constrained devices. I publishedan open source implementation of SUIT able to deliver arbitraryfirmware updates for the RIOT operating system. I show that firmwareupdates secured with SUIT can be achieved on a large variety ofdevices, including the smallest of microcontrollers.Small virtual machines hosted on embedded devices can act as sandboxfor the device, to isolate and reconfigure part of the application code. Ipresent the design of a minimal VM, rBPF, implemented and studiedexperimentally. I compare the performance to an approach hostinghigh-level logic in an embedded WebAssembly virtual machine. I showthat rBPF can be deployed with minimal impact on the firmware size.The capability to run applications inside a VM can be leveraged tobundle applications into small software components. I introduceFemto-Containers, which enables Functions-as-a-Service capabilitiesembedded on heterogeneous low-power IoT hardware. I show thatFemto-Containers significantly improve state of the art, by provid-ing FaaS-like capabilities with strong security guarantees on suchmicrocontrollers.I design and present Cubedate, a framework achieving strong securityguarantees and low overhead, for continuous deployment of softwareover the air on multi-tenant CubeSat. With the implementation here Ishow how SUIT and the other components presented in this work canbe used to maintain and adapt the logic running as CubeSat payload.

In nature, tissues are anisotropic, but most biomaterials used in regenerative medicine applications are not. Patterned biomaterials offer the opportunity to mimic spatially segregating biophysical and biochemical properties found in nature. Engineering such properties allows to study cell-matrix interactions in anisotropic matrices in great detail. The tunable artificial niche this research used is an alginate-based hydrogel with dual crosslinking: Diels-Alder covalent crosslinking (norbornene-tetrazine, non-degradable) and UV-mediated peptide crosslinking (matrix metalloprotease sensitive peptide, enzymatically degradable). This material platform allowed us to generate biomaterials with patterns in diverse multiple biophysical properties to stablish the potential application of this materials for guiding cell behavior in 3D matrices. First, we developed a hydrogel with spatial patterns in degradation and stiffness, featuring two distinct sections: a soft, non-degradable phase (Soft-NoDeg) and a stiff, degradable phase (Stiff-Deg). Rheology was used to characterize single-phase materials, while surface micro-indentation assessed the patterned hydrogels. Over time, the materials exhibited emerging stiffness and degradability patterns. Encapsulated 3D mouse embryonic fibroblasts (MEFs) were analyzed for anisotropic cell-matrix interactions using a novel image-based quantification tool. Live/dead staining revealed no differences in viability but distinct proliferation patterns, with higher cell numbers in Stiff-Deg materials by day 14. Cell morphology analysis showed larger projected cell areas in Soft-NoDeg materials at day 1, due to the difference in stiffness. By day 14, the enzymatic degradation of the Stiff-Deg materials allowed an increase in cell area comparable to Soft-NoDeg, alongside a decrease in circularity and an increase in filopodia number and length significantly higher compared to Soft-noDeg. This demonstrated that stiffness and degradability patterns can control anisotropic cell responses in 3D hydrogels, effectively quantified through image-based approaches. Building on these findings, the next phase compared anisotropic hydrogels with degradation patterns to single-phase materials (Deg and NoDeg), using primary human mesenchymal stem cells (hMSCs). The Deg areas of anisotropic hydrogels showed enhanced cell spreading, collective cell alignment, mechanotransduction, and osteogenic differentiation. Anisotropic hydrogels lead also to spatially patterned osteogenic differentiation, with mid-stage markers (Osteocalcin) expressed by day 14, unlike single-phase Deg materials that showed only early markers (Osterix). Mechanosensing analysis revealed enhanced YAP nuclear translocation and expression in Deg regions of patterned materials, an effect absent in MMP-scrambled peptide or non-RGD-modified hydrogels. These results underscore the potential of degradation patterns in guiding hMSC behavior and boosting osteogenic differentiation by an enhanced mechanotransduction due to the material’s anisotropy. Finally, we focused on interphases for endothelial cell guidance by patterned viscoelasticity and collagen network. To achieve the patterning of these two biophysical properties, we developed a covalently crosslinked alginate and collagen interpenetrating polymer network (IPN). The degree of alginate crosslinking influenced collagen network formation and matrix viscoelasticity. Norbornene (N)- or tetrazine (T)-modified alginate enabled crosslinking via UV (degradable) or non-UV (non-degradable) methods. Confocal reflectance and rheology confirmed that reduced alginate crosslinking improved collagen network formation and viscoelastic properties, while highly crosslinked matrices exhibited elastic behavior and lacked collagen fibers. Photopatterning enabled the creation of high- and low-crosslinked areas within a single matrix, confirmed through confocal reflectance and scanning electron microscopy. These patterns influenced collagen architecture and viscoelasticity, facilitating guided endothelial cell migration and collective invasion. Our findings highlight the importance of a degradable, viscoelastic, collagen-rich matrix in promoting endothelial cell invasion. In summary, this research demonstrates the versatility and potential of patterned biomaterials in guiding cell behaviors by mimicking the anisotropic properties of native tissues. By employing hydrogels with spatially defined stiffness, degradation, viscoelasticity, and collagen organization, we provided evidence that these properties can independently and collectively regulate key cellular processes such as mechanotransduction, alignment, differentiation, and invasion. These findings not only deepen our understanding of cell-matrix interactions in anisotropic environments but also establish a foundation for the design of advanced biomaterials for regenerative medicine applications. Future work will focus on refining these material platforms and exploring their applicability in more complex biological models and therapeutic settings.

Biomolecular condensates regulate essential cellular processes and are implicated in many diseases. Understanding how condensates form and function is key to uncovering how biomolecules organize into a functional cell. However, because condensates often form through non-stoichiometric interactions, classical stoichiometric models of molecular interactions are insufficient to explain the activities of condensates, motivating the development of the condensate microenvironment model. This model posits that condensates have properties emerging from the collective behavior of their constituent molecules, resulting in distinct physicochemical parameters within condensates that differ from the surrounding cellular environment. Notably, the emergent material properties, such as diffusivity and viscosity, are thought to influence condensate composition and function. However, current tools face four distinct but related challenges, limiting tests of the condensate microenvironment model: (1) difficulty in perturbing material properties; (2) limited ability to selectively target specific condensates; (3) difficulty capturing condensate dynamics in live cells; and (4) complexity in manipulating endogenous condensates. This thesis presents the development and application of a synthetic 17-amino-acid micropeptide, the killswitch, that perturbs the material properties of condensates in live cells. The killswitch works by immobilizing condensate components through phenylalanine-mediated self-association and can be targeted to diverse condensates using genetic or nanobody-based approaches. I first identified the molecular features that underlie killswitch function, then used the killswitch to perturb the material properties of nucleolar, oncogenic, and viral condensates in live cells. Immobilizing condensate components with the killswitch affected the abundance and mobility of untargeted condensate components, disrupting condensate function and leading to changes such as altered nucleolar protein composition and impaired adenoviral condensate organization. These results establish the killswitch as a powerful tool for perturbing condensate material properties and probing condensate microenvironments, enabling systematic investigation of how condensate dynamics, composition, and function are interrelated.

Weaning is still considered one of the most critical times in intensive pig farming. Due to the often still immature immune system, piglets often develop diarrhea which indicates microbial dysbiosis and intestinal inflammation. This condition is referred to as PWD and can cause huge economic losses for farms. Until 2006, antibiotics were often used to treat PWD and to help promote piglets’ growth during weaning. After this practice was banned throughout the European Union, the industry was faced with finding suitable alternatives. Pre- and probiotics are considered promising candidates. Studies have shown that probiotics can have positive effects on the health and performance of piglets by influencing the composition of the intestinal microbiome and promoting the abundance of beneficial bacterial species. Prebiotics in turn serve as a substrate for the intestinal microbiota promoting the colonization of potentially beneficial bacteria. Much research is conducted on the effectiveness of pre- and probiotics, but the results are often inconsistent and difficult to compare. There is no generalized approach regarding the application of pre- and probiotics. Depending on the probiotic strains and prebiotics used, potential effects highly depend on the dosage used, host-specific factors, genetic aspects and other environmental factors (e.g. husbandry and feeding).

The aim of this study was to develop a method to determine tailored combinations of pre- and probiotics to allow for targeted use in practice.

A novel ex vivo assay was established using a total of 60 fecal samples obtained from 20 German pig farms. Three commercially available probiotics (B. licheniformis/ B. subtilis, S. cerevisiae boulardii, E. faecium) and prebiotics (inulin, FOS, MOS) were selected alone or in combination to test their effectiveness to inhibit growth of three pathogenic strains (E. coli, C. difficile, C. perfringens) in fecal samples. Lag time was measured by using growth curves to estimate the effect of the feed additives on residual fitness of the pathogenic strains. Subsequently, the efficacy of the most promising combination was tested in an in vivo ETEC challenge trial. Forty piglets were weaned at 28 days of age and randomly assigned to one of five experimental treatment groups: Nonchallenged control (NC); challenged positive control (PC) challenged and vaccinated (CV), challenged and diet supplemented with pre- and probiotic mix (CM) and challenged, diet supplemented with pre- and probiotic mix and vaccinated (CMV). Piglets of CV and CMV were vaccinated parenterally prior to the trial at the age of 17 days. Piglets of PC, CV, CM and CMV were challenged five days after weaning with the ETEC strain (serotype O149:K91). The trial was conducted over a period of 29 days. Efficacy of dietary treatment against challenge was assessed by evaluation of growth performance, feed conversion and fecal score.

Using the novel ex vivo assay, farm-specific differences were identified regarding the pre- and probiotics used, as well as the pathogenic strains. FOS and its combination with probiotics reduced the residual fitness of the E. coli strain in some farms. In other farms FOS lead to increased fitness of the E. coli strain. Combinations of pre- and probiotics did not show additive effects but displayed farm dependent differences. The effects of pre- and probiotics on the residual fitness of the C. difficile strain were less pronounced. Distinct differences were found between the use of prebiotics alone and their combination with probiotics. During the preliminary assays, there had been insurmountable problems with the C. perfringens strain, which is why this strain was excluded from the subsequent assays.

The in vivo challenge resulted in a significant reduction of body weight of vaccinated piglets (p = 0.045) during the first week of the trial when compared to NC. It was also associated with an impaired gain to feed ratio (P = 0.012). Piglets of the pre- and probiotic group (CM) were able to maintain their body weight. Their ADG was not significantly different from the control groups. Between the 3rd and 4th week of the trial, no differences were observed between groups. Neither supplementation of pre- and probiotics, nor vaccination had a significant impact on fecal consistency or prevalence of diarrhea. No positive synergistic effect on performance or diarrhea was observed between vaccination and pre- and probiotic supplementation.

In conclusion, the study has demonstrated that the novel screening method represents a promising approach for determining farm-specific responses to pre- and probiotics to identify customized combinations for individual farms. Yet the effectiveness of the specially determined combination of pre- and probiotics was not confirmed by the challenge trial. Vaccination was also not convincing and rather resulted in reduced performance and increased diarrhea.

Based on our findings further research is warranted to verify potential synergistic effects of tailored combinations on a broader basis. An alternate study design and the use of different pre- and probiotics, as well as pathogenic bacteria may affect the intestinal response differently.

The progressing miniaturization of electrical components on a circuit inevitably leads to the research of components on a molecular level. Since the 1970s, when this new research field of molecular electronics was started, it has been advancing rapidly with the help of quantum chemical methods, such as the non-equilibrium Green's function formalism.

Molecular junctions are designed to fulfill the roles of classical electronic components. However, they can also exhibit special conducting properties due to quantum effects, e.g. quantum interference, which can lead to extremely low or high transmission. Molecular junctions can also be highly efficient spin filters. Since high electron mobility is very important in such systems, graphene nanoribbons are a popular base material for molecular electronic devices. Their electronic properties are highly tunable by their width, their edge shape and their functionalization through doping or other means.

In this thesis, the global and local transport properties of several different molecular devices are investigated with non-equilibrium Green's function methods. For this, a readily available Python-based tool for non-equilibrium Green's function from the atomic simulation environment program package is expanded on to allow the use of outputs from a wider range of quantum chemical software packages. First, a quinone functionality exhibiting destructive quantum interference due to its cross-conjugated pi system is examined between 2-zigzag graphene nanoribbon leads. The functionality is then applied to a 5-armchair graphene nanoribbon structure. Next, the strong spin filtering properties of a nitrogen doped 2-zigzag graphene nanoribbon device are investigated, which can be activated through simple chemical or physical transformations. In a third project, the cross-conjugation patterns in diboracene systems are explored, how different substitutions can effect the resulting quantum interference and how this effect can produce spin-selective transport in the unsubstituted systems in the triplet state.

The increasing switching frequencies of modern field-effect transistors require time-resolved methods, e.g. the driven Liouville--von-Neumann equation, to describe the transport behavior before the current equilibrates. In the final project of this thesis, the quantum chemical toolkit that can be used to describe molecular transport properties is expanded by introducing the novel resolution-of-identity stochastic time-dependent configuration interaction (RI-sTDCI) method for time-resolved transport simulations.

he ongoing opioid crisis in the US highlights the urgent need for safer analgesics, as conventional opioids acting through the µ-opioid receptor (MOR) are associated with severe adverse effects, including respiratory depression, tolerance, and addiction. The κ-opioid receptor (KOR) has emerged as a promising alternative target for pain management, offering potent analgesia with a lower risk of addiction. However, therapeutic use of the KOR has been hindered by its own undesirable side effects, such as dysphoria and sedation. Recent advances suggest that G protein-biased KOR agonists could mitigate these limitations. Since biased signaling arises from subtle and dynamic conformational changes within the receptor, molecular dynamics (MD) simulations provide a powerful means to analyze these processes at atomic resolution. Yet, the vast amount of data generated during MD simulation remains a challenge. To address this, new computational tools were developed to rationally understand and analyze molecular mechanisms underlying KOR-biased signaling.

The first tool, OpenMMDL, was developed to streamline the setup, execution, and analysis of MD simulations. OpenMMDL enables the exploration of protein-ligand interactions in a fully dynamic context, capturing the continuous structural fluctuations that define receptor-ligand interplay. Its intuitive, low-code interface allows both expert and non-expert users to conduct end-to-end simulation workflows, beginning from system preparation and trajectory generation to automated post-simulation analysis, with minimal technical overhead. The second tool. MDPath is an analysis framework designed to detect allosteric communication paths that underlie and stabilize distinct conformational ensembles of the receptor. Understanding communication networks is essential for identifying the subtle molecular mechanisms that give rise to signaling bias within GPCRs. Furthermore, MDPath enables residue-based path tracking; this capability provides a mechanistic bridge between specific protein-ligand interactions and long-range conformational transitions, offering a powerful means to dissect the dynamic processes governing different activation states.

By applying these new tools in combination with established molecular dynamics simulations, we investigated a novel series of dual-charged naltrexamine amides designed to exhibit G protein-biased activity at the KOR. The simulations revealed distinct allosteric communication paths that govern biased signaling, providing atomic-level insight into how specific structural features shape receptor dynamics. Comparative analyses with the unbiased ligand MP1104 and the G protein-biased agonist nalfurafine uncovered distinct allosteric routes within the KOR that can be exploited for the rational design of G protein-biased ligands, ultimately advancing a framework for the development of safer analgesics. Beyond these receptor-specific insights, this work establishes a broadly applicable computational workflow that integrates MD simulations, dynamic pharmacophore characterization, and allosteric path analysis. Such an approach provides a versatile foundation for addressing mechanistic questions not only in GPCR signaling but also across a wide range of protein systems where dynamic allostery governs biological function.

Blended care (BC) integrates digital components with face-to-face (f2f) psychotherapy and has been proposed as a promising approach to improve access, flexibility, and continuity of mental health care. While a growing body of research suggests that BC can achieve outcomes comparable to or better than traditional psychotherapy, considerably less is known about how therapeutic change unfolds in BC and which mechanisms are involved. Addressing this gap is essential for the informed development, implementation, and evaluation of blended interventions. This dissertation investigates mechanisms of change and facilitating factors in BC, with a particular focus on therapeutic alliance, self-efficacy, therapeutic agency, and empowerment. Across four empirical studies, the dissertation combines different study designs: participatory intervention development, systematic review and meta-analysis, longitudinal modeling, and qualitative content analysis. Study 1 used a participatory design approach to develop a blended intervention (TONI). Through interviews, focus groups, and iterative prototyping with patients and therapists, the study identified facilitators and barriers for BC implementation. Findings highlighted the importance of transparency, usability, and flexible alignment with patient needs, as well as concerns regarding workload, data security, and potential impacts on the therapeutic relationship. These insights informed the design of TONI, a transtheoretic and transdiagnostic BC intervention to be used in routine outpatient psychotherapy. Study 2 synthesized existing evidence on self-efficacy as a mechanism of change in internet-based interventions, including blended formats, using a systematic review and meta-analysis. Across 41 randomized controlled trials, digital interventions showed small to moderate positive effects on self-efficacy. Evidence for self-efficacy as a predictor of outcomes was mixed, while it emerged as a potential mediator. However, heterogeneity was high, and the methodological quality of mediator studies was largely insufficient. Study 3 examined longitudinal associations between therapeutic alliance, self-efficacy, and therapeutic agency and outcomes in blended versus routine outpatient psychotherapy. All constructs were associated with better outcomes. Temporal effects showed that increases in alliance, mental health self-efficacy, and agency predicted subsequent improvements in outcomes across both treatment formats. Differences in processes between BC and f2f psychotherapy were only found for general self-efficacy, suggesting that BC primarily mobilizes established therapeutic processes rather than distinct, format-specific patterns. Study 4 explored empowerment from the patient perspective using qualitative responses from a large outpatient sample. Patients described empowerment as a multidimensional experience encompassing increased everyday functioning, more effective coping strategies and a better self-concept. The factors patients perceived as contributing to empowerment, such as the therapeutic relationship, the structure of therapy and opportunities to apply strategies, partly overlapped with how empowerment itself was experienced. Digital components were rarely described as transformative but were perceived as supportive tools. Taken together, the findings suggest that BC draws on established therapeutic mechanisms rather than introducing fundamentally new processes. Digital components appear to support mechanisms such as alliance, self-efficacy, agency, and empowerment, without substantially altering their function. At the same time, variability in engagement, implementation, and measurement are methodological challenges in identifying mechanisms in BC. Future research should focus on mechanism-sensitive designs, implementation challenges, personalization strategies, and the careful integration of emerging technologies within BC.

This thesis investigates the interplay between tectonic processes, structural inheritance, and landscape evolution along the Chilean forearc across contrasting climatic zones. By integrating geomorphic mapping and morphometric analysis of remotely sensed datasets, structural field data, and seismostratigraphic interpretation for kinematic modelling, this research shows how transients of crustal deformation and upper-plate fault reactivation shape the forearc landscape over multiple timescales (from decades to several millions of years).

In Northern Chile (19-20°S), the long-standing hyperarid climate led to a sediment-starved trench that abuts a margin impacted by subduction erosion and aseismic ridge subduction. Here, the focus lies first on identifying and characterizing previously unmapped faults within the Coastal Range and Central Depression transition, which subtly deform Quaternary drainages. Results reveal the presence of slow-deforming, east-vergent compressive blind faults with two main tectonic phases: Late Miocene–Pliocene reverse faulting that inverted the Andean piedmont, followed by Pliocene–Quaternary transpression extending southward. These faults actively modulate topography, force Quaternary aluvial-alluvial drainages to reorganize, and pose unrecognized seismic hazards. Seismostratigraphic interpretation and kinematic modelling based on reprocessed seismic data and lithologic well reveal that the east-vergent faults are steep crustal structures inherited from Early Cretaceous extensional tectonics and have been episodically reactivated since the Late Cretaceous. Their reactivation reflects changes in plate convergence and the arrival of subducting ridges to the trench, causing intermittent uplift of the Coastal Range block. Available microseismicity and focal mechanisms show the reactivation and switching kinematics of these persistent upper-plate faults within the ongoing seismic cycle of subduction earthquakes.

In Southern Chile (37-39°S), the forearc has experienced intense glaciation since the Pliocene. The trench sediment infill is thick, and the margin is impacted by accretionary processes. Here, the focus lies in exploring the history, patterns, and drivers of the rapid Late Neogene uplift of the coastal Nahuelbuta Range and upper-plate fault reactivation along its boundary with the Central Depression. Geomorphic mapping and river profile analysis indicate an onset of uplift between 3 and 2.5 Ma, with subsequent discrete pulses of rapid uplift lasting ~1 Ma. Paleoclimate proxies show that these pulses correlate with Southern Hemisphere glacial cycles and consequent sudden increases in sediment influx to the trench, meaning larger volumes of subducted material. Further analytical models show that the duration and wavelength of transient uplift pulses are compatible with underplating nappes of sizes and depth matching thick and sediment-rich portions of the subduction channel.

This thesis emphasizes the oscillating nature of tectonically driven forearc landscape evolution across diverse geological timescales and climate settings. It highlights the importance of integrating geomorphologic analysis and field data with subsurface imaging to understand the interplay between upper-plate faults, the seismic cycle, and subduction zone processes, which contribute to crustal deformation and seismic hazard in active convergent margins.

The thesis aimed at the introduction of perfluoroalkoxy-groups in organic molecules. Herein, the forcus was set to trifluoromethoxylations. In the first project nucleophilic substitution on primary and secondary benzylic bromides by ORF-groups was realized. Here, AgORF salts prepared by RIEDEL and coworkers was used as ORF anion source. The reaction was performed in acetonitrile at ‒30 °C up to room temperature. The utility of AgORF salts as transfer reagents for ORF-groups was shown by 15 examples. Furthermore, an aliphatic, primary iodine, 1-iodohexadecane, was successfully functionalized. In the second project, a radical trifluoromethoxylation protocol using bis(trifluoromethyl)peroxide (F3COOCF3, BTMP) as OCF3-source was developed. Through C‒H activation with acetone or tetrabutylammonium decatungstate (TBADT) functionalization of benzylic, non-activated, and aldehydic C‒H bonds was realized. Herein, odorants, pharmaceuticals, and natural products were derivatized. Additionally, use of 13C-labeled BTMP enabled isotopic labeling and trifluoromethoxylation in a single step. Moreover, determination of the quantum yield indicated a radical chain mechanism. The new methodology was successfully applied to the gram-scale synthesis of an ibuprofenderivative. In the third part of this work, key intermediates towards the synthesis of daphnepapytone A were synthesized. This natural product was isolated by DAI, ZHAO, and coworkers in 2022 from Daphne papyracea and has an unprecedented, caged 5/6/4/5 scaffold. Furthermore, three of five stereocenters are quaternary ones. Starting from (+)-dihydrocarvone and ethyl vinyl ketone, the two natural products, daphbolide A and B, were synthesized in four and five steps, respectively. The two starting materials were further used for the synthesis of a C6-oxidized derivative of daphbolide A. As daphbolide A and its C6 oxidized derivative turned out to be challenging in C9-oxidation, a decalin-like moiety carrying the correct oxidation state at C9 was prepared from (R)-(‒)-carvone. Here, α,β- unsaturation turned out to be critical. Thus, α-santonin was used as cheap starting material already carrying a dienone mandatory for the synthesis. Using α-santonin, the corresponding guaianolide was prepared, which was also problematic for C9-oxidation.

Trees may experience numerous recurring attacks by herbivorous insects during their life. As perennial plants, they might benefit from adjusting their anti-herbivore defences to these experiences. This thesis investigated how defence responses of the field elm (Ulmus minor) against the herbivorous elm leaf beetle Xanthogaleruca luteola are shaped (primed) by previous elm leaf beetle infestations. Hence, the studies focused on a specific defence strategy, i.e. priming of anti-herbivore defence by plant responses to previous, infestation-indicating (warning) cues. One of the studies addressed short-term priming of defence elicited by the beetle’s egg deposition as warning cue and acting against the hatching larvae. Another study focused on long-term priming elicited by infestation of the tree in one year and acting against infestation in the subsequent year. The studies included insect performance assays as well as quantitative analyses of phytohormones, of phenylpropanoids as defence compounds, and of transcript levels of genes involved in the phenylpropanoid pathway and epigenetic modifications. Phenylpropanoids may have anti-microbial effects, and microbes associated with insects may significantly affect plant – insect interactions. However, prior to this thesis, almost no knowledge about microbes associated with the elm leaf beetle was available. Therefore, a final study analysed the microbiome of X. luteola. The study on short-term priming, which examined how the plant’s response to elm leaf beetle eggs influences the subsequent elm's defence against newly hatched larvae feeding on the leaves, revealed the following results. Phenylpropanoid biosynthesis was more strongly induced in leaves upon which eggs had been laid prior to larval feeding. The results indicated that the phytohormone salicylic acid (SA) is involved in regulation of this response. In detail, SA concentrations were higher in leaves subjected to egg laying and 24 hours of subsequent larval feeding compared to leaves that had experienced larval feeding alone. Interestingly, at this time, SA concentrations were not elevated in leaves exposed only to egg laying without larval feeding. By contrast, the phytohormone jasmonic acid (JA) was induced by larval feeding after 24 hours, regardless of prior egg laying. A PCR analysis revealed that egg laying intensified the feeding induced expression of a gene encoding phenylalanine ammonium lyase (PAL), the input enzyme of the phenylpropanoid biosynthetic pathway, compared to the expression in egg free leaves. Accordingly, the concentrations of certain phenylpropanoids, namely kaempferol and quercetin derivatives, were elevated in previously egg deposited leaves after 24 hours of larval feeding compared to untreated control leaves. This was not the case in feeding damaged leaves without prior egg deposition. The study on long-term priming, which examined whether an infestation involving adults, eggs, and larvae of the elm leaf beetle in one year could improve the elm's defences against a second elm leaf beetle infestation after a winter dormancy, provided the following results. Indeed, elm leaf beetles developed slightly worse on previously infested trees compared to control trees during the second year. Larvae took longer to pupate, while the resulting beetles only reached a lower weight. In addition, females that had developed from larvae on the previously infested trees laid most of their eggs later, though the total number of eggs did not differ between groups. Concentrations of kaempferol and quercetin derivatives were slightly enhanced in newly infested leaves of previously infested trees compared to those of previously naïve trees. This suggests that elm leaf beetle infestations may enhance elm defences against subsequent elm leaf beetle attacks, further reinforcing short-term egg primed elm defences. Remarkably, since elms are deciduous, information about the previous infestation could not have been stored within the previously infested leaf material, but must have been transferred to the newly sprouted leaves. In a third study, we analysed the microbiome of elm leaf beetles using culture-dependent and culture-independent methods to determine whether they harbour a resident microbiome. We studied all elm leaf beetle life stages from adult beetles to beetles of a second generation. Furthermore, the microbiome of the leaf material, on which the elm leaf beetles were kept, was studied. No persistent bacterial community was detectable in the different elm leaf beetle life stages and on elm leaves. Fungi were prevalent in feeding stages of the elm leaf beetle and on elm leaves. These fungi could mainly be assigned to Penicillium Ianosocoeruleum and Aspergillus flavus. Notably, P. Ianosocoeruleum was predominantly found on feeding damaged leaves, indicating that it thrives in the presence of elm leaf beetles. Therefore, we wanted to know whether these fungi influence the development of elm leaf beetles. We found that larvae feeding on leaves inoculated with P. Ianosocoeruleum gained more weight, and the resulting beetles produced larger egg clutches compared to elm leaf beetles feeding on surface-sterilised leaves or leaves inoculated with A. flavus. Thus, elm leaf beetles appeared to benefit from a colonization with P. Ianosocoeruleum. In summary, this thesis provided the following insights into the interactions among elms, elm leaf beetles and associated microorganisms: a) Elm’s defences against elm leaf beetle larvae are short-term primable by the beetle’s eggs. Primed elm leaf responses comprise enhanced phenylpropanoid biosynthesis. SA concentrations are enhanced in egg primed, feeding induced leaves when compared to egg-free, feeding damaged leaves. b) Elm’s defences against elm leaf beetles are also long-term primable by insect infestation (adults, eggs, larvae) in a preceding year. Hence, the priming information is transmitted from leaves that are shed in one year to newly grown leaves in the following year. While concentrations of phenylpropanoids were moderately enhanced in infested leaves of long-term primed trees, concentrations of JA and SA did not differ between leaves of long-term primed and non-primed trees. c) Elm leaf beetles do not harbour a persistent bacterial microbiome. The fungus P. Ianosocoeruleum proliferates in the presence of elm leaf beetles (increased presence on feeding damaged leaves and in feeding stages of the elm leaf beetle). d) Penicillium Ianosocoeruleum positively influences the growth and reproduction of elm leaf beetles.

In meinem Promotionsvorhaben in der Sozial- und Kulturanthropologie untersuchte ich die spezifischen Affekte in Multi-Rassifizierungs- und Rassismuserfahrungen in den Lebensge-schichten von Frauen mit einem koreanischen und einem nicht-koreanischen Elternteil in Deutschland. Während meiner ethnographischen Feldforschung von Juni 2020 bis Juni 2021 ging ich – hauptsächlich mittels einer „Digital ethnography“ – den Multi-Rassifizierungspro-zessen in den Leben meiner Gesprächspartnerinnen nach und wie sie je nach Lebensphase er-lebt, gefühlt und gewertet wurden.

Hierbei wurde aus affekttheoretischer und rassismuskritischer Perspektive schnell ersichtlich, wie ein „Doing race“ auf vielschichtige und fast unmerkliche Arten und Weisen integraler Be-standteil aller Lebensbereiche, über den gesamten Lebensverlauf meiner Gesprächspartnerin-nen war und ihre Leben in bestimmte Richtungen lenken konnte. Die Frauen aus meiner Un-tersuchung wurden noch vor der Geburt bis ins Erwachsenenalter kontinuierlich in der Familie, von Freund*innen, in der Öffentlichkeit sowie im Schul- und Arbeitskontext als multi-rassifiziert ‚illegitim‘, ‚defizitär‘ oder ‚faszinierend‘ herausgestellt. Meist erfolgte diese Herausstellung über eine „Sezierung“ der Körper und Sprachen der Frauen, die dann als von der mono-rassifiziert-lingualen Norm ‚abweichend‘ galten. Ihre daraus resultierenden meist negativen multi-rassifizierten Affekte wurden dominanzgesellschaftlich, v.a. über Beschämung, ebenfalls als ‚illegitim‘, ‚übertrieben‘ und ‚unangemessen‘ re-produziert. Durch dieses konstante „affektive Gaslighting“ meiner Gesprächspartnerinnen, begannen sie mit zunehmendem Alter selbst an der Berechtigung ihrer negativen Affektivität zu zweifeln. So war ein starkes Ringen mit den multi-rassifizierten Affekten zwischen „minor“ Gefühlen, wie Scham, und „major“ Gefühlen, wie Wut, zu beobachten, worin meine Gesprächspartnerinnen oft stagnierten. Dies führte in vielen Fällen zu einem affektiven Orientierungsverlust, wodurch Rassismus als Norm in Deutschland unbestritten bestehen bleiben kann. Dennoch wurde im Ringen ein geringer Bewegungsradius erkennbar, der es den Frauen erlaubte sich je nach Historie, Lokalität, Relation und Kontext aus der Stagnation partiell zu lösen und Herrschaftsverhältnisse zu hinterfragen, was jedoch nicht ohne immense affektive Anstrengungen einherging.

Als erster Beitrag dieser Art bringt meine Arbeit nicht nur mehrere und zu Teilen vernachlässigte Forschungsfelder in Deutschland neu zusammen, sondern leistet auch einen gesellschaftlich höchst relevanten Beitrag zur Fortentwicklung des Verständnisses, wie ein „Doing race“ auf fast unmerkliche Weisen Alltagswelten im gegenwärtigen Deutschland sozial und emotional strukturiert.

The rapid proliferation of single-cell RNA-seq data requires novel algorithms for identifying cell types and their defining marker genes. While clustering algorithms aim to discover biologically meaningful cell groups, most require prior knowledge of the number of cluster and fail to identify cluster-specific genes. The absence of a consensus definition for cell types further underscores the need for transparent algorithms that explicitly report the genes driving their clustering decisions. Current workflows require separate downstream analyses to identify marker genes and annotate cells. However, this approach reuses the same data for both clustering and statistical testing, potentially inflating the statistical significance of p-values. Moreover, researchers routinely assess clustering quality using non-linear dimensionality reduction methods like UMAP or t-SNE, whose embeddings can distort the true cluster structure of the data and vary substantially with hyperparameter choices.

This thesis introduces CAbiNet and CAdir, two methods that aim to solve these problems by simultaneously clustering cells and their marker genes by leveraging the properties of correspondence analysis (CA). This biclustering approach circumvents statistical issues arising from data reuse, while providing intuitive insights into the cell clustering through the co-clustered marker genes. Because they are specifically designed for sparse scRNA-seq data, both methods outperform existing biclustering tools and provide visualizations for interpreting and evaluating results. CAbiNet unfolds the data manifold using a cell-gene graph on which a joint clustering of cells and genes can be performed. CAbiNet's unique graph combining cells and genes is ideally suited to unraveling complex non-linear cell-gene relationships as well as developmental or differentiation trajectories. The same graph can then be visualized in a two-dimensional embedding, called biMAP, in which both cells and genes can be plotted and which highlights the relationship between marker genes and their respective clusters. CAdir, on the other hand, exploits the geometry induced by CA to cluster cells and their marker genes based on the directions they lie in CA space. This approach enables extremely fast biclustering of cells and genes and can therefore be applied to even cell atlas-sized datasets. The angle between directions can be used to dynamically determine the number of clusters in the data in an almost parameter-free manner without user intervention. Because CAdir clusters based on directions in CA space, Association Plots can be used to directly visualize each cluster direction to assess cluster quality and to discover cluster specific marker genes based on the embedding. We demonstrate for both methods how they can be used to generate biologically meaningful insights on real, experimental datasets. Finally, in comprehensive benchmarks, CAbiNet and CAdir are compared to other state-of-the-art clustering and biclustering methods, highlighting their competitiveness and versatility in clustering a wide variety of datasets.