16580 research outputs found
Sort by
Mechanism of action of adipose mesenchymal stromal cells reducing cisplatin-induced proximal tubular epithelial cell injury
No full textKidney disease is a serious health problem worldwide with an increasing prevalence. New research has made it possible to understand the pathophysiology of kidney disease better. However, there are still limited treatment options to stop or reverse the progression of the disease. Therefore, developing new strategies for treating kidney diseases is a critical area of research. In this context, cell-based therapy could be a promising strategy and is currently the focus of preclinical studies.
In our study, we asked ourselves whether mesenchymal stem cells could offer a novel cell-based therapy modality. Many questions remain open about choosing the source that best suits for your scientific question or using paracrine factors rather than cells. Moving on to the use of cell-free therapy raises other questions about the decision to use the whole secretome or just one component.
To find an answer to these questions, we structured this study in different levels, starting from the study of different cell types, passing later in the secretome and then moving on to a situation of injury.
First, we established harmonised tissue culture conditions for the expansion of adipose, bone marrow and umbilical cord MSC between three independent centres to study the reproducibility of these procedures and their impact on their biological characteristics and functionality both in vitro and in vivo. This part of the thesis highlights the importance of adopting harmonised protocols that reduce, but not eliminate, site-to-site variation while specific differences in donors remain evident. Despite the use of a common protocol, the different types of MSCs have shown individual properties, which may have benefits in specific therapeutic settings.
The aim of the secretome study was to test the in vitro efficacy of the different components for different aspects of tissue regeneration and at the same time understand how the different isolation methodologies can influence the results. Extracellular vesicles (EVs), regardless of the method of isolation, failed to replicate the results obtained by the secretome, in contrast, the protein fraction showed an effect very close to that of the secretome. We have shown that depending on the isolation method, contamination with protein residues can alter the results and misinform about the true efficiency of EVs. On the other hand, increased EV concentration in some applications increased effectiveness, indicating that purity and dose affect the final functionality.
At the end, we aimed to test the secretome in a situation of cellular damage in order to identify the beneficial effects and the molecular mechanisms involved to propose a therapeutic strategy. The administration of the conditioned medium protected the kidney cells by reducing cisplatin cytotoxicity, maintaining cell viability, stimulating cell migration, reducing apoptosis and apoptosis-related protein. We demonstrated that the conditioned medium reduced cell death by inhibiting the expression of mir-181a. We propose a so-called type II regulatory circuit by which A-MSC CM and cisplatin affect p53 expression/apoptosis and the counteracting miR-181a expression
Der Gewalt auf der Spur: Die empirische Evaluation eines rechtsmedizinisch-kriminalistischen Triage-Instruments in der polizeilichen Praxis
No full textTrotz der zentralen, mitunter verfahrensentscheidenden Bedeutung der Rechtsmedizin für das gesamte Strafverfahren ist die Einbeziehung dieser forensischen Bezugswissenschaft bei schweren Gewaltstraftaten gegen (über)lebende Opfer in Deutschland bis zum heutigen Tag nicht flächendeckend standardisiert. Im Wesentlichen basiert sie auf den persönlichen Präferenzen und Erfahrungen der sachbearbeitenden Personen der Strafverfolgungsbehörden. Um jedoch überhaupt erst zu der Frage der rechtsmedizinischen Einbeziehung in das Ermittlungsverfahren zu gelangen, muss die Gewalthandlung als solche zunächst von den ersteintreffenden Polizeikräften, in der Regel aus den Reihen der kriminalistisch weniger umfassend ausgebildeten Schutzpolizei, erkannt werden. Auch hierfür existieren aktuell noch keine, an dieser Zielgruppe sowie den besonderen Strukturen und Abläufen der Polizei orientierten, Leitlinien bzw. fachliche Hilfestellungen. Damit sind sowohl die Detektion als auch die Bearbeitung und Dokumentation von schweren Gewaltdelikten, zumindest aus wissenschaftlicher Sicht, nahezu noch komplett unreguliert. In der Konsequenz bleiben diese Taten oft gänzlich unentdeckt oder werden aufgrund des fehlenden forensischen Verständnisses unterversorgt. Insbesondere vor dem Hintergrund, dass in über 99 % der jährlich in der Polizeilichen Kriminalstatistik erfassten Fälle von Gewaltkriminalität die Geschädigten die körperliche Misshandlung überleben, existiert an der Schnittstelle von Polizei und klinischer Rechtsmedizin ein dringender Handlungsbedarf, aber zugleich auch großes Potenzial zur Verbesserung der gesamten Strafverfolgung und des Opferschutzes in Deutschland.
In der vorliegenden Arbeit wurde daher empirisch-explanativ erforscht, wie sich die Zusammenarbeit der Polizei, als primär kriminaltaktisch agierende Strafverfolgungsbehörde, mit der Rechtsmedizin durch den Einsatz eines wissenschaftlich fundierten kriminalistisch-rechtsmedizinischen Triage-Instruments zur Detektion und Klassifikation von Gewalthandlungen weiter verbessern lässt. Dazu wurde ein bereits in einer vorausgehenden Arbeit entwickelter Algorithmus bzw. Kriterienkatalog mit insgesamt acht Teilfragen in einem zweiphasigen experimentellen Forschungsdesign mit mehreren Polizeibehörden evaluiert. In einer ersten Phase wurde mit Unterstützung von über 60 Studierenden der Hochschule des Bundes (Bundeskriminalamt) und der Hochschule für öffentliches Management und Sicherheit (Landespolizei Hessen) eine erste praktische Anwendungsschulung des als kriminalistischen Triage (KriT) bezeichneten Algorithmus mit anschließender gruppierter Online-Befragung mittels Vignettenexperimenten durchgeführt. In der anschließenden Feldphase erprobten, in Kooperation mit dem Innenministerium Baden-Württemberg, über 20 Angehörige des Kriminaldauerdienstes (KDD) der Polizeipräsidien Reutlingen und Offenburg das Instrument auf seine Praxistauglichkeit in ihrer alltäglichen (kriminal)polizeilichen Ermittlungsarbeit.
Die in beiden Phasen erhobenen Daten wurden im Zuge der Forschungsarbeit getrennt voneinander aufbereitet, ausgewertet und ausführlich dargestellt. Hinsichtlich der zuvor aufgeworfenen Problemstellung zeichnete sich dabei ein äußert positives Bild ab. So war die überwiegende Anzahl der Beamt*innen bereits nach einer zweistündigen Anwendungsschulung in der Lage, den Algorithmus selbstständig anzuwenden, schwere Gewaltdelikte zuverlässig zu detektieren und entsprechend den Hinweisen auf der Taschenkarte zielführend weiter zu bearbeiten. Die Ergebnisse der als Feldphase bezeichneten zweiten Phase lassen außerdem auf einen, über alle erhobenen Deliktfelder hinweg, messbaren und statistisch signifikanten Anstieg der rechtsmedizinischen Beteiligung, in der durch den KDD bearbeiteten Frühphase des Ermittlungsverfahrens, schließen. Diese positiven Auswirkungen wurden durch die Untersuchungsteilnehmenden in einer schriftlichen Expert*innenbefragung zu möglichen Veränderungen Ihrer Arbeitsweise durch die Etablierung der KriT auch qualitativ bestätigt.
Basierend auf diesen Ergebnissen und den aus den Daten sowie persönlichen Gesprächen gewonnenen Erkenntnissen wurde das kriminalistisch-rechtsmedizinische Triage-Instrument zielgerichtet weiterentwickelt, was im Kern primär eine Veränderung des Layouts sowie leichte inhaltliche Anpassungen bedeutete. Als Produkt der Arbeit wurde der weiterentwickelte und in Forensic Violence Detection Algorithm (ForViDeA) umbenannte Algorithmus sowie eine darauf basierende Adaption für die Schnittstellenzusammenarbeit von Notfallmedizin und Rechtsmedizin in Form eines mehrstufigen, als Forensic Violence Detection Score (ForViDeS) bezeichneten, Scoring-Systems präsentiert.
Zusammen mit dem, durch aktuelle Gesetzesinitiativen und verbesserte Finanzierungsmöglichkeiten, bereits katalysierten Aufbau einer klinisch-rechtsmedizinischen Versorgungsinfrastruktur in Deutschland existiert damit eine echte Chance, die individuelle Behandlung von Gewaltopfern und die Qualität der Strafverfolgung langfristig zu verändern. Damit haben die Arbeit und vor allem die vorgestellten Instrumente das Potenzial, zu mehr Gerechtigkeit und der Verbesserung des Opferschutzes bei (über)lebter körperlicher Gewalt beizutragen. Dafür bedarf es jedoch auch in Zukunft noch sehr viel Überzeugungsarbeit bei allen beteiligten Professionen sowie weiterer empirischer Forschung. Die Arbeit hat das dafür notwendige wissenschaftliche Fundament gelegt, sodass nun der Fokus auf die Weiterentwicklung, Qualitätssicherung und vor allem Verbreitung der Instrumente in den jeweiligen Einsatzfeldern gelegt werden kann
Context-aware clinical photoacoustic imaging
No full textPhotoacoustic imaging (PAI) is an emerging biomedical imaging modality that harnesses pulsed laser light to generate ultrasound waves through thermoelastic expansion, enabling high-resolution, non-invasive visualization of tissue structure and function at
clinically relevant depths. By combining the optical contrast of molecular imaging with the spatial precision of ultrasound, PAI offers a unique capability to assess physiological and pathological processes in vivo. Its relatively low cost, safety, and imaging speed make it a highly promising tool for a wide range of clinical applications, including oncology and vascular medicine. The clinical adoption of PAI, however, has been limited by challenges in interpreting its data without sufficient spatial, temporal, and biophysical context. This thesis addresses this limitation by developing methods that incorporate contextual information across these dimensions, enabling more accurate and clinically meaningful PAI analysis.
The lack of spatial context was addressed with a framework for reconstructing threedimensional (3D) volumes from sets of two-dimensional (2D) images. The central innovation of this approach lies in the use of an optical pattern that encodes spatial information through specific light-absorption characteristics. An extension of the pattern, adding fiducial markers, further enables the multimodal fusion of PAI with magnetic resonance imaging (MRI) and computed tomography (CT), thereby situating PAI within the established clinical imaging landscape. The lack of temporal context, addressed by pattern-based longitudinal registration of 3D PAI volumes, enables a more comprehensive assessment of disease status and progression. Third, a digital twin model was introduced to analyze unexpected clinical observations by disentangling physiological mechanisms from photoacoustic image formation processes. To demonstrate the broad applicability of these methods, they were validated in diverse clinical settings, with applications ranging from cancer therapy to vascular disease diagnosis. In a clinical study on peripheral artery disease, optical pattern-based PAI successfully detected ischemia and muscular heterogeneities, indicating benefits over conventional 2D approaches by combining spatial and temporal context. This thesis also presents the first evidence that PAI can non-invasively capture molecular changes induced by radiotherapy in patients with head and neck cancer. In this study, digital twin modeling further provided a mechanistic explanation for unexpected oxygenation measurements, revealing that these anomalies arose from signal distortions in regions with low blood volume.
In conclusion, this work establishes the concept of context-aware PAI, integrating spatial and temporal, multimodal, and biophysical information to enhance both interpretability and clinical trust. By demonstrating feasibility in clinical studies, it outlines a pathway for translating context-aware PAI into routine medical practice
Fast and Slow: the Evolution of Sex-Biased Expression and Liver Zonation across Mammals
No full textGene expression programs are central to the emergence of phenotypic diversity across species, shaping how cells acquire their identities and functions during development and evolution. In this thesis, I explore two complementary dimensions of how such programs evolve in mammals: (i) the developmental establishment and evolutionary dynamics of sex differences across organs, and (ii) the origin and molecular evolution of the liver’s spatial cell architecture.
In the first part, I used comparative transcriptomic datasets from males and females spanning developmental time series of five major organs (brain, cerebellum, heart, kidney, and liver) in five mammals and one bird. Through this analysis, I showed that sex-biased gene expression is widespread but highly variable across organs and species, and often restricted to specific cell types. Its onset is not gradual but occurs abruptly around sexual maturity, coinciding with the increase of circulating sex hormones. While the identity of sex-biased genes evolves rapidly and the underlying mechanisms differ between organs, the cell types that exhibit sexual dimorphism are deeply conserved, indicating that molecular programs evolve fast, but the cellular framework they act within changes slowly.
In the second part, I investigated the evolutionary origins and dynamics of liver cell organization using single-nucleus transcriptome and chromatin accessibility data from 17 species—16 mammals and one bird—complemented with spatial transcriptomics data. This analysis demonstrated that liver zonation, the compartmentalization of hepatocyte functions along the porto-central axis, is a mammalian innovation absent in birds and fish. Zonation is driven by the emergence of WNT and R-spondin signaling from central vein endothelial cells, which activate central hepatocyte gene expression via the transcription factor TCF7L2. Once established, this architecture has been remarkably conserved across mammals for ~180 million years. Yet, beneath this conserved architecture, the genes showing zonation patterns show a fast turnover, reflecting fast molecular evolution operating within a slow-evolving structural framework.
Overall, this work advances our understanding of the principles that govern gene expression evolution in mammals, showing that although expression programs can change rapidly, functional outcomes evolve more slowly, constrained by developmental, physiological, and ecological demands
Xeno-learning: knowledge transfer across species in deep learning-based spectral image analysis
No full textDifferentiating between tissue types accurately and reliably, in
particular with regard to pathological regions and critical anatomical
structures, presents a perennial challenge in surgical interventions.
Conventional RGB imaging is limited to the visible spectrum and thereby
often fails to provide the necessary contrast for informed
decision-making. hsi in contrast captures rich spectral signatures
across dozens to hundreds of narrow wavelength bands, enabling superior
tissue classification and organ boundary delineation. The widespread
clinical adoption of hsi is however significantly hampered by the
scarcity of large, annotated datasets that are crucial for training
robust dl algorithms, which have shown great promise in the field of
medical imaging. This issue is exacerbated by the
inherent unpredictability of real-world intraoperative conditions as
these can lead to significant domain shifts in spectral data.
This thesis introduces two novel augmentation-based approaches designed
to overcome the limitations of sparse data availability for robust
semantic scene segmentation in hsi. The core idea is to inject relevant
data samples during either model training or inference with the purpose
of enhancing model generalizability.
To begin with, we address the issue of pathology-related domain shifts.
Domain shifts such as changes in tissue perfusion can be caused by
surgical interventions and underlying health conditions. This results in
malperfused organs that largely deviate from the physiological data on
which models are usually trained. To reduce such occurrences, we propose
a new tta strategy that employs physics-informed synthetic spectral
data. We utilize physics-based simulations to produce synthetic data
that compensates for perfusion-related variations as opposed to
generative models that require big data sets. This tta method adjusts
the distribution of unseen test images and aligns them with the training
data, thus improving the robustness of models trained solely on
physiological data to pathology shifts. We illustrate that this approach
effectively bridges the domain gap between physiological and malperfused
datasets and thus enables a more reliable segmentation performance in
different clinical scenarios.
Furthermore, inspired by the concept of xenotransplantation, this thesis
proposes a novel ”xeno-learning” framework for cross-species knowledge
transfer in medical imaging. Acquiring diverse human medical imaging
data is extremely challenging due to ethical and practical constraints.
Animal models (e.g., pig and rat), however, offer more readily available
and diverse datasets, including a wider array of tissue conditions. Our
main observation is that, despite the fact that the absolute spectral
characteristics of tissues may differ across species, the relative
spectral changes caused by physiological changes (e.g. changes in organ
perfusion or the administration of contrast agents like icg), tend to
remain consistent. This forms the basis of our xeno-learning framework.
We develop a novel ”physiology-based data augmentation” technique that
transfers knowledge about domain shifts learned from animal data to
augment human data during training. This enables models to generalize
effectively to unseen pathological human test datasets. We demonstrate
that xeno-learning not only enables successful knowledge transfer but
also outperforms traditional adversarial domain adaptation methods,
highlighting its potential for improving model generalizability between
species
To summarize, this thesis establishes the feasibility of automatic
semantic scene segmentation on human physiological HSI data and offers
innovative solutions that address critical data scarcity and domain
shift challenges. The proposed tta strategy and xeno-learning framework
offer robust pathways for enhancing the generalizability and clinical
applicability of DL-based spectral image analysis. This allows for
advancements in robotically-assisted surgery, organ-aware perfusion
assessment, and ar/vr-based surgical guidance
In situ cryo-electron tomography of SARS-CoV-2 replication pore complex and virus assembly
No full textViruses have evolved distinct strategies to replicate their genomes and assemble their progeny virions in infected cells. Viral proteins often hijack host cellular membranes to facilitate replication and assembly. These processes are tightly coupled and require interactions between viral proteins along with host proteins, which are recruited to assist in viral assembly. Betacoronaviruses have caused previous deadly epidemics, including the ongoing Middle East Respiratory Syndrome (MERS) epidemic. The Coronavirus Disease 2019 (COVID-19) pandemic was caused by the betacoronavirus Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2). Coronaviruses harbor, compared to other RNA viruses, a large RNA genome, which encodes an unusually high number of viral proteins involved in RNA synthesis and RNA-processing. Viral RNA synthesis is associated with replication organelles, which are built from ER membranes and consists of double-membrane vesicles (DMVs) containing pores. Coronavirus replication provides the RNA genome for assembly of progeny virions. The viral budding involves ER-Golgiintermediate compartments (ERGICs), which provide the membrane for assembly of the viral envelope. However, many aspects of viral replication and assembly are unknown, including the function and biogenesis of the replication pore complex inside DMVs. Furthermore, what drives viral assembly and how budding is mediated remains to be elucidated. In this thesis, I investigate the structural aspects of DMV pore formation and viral assembly. In situ cryo-electron tomography (cryo-ET) provides unique structural information, which can be used to build structural models or describe biological processes. I discovered the minimal requirements for DMV pore assembly and showed that the pore structural integrity is essential for DMV biogenesis. I provided new molecular insights into viral budding at the ERGIC and discovered the importance of the envelope protein in this process. Overall, this thesis extends the current knowledge in DMV biogenesis and provides a model for SARS-CoV-2 assembly, where the envelope protein mediates membrane scission and alters membrane curvature at the bud neck. The established SARS-CoV-2 virus-like particle (VLP) system can be used in future experiments to investigate the involvement of host proteins in viral budding and may provide a valuable tool for diagnostics and vaccine development
High-Performance Graph Processing
No full textWith the increasing importance of graphs and algorithms that use them in many new use cases and the increasing availability of data, the number of applications relying on a graph data model is rising. Graphs are not only used in many sciences like biology, but also in the public and commercial sectors to represent, for example, social, business, or traffic networks. Applications use these graphs using queries that apply graph algorithms on them, while they run on systems shared by multiple users.
For the performance, this means that the overall system throughput is more important than the elapsed time of a single isolated graph query. The issue is that parallelizing a graph algorithm efficiently is already complex due to the data-dependent nature of many graph algorithms and the broad range of graph sizes and properties. Likewise, there has been a trend to more complex systems with an ever-increasing number of cores, with more complex cache and memory structures and technologies. In order to deal with this complexity, graph processing engines have been proposed that provide primitives for efficient parallel graph processing. Unfortunately, these systems typically focus on the execution of a single query at a time, so they do not try to improve overall throughput during concurrent usage.
In this work, we investigate high-performance throughput-oriented graph query processing on scale-up systems. Therefore, we first investigate the behavior of a state-of-the-art graph processing engine using different graph sizes and algorithms when multiple instances of it are used concurrently on the same system. These experiments provide insight into the potential, but also the issues of concurrent execution. Afterwards, we investigate parallel execution inside a single query and, in particular, contention on atomic update operations as they are common in many graph algorithms. We propose a simple buffering schema for atomic updates to reduce contention and analyze it in different scenarios that mimic common update patterns.
Finally, based on the previous investigations, we propose a throughput-oriented runtime scheduler for graph queries that automatically controls the degree of parallelization to avoid inefficiencies. The underlying concept consists of multiple parts: (1) sampling is used to determine graph properties, (2) cost estimations and parallelization boundaries are derived from graph, algorithm, and system properties, (3) suitable work packages are generated based on the cost, and (4) executed controlled by a runtime component that controls the parallelism. We evaluate the proposed concept using different algorithms on synthetic and real-world datasets, with up to 16 concurrent sessions (queries). The results demonstrate robust performance despite these various configurations, which is always close to or even slightly ahead of manually optimized implementations
Geschichte oder Gegenwart? Berufsbezogene Überzeugungen von KZ-Gedenkstätten-Guides in Österreich
No full textIm Diskurs zum Lehren und Lernen über die Massenverbrechen des NS-Regimes kommt Gedenkstätten eine besondere Bedeutung zu. Dennoch liegen, insbesondere für Österreich, nur wenige empirische Befunde zur pädagogischen Praxis an diesen Lernorten vor. Anhand von Österreichs bedeutendster NS-Gedenkstätte, der Gedenkstätte Mauthausen, werden im Rahmen der vorliegenden Dissertation die berufsbezogenen Überzeugungen der dort tätigen Guides untersucht. Die vorläufigen Ergebnisse legen nahe, dass der Einfluss des vorherrschenden Diskurses zur ‚Holocaust Education‘ in Abhängigkeit von spezifischen berufsbiographischen Merkmalen der Guides stark variiert und teilweise gänzlich aufgehoben wird
Comparison of Regression and Machine Learning Methods for Variable Selection to Develop a Clinical Prediction Model
No full textClinical prediction models (CPMs) are increasingly finding their way into healthcare as they provide a prediction of a clinical outcome. Thus, they can support the decision-making of healthcare providers and their patients. A common challenge in developing a CPM is the identification of important predictor variables. To conduct variable selection and develop a CPM, a wide variety of methods exists. While some are based on traditional regression methods,
others belong to the field of machine learning. Therefore, this thesis aimed to compare different variable selection methods in the development of a CPM for a continuous outcome
in low-dimensional data and to provide recommendations for practical application.
Initially, a simulation study design was developed to provide a comparison as fair as possible and to examine the strengths and weaknesses of the different methods. For this, four different data-generating processes with increasing complexity were developed, which contain realistic data structures as well as relevant challenges. Based on the simulated datasets, the following established and widely used methods were compared: linear regression with stepwise selection (LMSS), regularized linear regression with elastic net penalty (ENET), gradient boosting with linear regression models (GBM) and decision trees (GBT) as base learners, and the Boruta (RFB) as well as the Hapfelmeier (RFH) method for the random forest. Moreover, the multivariable fractional polynomials (MFP) regression model was applied as a benchmark. All methods selected an increasing number of variables as the sample size of the dataset increased. While LMSS, RFB, and RFH achieved better results regarding the correct inclusion of predictors and correct exclusion of non-predictors, ENET, GBM, and GBT selected nearly all variables. LMSS revealed the best selection properties in the scenarios with low complexity and also identified predictors with non-linear functional form. In the
more complex scenarios, the true inclusion frequency of predictors with non-linear relations to non-predictor variables decreased, especially for LMSS. RFB and RFH achieved the best selection properties in the scenarios of the greatest complexity. The performance regarding the predictive accuracy in test data generally improved as the sample size increased and was similar for all methods. However, while ENET, GBM, and GBT achieved good calibration
by inherently utilizing regularization, RFB and RFH were suggested to be under- and LMSS overfitted in certain scenarios, respectively. In addition to the simulation study, the methods were applied to a real dataset to develop a CPM for intraoperative blood loss during liver
transplantations. The developed simulation design is freely available and can be used for further research, e.g., for investigations of additional methods. Furthermore, the design can be extended to include additional aspects, such as more variables.
This thesis provides a comparison of traditional regression and machine learning methods for variable selection in the development of a CPM for a continuous outcome in low-dimensional data. A sample size of 250-500 observations is required for all methods to identify predictors
sufficiently and achieve adequate predictive accuracy. LMSS can be recommended for data with low complexity as well as, possibly with adaptions, for more complex data structures.
RFB and RFH are recommended for more complex data structures, particularly when interactions between variables might exist