SBI – Department of Systems Biology and Bioinformatics
Faculty of Computer Science and Electrical Engineering
University of Rostock
Ulmenstrasse 69 | 18057 Rostock
Germany
+49 381 498-7571
olaf.wolkenhauer@uni-rostock.de
Automatic Classification of the Bronchus Structures using a Rule-Based Approach
Abstract: Knowledge of a patient’s bronchus structure is essential for surgery, for example, where lung segments have to be removed. However, the deep bronchus structures are hard for surgeons to identify by only viewing CT image slices. This thesis presents a pipeline, which automatically classifies the bronchus down to the tertiary bronchi using a rule-based approach. The bronchus classification with the lowest cost is quickly found by using Dijkstra’s algorithm to search over all possible classifications matching the rule set. The algorithm is applied to a data set of 76 segmented High Resolution Computed Tomography (HRCT) images provided by the Department of Thoracic Surgery at University Medicine Rostock. Furthermore, to analyse variations in the bronchus, the secondary bronchi were clustered into groups sharing the same deep structure. Finally, the pipeline was used to manually create classification ground truths for all patients up to the tertiary bronchi. The algorithm was evaluated on the created ground truth data, achieving a classification accuracy of 98.4% for the secondary bronchi and 84.2% for the tertiary bronchi. The annotated 3D models provided by the pipeline greatly enhance a surgeon’s ability for intuitive visualisation of the bronchus structure.
(Bachelors Thesis)
Functional Characterization of Long Non-Coding RNAs from Stem Cell Derived Cardiomyocyte Cell Types
Bachelor thesis within the study degree of Medical Biotechnology
PhD theses
Accessible Tools for Biomolecular Data Analysis using Large-Scale Knowledge Graphs
Biomolecular experiments generate data to improve our understanding of biological systems and their response to external stimuli. In the context of diseases, the goal is to use this information to modulate such systems in a desired fashion. However, data interpretation proves challenging due to the heterogeneity of data types across multiple levels of structural and functional organization. With a substantial number of potentially significant molecules, cell types, processes, and disease phenotypes, there is a need for tools to explore data from various sources and types. To this day, there is no single all-encompassing multi-level approach to analyzing such heterogeneous biomolecular data. In this PhD research project, I constructed knowledge graphs (KGs) to represent disease-related multi-level processes in a standardized format. I then developed tools to explore these knowledge graphs together with experimental data. The KGs are accessible through public, interactive, and community-driven platforms, referred to as “Disease Maps”. I demonstrate the approach with three Disease Maps. First, I introduce the Atlas of Inflammation Resolution (AIR) as a Disease Map of the molecular and cellular processes involved in acute inflammation and inflammation resolution. I present a novel enrichment-based analytical approach called 2DEA, which was integrated into the AIR. The approach facilitates in silico perturbation experiments and inferences from experimental data. I demonstrated the applicability of the AIR and the 2DEA through two studies in which I evaluated the mode of action of multi-target anti-inflammatory drugs and investigated cell type-specific gene regulation of lipid mediator synthesis. Secondly, I present the Sarcopenia Map that links molecular processes of food intake, gastrointestinal diseases, and sarcopenia through Boolean modeling. Finally, I describe the MASLD Map, which combines multicompartmental and Boolean approaches to study spatiotemporal mechanisms in steatotic liver diseases. The results of my work have been developed in several interdisciplinary collaborations with experimental, clinical, and industry partners. With the outcomes presented in this thesis, I provide accessible tools for biomolecular data analysis using large-scale knowledge graphs. They support the integrative analysis of diverse data types across multiple levels of functional and structural organization in biological systems.
Improved imbalanced classification through convex space learning
Imbalanced datasets for classification problems, characterised by unequal distribution of samples, are abundant in practical scenarios. Oversampling algorithms generate synthetic data to enrich classification performance for such datasets. In this thesis, I discuss two algorithms LoRAS & ProWRAS, improving on the state-of-the-art as shown through rigorous benchmarking on publicly available datasets. A biological application for detection of rare cell-types from single-cell transcriptomics data is also discussed. The thesis also provides a better theoretical understanding behind oversampling.
Defense: 16 Dec. 2021
Customized Workflow Development and Omics Data Integration Concepts in Systems Medicine
The ever-increasing amount and diversity of biological and medical data is a major challenge in computational analyses. Computational methods have to be combined into analysis workflows for seamless, swift, and transparent computation. In this work, numerous workflows were developed for the general processing of bulk RNA sequencing (RNA-Seq), single-cell sequencing experiments, and non-coding RNA identification. Mathematical concepts of machine learning and univariate meta-analyses were successfully implemented to independently investigate the role of cell therapies in cardiac regeneration.
Defense: 30. November 2020
Improving Reproducibility and Reuse of Modelling Results in the Life Sciences
Research results are complex and include a variety of heterogeneous data. This entails major computational challenges to (i) to manage simulation studies, (ii) to ensure model exchangeability, stability and validity, and (iii) to foster communication between partners. I describe techniques to improve the reproducibility and reuse of modelling results. First, I introduce a method to characterise differences in computational models. Second, I present approaches to obtain shareable and reproducible research results. Altogether, my methods and tools foster exchange and reuse of modelling results.
Defense: 30 August 2018
An investigation of microRNA target regulation mechanisms using an integrative approach
2015
Transcriptional Responses to Radiation Exposure Facilitate the Discovery of Biomarkers Functioning as Radiation Biodosimeters
The development of new methods for a retrospective quantification of the radiation dose of exposed individuals is of widespread interest. To this end, I developed a computational framework for biomarker discovery and radiation dose prediction and successfully identified gene signatures with which low and medium to high radiation doses can be accurately quantified. To enhance our understanding of the radiation-induced transcriptional response, I additionally analyzed microarray data of human PBLs after ex vivo gamma-irradiation and characterized affected functional processes and pathways.
Defense: 13 May 2014
Compartmental Modelling of the Wnt pathway
The Wnt pathway plays a critical role in development and disease. The key player of the pathway is b-cat. In the nucleus, the complex formation of b-cat and TCF initiates target gene expression. Its activity is mainly regulated by retention and degradation by its antagonists APC, Axin and GSK3. Based on experimental findings, I develop and investigate compartmental models in order to analyse of the role of nucleo-cytoplasmic shuttling of these proteins in Wnt signalling. I show that the compartmental separation of b-cat and its antagonists yields an increase of the b-cat/TCF concentration.
Defense: 19 December 2013
Characterisation of Signalling Pathways of pancreatic epithelial and cancer tissue using systems biology
2013
A Systems Biology Approach to Understand the Influence of SORLA on Amyloidogenic Processing in Alzheimer's Disease
In this dissertation APP processing is modelled and the influence of SORLA in Alzheimer’s disease is evaluated. A model with both monomeric and dimeric APP in a single compartment best fitted the experimental data suggesting that SORLA prevents APP oligomerization and affects β-secretase indirectly. Including compartmental details into the combined model, the multi-compartmental model determines the relative contribution of SORLA to each APP-cleavage step showing that SORLA specifically impairs the processing of APP dimer and alters the dynamical behavior of beta-secretase.
Defense: 29 January 2013
Regulation of the general stress response of Bacillus subtilis
The bacterium Bacillus subtilis lives in the environmentally diverse soil habitat. Various environmental signals are sensed by the stressosome protein complex. Stimulation of the stressosome modifies its phosphorylation patterns and activates a signalling cascade that lead to protein expression changes. The thesis introduces models to study structural processes and signalling principles of the stressosome. Overall, the thesis enhances our understanding of the SigmaB-mediated general stress response and raises our awareness of the environmental integration of B. subtilis.
Defense: 4 February 2013
A systems biology approach to unravel the cellular function of microRNAs
2012
A systems biology approach to dynamic modelling of the AMP-activated kinase pathway
2011
Stochastic Modelling of Subcellular Biochemical Systems
2009
Random set theory applied to the forecasting of stochastic point processes
2002
Master theses
Machine-Learning-Assisted Construction of a Disease-Specific Molecular Interaction Network from Human Blood RNA-Seq Data
Master thesis within the study degree of Medical Biotechnology
Comparison of Deep Learning Approaches for Cardiomyocyte Evaluation
2020
Arrhythmias are severe cardiac diseases and lethal if untreated. To serve as an in vitro drug testing option for anti-arrhythmic agents, cardiomyocytes and especially pacemaker cells are being generated in vitro from induced pluripotent stem cells (iPSCs). These generated cardiomyocytes resemble fetal cardiac tissue rather than adult cardiomyocytes. An automated tool for evaluations of cardiomyocytes would help the establishment of new generation protocols. In this work, a novel approach for this task is presented and evaluated.
Different convolutional neural networks (CNNs) including transfer models and native 2D and 3D models were trained on fluorescence images of cardiomyocytes, which were rated based on their sarcomerisation and the orientation of sarcomeres (directionality) beforehand. The CNNs were trained to perform classifications on sarcomerisation and directionality ratings and cell source, as cardiomyocytes from five different sources were used in this work. In this thesis, it could be shown that cellular fluorescence images can be analysed with CNNs. This classifier can be used to make trustworthy predictions on the quality of a cardiomyocyte which will hopefully benefit the generation of cardiomyocytes from iPSCs. This classifier is currently being
Curation of an immune cell interactome and its analysis
2019
Master thesis within the study degree of Medical Biotechnology
The interconnectivity of immune cells has been the subject of research numerously due to its importance in different diseases such as autoimmune defects, (microbial) infections and cancer. Various cell types have already been identified that are regulated by a complex network of cytokines and small molecules, of which many may not have been discovered yet. Therefore, it is of great interest to understand these mechanisms as they form the basis for drug development and therapy design. In this project, methods were described to create and analyze a cell interactome of molecular intra- and intercellular communication processes. Many molecular interaction maps (MIMs) have already been developed to evaluate molecular processes in certain diseases or cells. However, they either lacked essential information necessary for accurate modeling of cell-cell interactions or were poorly clinically assessed. Here, systems biology-based rules were defined to model the molecular pathways of intercellular interactions of cells in detail. By mapping expression data of immune cell samples, individual cellular MIMs were created automatically and validated by comparing the results with the current knowledge in the field of immunology. In addition to analyzing intracellular signaling pathways, intercellular communication processes were investigated by connecting the MIMs. The outcomes of this work improve the system biology modeling of molecular interaction networks and further provide the basis for the efficient development of complex intercellular networks to investigate biological and molecular processes in silico.
A Machine Learning approach to identify White Blood Cells
2018
Design of the TriplexRNA: a database of cooperating microRNAs and their mutual targets
MicroRNAs (miRNAs) are an integral part of gene regulation at the post-transcriptional level. Single miRNAs can repress the expression of many genes, however, experimental evidence suggest that their repressive effect alone is rather mild compared with that carried out by cooperating miRNAs. Although synergistic target regulation has been confirmed as a prevalent pattern in gene regulation, the mechanism behind this phenomena are little known, and it is unclear which miRNA pairs and gene targets are involved in the formation of RNA triplexes. The growing amount of data on miRNA-mediated gene expression is motivated by the role they play in cellular functions, where their presence can be correlated with cancers, and cardiovascular diseases, making them good biomarkers for diagnostic and prognostic purposes. However, existing databases relate disease associations with the duplex model: one miRNA inhibiting the expression of one or many genes, while none of them provide an exploration of the more effective regulation operated by synergistic pairs. To overcome this gap, the present work describes the design and implementation of the TriplexRNA: a database of cooperating miRNAs and their mutual targets, which we realised to enable researchers explore novel patterns in gene regulation. The database can be accessed at https://triplexrna.org.
Next Generation Sequencing Data Analysis of Stem Cell Derived Cardiomyocyte Cell Types
2014
M.Sc. within the study degree of Medical Biotechnology
Analysis of the carbon flow in response to iron supply in Clostridium acetobutylicum
2012
Effect of enzyme-complexes on the efficiency of metabolic pathways
2012
Cloud based analysis of Health Data with RESTful Web Services
Computational identification of ADAR editing sites in microRNA precursors
Adaptation in networks of heterogeneous mitochondrial populations
2012
Detection of potential drug targets in cancer signaling by mathematical modelling and optimization
2012
Investigation of mammalian microRNA biogenesis by use of mathematical modeling
2011
Stochastic Modelling and Simulations: An Extended Crossover method for modelling noise in stiff/bistable biochemical systems.
Electrical Engineering, 2011
Data Mining and Clustering in Model Calibration for Systems Biology
2009
Mathematical modelling and mathematical optimisation of biochemical systems: a method of analysis in biotechnology and biomedicine
2007
Development of a SBToolbox module to analyse Power-Law Models
2006
Significance Analysis for Microarray Data
2003
Studienarbeiten
Identification of activated signaling pathways in cardiac stem cell types by using network analysis
2017
Bachelor thesis within the study degree of Medical Biotechnology
Visualising sub-networks in SBML models
2015
further information at the SEMS Webpage.
Classifying reactions of computational models
2015
Further information at the SEMS webpage.
The thesis is available online (PDF).
Bestimmung von Distanzen zwischen Versionen XML - kodierter biologischer Modelle
2012
The impact of iron limitation on growth and product spectrum of Clostridium acetobutylicum
2012
Ansäuerungsleistung und Sauerstoffverbrauch von Säugetierzellen im extrazellulären Medium
2007
Bifurkation analysis of the Hes1 oscillator
2007
Molecular Dynamics Simulation
2007
Konzipierung und Erstellung eines Interfaces zum Export von Matlab SBmodel Modellen nach SBML
2006
Detektion charakteristischer Muster in Proteininteraktionsnetzwerken
2005
Diplomarbeiten
Versioning Concepts and Technologies for Biochemical Simulation Models
2010
Biological Aging, Negligible Senescence and the Concept of Renewal: A Complex Systems Approach by Memory Evolutive Systems
2010
Optimierung von Stimulusprofilen zellbiologischer Experimente durch mathematische Modellierung
2010
Simulation der Messung eines Cell Monitoring Systems (Modellbildung und Parameterschätzung)
2008
A JAVA framework to integrate ncRNA detection methods
2008
Erstellung eines Parameter Estimation Frameworks für die Systems Biology Toolbox for Matlab
2007
Combined Visualisation of Pathway and Protein-Protein Interaction Data
2006
Zur Bestimmung sequenzabhängiger Zusammenhänge zwischen Transkriptom und Proteom
2004