Spatiotemporal single-cell Bioinformatics
Dr Fabian Kern
Infections and therapies using anti-infectives involve spatially and temporally dependent biochemical processes in the human body with many yet to discover molecular relationships. We develop progressive bioinformatics-driven approaches as for example domain-specific neural networks based on RNA profiles, for the interpretation of healthy and diseased cell functions. Advanced computational models will allow us to gain a deeper mechanistic understanding of the cellular host signaling cascades induced by pathogenic microbes and novel drug candidates. We thus aim to promote the translational process for compounds developed in the HIPS toward successful clinical applications.
Our research and approach
High-throughput sequencing and ultra-efficient bioinformatics software drive most branches of research in modern biomedicine. Essential pathogenic factors, for instance virulence or drug resistance are governed by complex gene-regulatory pathways as well as the type of host cell infested. The underlying molecular processes can be deciphered using bioinformatics methods. Further research on antibiotics is expected to comprise spatially and temporally resolved interdisciplinary studies to comprehensively assess the effectiveness of new candidate compounds with respect to pathogenic microbiota and infection course of disease. Therefore, we consider such complex but more realistic studies essential to calculate the chances of success for our translational efforts and new intervention strategies.
The research group led by Dr. (PhD) Fabian Kern focusses on the bioinformatical analysis of (single-cell) RNA-sequencing data, amongst others utilizing lab infection models and human samples from the clinics. Moreover, we are interested in gaining a better understanding of microbial communities upon infection of the host that surround biological interfaces such as the gut or blood-brain barrier. We make us of paired metatranscriptomics data to logically connect the microbial activity with the in-parallel RNA-profiles of the primary or secondary affected host cells. Our goal is to identify reproducible cellular changes over space and time caused by infections and anti-infectives via comprehensible machine-learning models and graph-based algorithms. In this way, we also try to enable a stepwise interpretation of complex human phenotypes like accelerated aging or chronic diseases of the elderly that are caused potentially by severe infections.
The rapid development of RNA-sequencing assays at unprecedented throughput and resolution enables us to capture detailed RNA-profiles of many thousands of single-cells from human tissue and body fluids. New experimental approaches combining already established methods to spatially tag RNA molecules and high-throughput sequencing (Spatial Transcriptomics) open a myriad of potential viewpoints to study the spatial patterns for and around center of infections in up to hundreds of tissue samples. For a deeper data analysis of sequencing data and to discover functional relationships advanced bioinformatics tools are required. Moreover, many known but also yet hidden technical factors in these experiments play an important role during data generation, considerably influencing our interpretation at a later stage. Thus, we make use of different state-of-the-art technologies by leading companies in the field to test and compare our software under varying experimental setups. Primarily we are looking for interesting and challenging applications around infection and antibiotics research as to cartography cellular impacts of new drugs and compounds.
RNA-profiles at the single-cell level and sub-cellular areas in a tissue slice are captured and spatially tagged on a two-dimensional matrix followed by sequencing (left). The spatially resolved cellular gene expression profiles are then independently clustered according to their similarity and function, and then reconnected with their spatial coordinates (right). © BGI-Shenzhen
Following the initial processing and data curation of (single-cell) RNA-seq data, one may successively introduce further layers of abstraction. To this end, we are searching for new algorithms to distinguish pathogenic cell profiles from healthy ones as to find unique disease signatures. Modelling cells and cell communities as nodes in local or global graphs is a promising approach that recently received more attention in the field. Edges between the cell nodes are determined by ligand-receptor pathways or similar biochemical features that are routed by gene expression. As one derivative of this approach to find gene signatures (or panels) we might be able to decipher which part of the transcriptome is afflicted by pathogens, but also which communities and intercellular communication pathways are the most affected. In a next step, new anti-infectives could be judged according to their effectiveness and compliance, allowing us to monitor molecular repair by treatment success via the previously identified gene panels.
Biological barriers are an essential mean of the human body to enable necessary physiological conditions. First and foremost, they keep tissue compartments and different zonation areas sterile and free of circulating microbes. Numerous examples of human diseases exist where impaired biointerfaces are one of the underlying causes. At the same time, an on-going challenge in drug development is to transfer the chemical compounds beyond these barriers. Thus, we are interested to uncover gene expression programs of the often very specialized cells, like endothelial cells, residing at or below tissue interfaces supporting us to improve antibiotics. Further, we see potential in new compounds able to repair dysregulated barrier cells recovering them into their healthy state.
A manifold of biological barriers exists in the human body, at which different pathogenic microbes act. Immune and chronic inflammatory processes are mediated by cellular signaling pathways, even in non-infected areas of deeper tissue layers. We are interested to uncover the associated mechanisms with high spatial and temporal accuracy. © Biorender.com
Nowadays Bioinformatics routinely involves processing enormous amount of data of various flavors and counting. Thus, we need of open, flexible, and more efficient formats for data storage and exchange. Our group makes use of an already broad experience on developing peer-reviewed online resources like databases and web servers for interactive computing, to continuously share new software and experimental data sets with the scientific community. We place value on providing sufficiently preprocessed and comprehensive data collections. Most importantly, we commit ourselves to enable transparency and reproducibility in our studies and always try to find new ways on how to improve our existing measures.
With the aid of large plasmid sequence databases, a detailed taxonomic tree of bacterial strains can be reconstructed. Figure adapted from our publication: Georges P Schmartz, Anna Hartung, Pascal Hirsch, Fabian Kern, Tobias Fehlmann, Rolf Müller, Andreas Keller, PLSDB: advancing a comprehensive database of bacterial plasmids, Nucleic Acids Research, Volume 50, Issue D1, 7 January 2022, Pages D273–D278, https://doi.org/10.1093/nar/gkab1111. © Oxford University Press