Berlin 2018 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
BP: Fachverband Biologische Physik
BP 9: Postersession I
BP 9.42: Poster
Monday, March 12, 2018, 17:30–19:30, Poster A
Network coherences - a universal approach to quantify the match between “omics” data and a biological network — •Piotr Nyczka1, Marc-Thorsten Hütt1, Kristina Schlicht2, Carolin Knecht2, and Michael Krawczak2 — 1Departement of Life Sciences and Chemistry, Jacobs University, Bremen, Germany — 2Institute of Medical Informatics and Statistics, Christian-Albrechts-University Kiel, Germany
Network-based analyses of “omics” data are a cornerstone of systems medicine. Their goal is to quantify and statistically evaluate the clustering of biological signals (e.g., co-expression of genes) in a network (e.g., a metabolic network or a protein-interaction network). Network coherences are topological indices evaluating the connectivity of subnetworks spanned by the “omics” signal of interest [1,2]. They have been used very successfully to identify scientifically relevant patient subgroups in disease cohorts [2-4].
Here, we aim at a deeper theoretical understanding of network coherence. Using various random walk models on graphs, we test, refine and calibrate this method. In this way, the dependence of a given network coherence upon the number of (e.g., disease-associated) genes, the topology of the underlying biological network or the fragmentation of the functional signal in the network can be studied numerically and compared to analytical predictions.
Our method allows us to detect functional signal even in very noisy data. The main novelty of this approach lies in taking into account collective expression profiles of the whole group of patients and contrast it with individual ones. In order to find relevant signals we “tune” parameters of the collective expression extraction procedure with respect to maximization of the network coherence. This allows us to pick up structures which are not detectable when dealing with individual patients separately.
Based upon our results, we also present a range of applications of (in particular metabolic) network coherence to the analysis of transcriptome profiles in chronic inflammatory diseases. This investigation seems to have huge potential for further development and following applications also outside of this specific field.
[1] Sonnenschein et al. (2011) BMC Systems Biology 5, 40.
[2] Sonnenschein et al. (2012) BMC Systems Biology 6, 41.
[3] Knecht et al. (2016). Scientific Reports, 6.
[4] Häsler et al. (2016). Gut, gutjnl-2016-311651.