Networks in the Life Sciences – Genomics, Proteomics and Systems Biology
14th EMBL PhD Symposium

25th–27th October 2012

Speaker: Prof. Dr. Andreas Schuppert

  • AICES, RWTH Aachen University
  • Director, Bayer Technology Services GmbH
  • International review panel member, EPSRC (Engineering & Physical Sciences Research Council)

Background

Andreas Schuppert Director of Bayer Techonology Services GmbH and professor at Aachen Institute for advanced study in Computational Science, Andreas Schuppert research interests include various data driven modeling approaches and applications, using combinations of statistical, thermodynamic & data mined information he studies both complex functional networks & regulatory networks.

Talk: Modelling of complex phenotypes by functional networks

Efficient modelling techniques is a prerequisite for the prediction, optimization and control of complex biological systems. However, the available understanding of complex systems, especially in biomedicine, is often not sufficient to establish detailed, science-based and reliable models for system behaviour. Therefore reliable, quantitative models are lacking for most medical and biological applications.

In contrast, data based models neglecting all mechanistic information suffer from the loss of extrapolation as well as a very high demand on data, which is required for reliable modelling of high dimensional systems. These drawbacks can be cured using structured functional models which provide a systematic approach for integration of mechanistic sub-models, a priori knowledge of system structures, and data based modelling. Functional networks provide a systematic, gradual interpolation technique on a mesoscopic scale, between fully mechanistic and pure black box modelling.

A first step towards reconstruction of functional networks from combinatorial data – the inverse problem – has recently been presented. We will discuss a direct reconstruction algorithm allowing us to unravel meso-scale network structures of biological systems from combinatorial input-output data. Meso-scale networks allow a rough representation of the topological structure of biological mechanisms. Although they lack of detailed understanding, they provide an unbiased overview over the activated pathways and their interaction. As these data can be achieved from high throughput screening experiments, the meso-scale analysis of the functional structure provides a complementary approach to established network reengineering methods and provides insight into systems of very high complexity.

Examples from gene diagnostics and drug response analysis demonstrate the application of direct inverse problem algorithms in biomedicine.