A major challenge of systems biology is to infer biochemical interactions from large-scale observations, such as transcriptomics, proteomics and metabolomics. Inferred interactions are combined to yield biochemical network models.

The main idea behind network inferrence from large-scale observations is rather simple: 1) establish cause effect relationships between genes, using for example genetic perturbations and then 2) distinguish between direct and indirect causal effects. The final network model contains only links that correspond to direct causal effects.

Several methods for inference of such networks from experimental data have been proposed in the recent literature. Sometimes it is assumed that biochemical networks can be modeled as directed acyclic graphs (DAGs) which makes techniques such as Bayesian Networks applicable. However, cyclic network structures, such as feedback loops, are ubiquitous in biology and are associated with many of the specific properties of living systems, and therefore analyses should be independent of such assumptions.

Our main focus is on developing algorithms for gene network inference by integrative analysis of gene expression and genotyping data, so called Genetical Genomics. In this setup genotypes can be seen as multifactorial perturbations causing variation in gene expression levels which can be used to establish cause effect relationships. We employ techniques such as Gaussian Graphical Models and the related technique of Structural Equation Modeling. In contrast to Bayesian Networks these approaches do not rely on the incorrect assumption that biochemical networks are acyclic.