## Introduction to BioNet

The integrated analysis of microarray data in the context of biological networks (e.g protein-protein interaction networks) has become a major technique in systems biology. The primary objective of integrated network analysis is the identification of functional modules (significantly differentially expressed subnetworks) within large networks. Therefore, the nodes of the network have to be weighted by a score according to the functional relevance of each gene product and subsequently a proper network search algorithm is required to find the maximum-scoring subgraph (MSS). Recently, we have devised an algorithm to this NP-hard problem, that computes provably optimal and suboptimal solutions to the MSS problem in reasonable running time using integer linear programming or a fast heuristic approach. The BioNet package allows the scoring of the network by a modular scoring function based on signal-noise decomposition of the p-values. Multiple p-values, derived from various sources, can be aggregated beforehand into one p-value using aggregation statistics. Based on the scores provably optimal and suboptimal solutions can be calculated and the resultant modules can be visualized in 2D or 3D.

## Figure: 3D visualization of the resulting module

The figure shows the resulting module in a 3 dimensional plot for an example data set on diffuse large B-cell lymphomas, used in the study of Dittrich et al. (2008). Differential expression is depicted by node colouring (red: upregulated in tumor subclass I (GBC), green: upregulated in tumor subclass II (ACB)). Known disease-relevant modules (shaded) (Rosenwald et al., 2002) are captured and extended by the integrated network approach.