Estimating Hidden Influences in Metabolic and Gene Regulatory Networks

We address the applicability of blind source separation (BSS) methods for the estimation of hidden influences in biological dynamic systems such as metabolic or gene regulatory networks. In simple processes obeying mass action kinetics, we find the emergence of linear mixture models. More complex situations as well as hidden influences in regulatory systems with sigmoidal input functions however lead to new classes of BSS problems. The field of independent component analysis (ICA) as solution to BSS problems has been in the focus of rather intense research during the past decade [1]. With the nowadays available robust algorithms in particular for the linear case, more and more people turn towards model generalizations and applications of ICA. One area of application of ICA and machine learning in general has been bioinformatics [2], which mostly deals with the analysis of large-scale high-throughput data sets from genomics. With the basic methods being robustly established, a trend in ...

• Type: Journal
• Year: 2009
• Where: IDA
• Authors: Florian Blöchl, Fabian J. Theis
• Keywords: BSS Problems, IDA 2009, Information Technology, Mass Action Kinetics, Standard Bss Models