Eugene van Someren, Lodewijk Wessels and Marcel Reinders
Information and Communication Theory Group, Delft University of Technology
Genetic network modeling is the field of research that tries to find the underlying network of gene-gene interactions from time-course gene expression data. Interactions between genes are represented by the influence a gene's expression level exerts on the expression level of the genes it controls. Knowing this network of interactions provides a wealth of information, such as in which processes a gene is involved, the function it has in these processes, which genes are the initiators of certain pathways, what is the influence of deleting a certain gene, etc. However, the algortihms for learning conventional network models cannot be as easily applied to microarray data as was the case with clustering and classification techniques. Genetic network models suffer greatly from the dimensionality problem, i.e. the fact that the number of measured genes by far exceeds the number of measured time-points. This fundamental problem causes the results to be inaccurate and unreliable. Fortunately, research is being directed towards extracting as much information as possible from these "limited" datasets and solutions are being proposed that alleviate this problem by introducing biologically motivated constraints. In this talk we will introduce genetic network modeling, briefly explain the underlying difficulties and present up-to-date developments in finding reliable regulatory interactions.