ABSTRACT
In this contribution we introduce a statistical framework for modelling and analyzing network data. In particular present protein interaction network datasets are typically incomplete and noisy, and this needs to be incorporated into the analysis explicitly and from the outset. Here we show how this is done and suggest computational approaches based on multi-model inference which are suitable for inferring properties of the true network from present partial network datasets. This new approach is illustrated by applying it to data from the yeast protein interaction network.
