Statistical and Machine Learning Approaches for Network Analysis

4.2 Methods

4.2.1 General Approach

Interactome filtering may involve a selection of interaction data in terms of biological processes; as a result, a specific stratified interactome analysis could be performed. The advantage of such data disaggregation that we call “PIN fragmentation” is that the complexity levels of usually aggregated data would be avoided in favor of a reduction of dimensionality. The loss of information inherent to the integrated data may be balanced by the fact that the specific information layers that are needed may be promptly used. Thus, signaling or cell cycle or other processes can be examined once they have been retrieved, and can always be compared or combined, if needed.

Recent works [10–12] has inspired our approach. These authors compare topological characteristics of protein and metabolic filtered PIN from Escherichia coli and Saccharomyces cerevisiae model organisms, while we look specifically at cell cycle PIN data for this work, and plan to extend the fragmentation analysis in parallel studies. We thus built a compilation of filtered PIN from the yeast reference datasets of Reguly et al. [13], and used as a benchmark the literature-curated interactome (LIT-Int) obtained from small-scale experiments. Based on the reference dataset (say, rPIN) we applied a PIN fragmentation to extract subinteractomes specialized by cell biological processes.

The focus on the cell cycle process has the following rationale. Starting from the available fragmented PIN ...