Toward large-scale computational prediction of protein complexes

Abstract

Cellular functions are often performed by multiprotein structures called protein complexes. These complexes are dynamic structures that evolve during the cell cycle or in response to external and internal stimuli, and are tightly regulated by protein expression in different tissues resulting in quantitative and qualitative variation of protein complexes. Advances in high-throughput techniques, such as mass-spectrometry and yeast two-hybrid provided a large amount of data on protein–protein interactions. This sparked the development of computational methods able to predict protein complex formation under a variety of biological and clinical conditions. However, the challenges that need to be addressed for successful computational protein complex prediction are highly complex. The post-genomic era saw an emerging number of algorithms and software, which are able to predict protein complexes from protein–protein interaction networks and a variety of other sources. Despite the high capacity of these methods to qualitatively predict protein complexes, they could provide only limited or no quantitative information of the predicted complexes. Recently, a new large-scale simulation of protein complexes was able to achieve this task by simulating protein complex formation on the proteome scale. In this chapter, we review representative methods that can predict multiple protein complexes at different scales and discuss how these can be combined with emerging sources of data in order to improve protein complex characterization.