ABSTRACTMembrane proteins play essential role in cellular mechanisms. Despite that and the major progress in experimental structure determination, they are still significantly underrepresented in Protein Data Bank. Thus, computational approaches to protein structure determination, which are important in general, are especially valuable in the case of membrane proteins and protein-protein assemblies. Due to a number of reasons, not the least of which is much greater availability of structural data, the main focus of structure prediction techniques has been on soluble proteins. Structure prediction of protein-protein complexes is a well-developed field of study. However, because of the differences in physicochemical environment in the membranes and the spatial constraints of the membranes, the generic protein-protein docking approaches are not optimal for the membrane proteins. Thus, specialized computational methods for docking of the membrane proteins must be developed. Development and benchmarking of such methods requires high-quality datasets of membrane protein-protein complexes. In this study we present a new dataset of 456 non-redundant alpha helical binary complexes. The set is significantly larger and more representative than previously developed ones. In the future, this set will become the basis for the development of docking and scoring benchmarks, similar to the ones developed for soluble proteins in the DOCKGROUND resource http://dockground.compbio.ku.edu.