Membrane proteins play important roles in cell survival and cell communication, as they functionas transporters, receptors, anchors and enzymes. They are also potential targets for drugs that block receptors orinhibit enzymes related to diseases. Although the number of known structures of membrane proteins is still smallrelative to the size of the proteome as a whole, many new membrane protein structures have been determinedrecently.Scope of the article:We compared and analyzed the widely used membrane protein databases, mpstruc,Orientations of Proteins in Membranes (OPM), and PDBTM, as well as the extended dataset of mpstruc based onsequence similarity, the PDB structures whose classificationfield indicates that they are“membrane proteins”and the proteins with Structural Classification of Proteins (SCOP) class-f domains. We evaluated the relationshipsbetween these databases or datasets based on the overlap in their contents and the degree of consistency in thestructural, topological, and functional classifications and in the transmembrane domain assignment.Major conclusions:The membrane databases differ from each other in their coverage, and in the criteria that theyuse for annotation and classification. To ensure the efficient use of these databases, it is important to understandtheir differences and similarities. The establishment of more detailed and consistent annotations for the se-quence, structure, membrane association, and function of membrane proteins is still required.General significance:Considering the recent growth of experimentally determined structures, a broad survey andcumulative analysis of the sum of knowledge as presented in the membrane protein structure databases can behelpful to elucidate structures and functions of membrane proteins. We also aim to provide a framework forfuture research and classification of membrane proteins.