As for all proteins, there are three basic problems facing bioinformatics in the field of membrane proteins: to reliably identify membrane proteins in genome-wide data sets, to predict as much as possible about their structure, and to predict as much as possible about their function. Function prediction will not be further treated here since the methods used (sequence alignments, phylogenetic profiles) are not particular to membrane proteins. The other two areas, however, require a different set of methods than those normally applied to soluble proteins. As a background to the prediction problem, the structural principles inherent to integral membrane proteins and some elements of the cellular mechanisms responsible for their biogenesis will be reviewed in Sects. 2.2 and 2.3, followed by a discussion of current methods for discriminating between membrane proteins and soluble proteins, for predicting the topology of membrane proteins, and for modeling their 3D structure when no homologues with know structure are known. For more thorough reviews of the membrane protein field, see, e.g., von Heijne (1999, 2000).