Most subcellular organelles are expected to be similar among different cell types; however, a recent study suggests a surprising amount of variation in the protein composition at the nuclear envelope. Therefore, to comprehensively identify proteins in subcellular organelles proteomics datasets may need to be generated from multiple cell types. In this chapter we describe a proteomics study that expanded the number of nuclear membrane proteins by 5-fold using a "subtractive" methodology in which a subcellular organelle is partially purified biochemically and partially purified in silico. The biochemical fraction of interest and a separate fraction enriched in proteins known to contaminate it, in this case nuclear envelopes and microsomes respectively, are first isolated and separately analyzed by mass spectrometry. For in silico purification, proteins appearing in both fractions are subtracted from the dataset in order to identify proteins that are unique to the organelle being investigated. This approach identified 67 novel putative nuclear envelope transmembrane proteins in rodent liver. Further analysis of their expression levels in other tissues indicates that several are preferentially expressed in liver cell types, which in turn predicts considerable variation in the nuclear envelope proteome among different cell types. Finally, we discuss several issues associated with confirming that these peptide-based identifications represent proteins that truly localize to the nuclear envelope. These studies have complicated rather than simplified our view of the nuclear envelope, but proteomics has set the stage for beginning to understand this highly complex subnuclear organelle.