The attachment of a glycosylphosphatidylinositol (GPI) is one of the most important posttranslational protein modifications and plays an important role in many vital reactions. It is well known that most GPI-anchored proteins (GPI-APs), such as receptors, nucleotidases, proteinases, immune antigens, and prion protein, are associated with serious human diseases. Consequently, the annotation of novel genes encoding GPI-APs in mammalian genomes, including the human genome, is warranted. A GPI-AP database of candidate genes is also required to accommodate the accumulation of genes related to human diseases. In this study, candidate GPI-APs in mammalian genomes were comprehensively predicted with a computational method based on their hydropathy profiles, residue sizes, and position-specific amino acid propensities. We have created a cDNA library of GPI-AP candidate genes related to refractory diseases to confirm the prediction results and have developed a GPI-AP database containing clones of the candidate genes to support drug discovery and development.