Mapping the InParanoid datasets onto paraclusters it is apparent that there is almost complete turnover of paracluster gene members across different kingdoms and phylum, implicating many paracluster genes as species or clade specific. Even though 21% of human paraclusters SU5416 contained at least one out-paralog in common with paraclusters of nonvertebrate species, only 1.6% of human clusters had two or more out-paralogs in common. This suggests that when paraclusters sharing a common ancestry are both undergoing expansion in two distantly related species, these expansions are largely independent processes in the two species. This agrees with our observation that vertebrate paraclusters are highly distinct in gene content from those found in the invertebrates, plants, and fungi. This is true even though some functional classes are conserved and some of the most prolific superfamilies are commonly present across all the species we tested.. The immunoglobulin superfamily domain is one of the most versatile XL-184 c-Met inhibitor domains in the genome, leading to a variety of functions including cell-cell recognition, cell-surface receptors, muscle structure and, in higher organisms, the immune system. This diversity emerges from the combinations and arrangements within genes of four subtypes of immunoglobulin domains: variable, constant-1, constant-2 and intermediate and paraclusters of the immunoglobulin superfamily consist of genes with highly variable combinations of these subtypes. Paraclusters of the immunoglobulin superfamily are particularly unique in their genomic arrangements, consisting of large clusters, up to 5 Mb in size, consisting of highly diversified family members with many unrelated genes interspersed among them. The diversity of these genes within these paraclusters is great enough to make detection of the paralogous relationships among them difficult when only using sequence similarity approaches; they are more readily detected using the domain datasets. To better understand how these gene clustering results based on structural annotation data differ from the results of sequence similarity approaches, we compared human paraclusters with human datasets downloaded from the Duplicated Gene Database which groups genes with sequence similarity by proximity with no applied measurement of statistical significance. There were 479 clusters reported by the DGD that were not detected as paraclusters because they were either too distantly spaced or lacked statistical significance. Additionally, there were 204 groups within DGD that contained additional genes not found within paraclusters, primarily as a result of additional genes positioned too distal for detection. Reversing the direction of comparison, we found 236 paraclusters not included in the DGD; 132 of these involved uncharacterized genes, most likely not included in the DGD datasets which are based on build 65 of Ensembl, and 22 clusters involved annotation errors.