They looked at the Pfamdatabase, a repository of the amino-acid sequences of 260, 000 protein domains, to find out how easily, using the 30% figure as a guide, these could be clustered into groups that were single units from an analytical standpoint.