We defined alpha/beta proteins as proteins with at least 20% of the residues in a helix, and all beta proteins as those that have less than 20% of the residues in a helix. There are several reasons why we did not use 0% vesus not 0% as criterion. When using completely automated ways to determine secondary structure (such as dssp), quite often a few residues are labelled as being in a helix when in truth they aren't. Even when looking at the 3D structure itself, there might be some ambiguity whether or not certain residues form a helix. And even if they do, if only 5 out of 150 residues are in a helix, should this protein really be classified as alpha/beta? From previous studies, we found the 20% cutoff to be a reasonable choice. Of course it is debatable whether or not this choice is optimal, but the database and scores were derived using this definition.