Approach used to characterize the similarity of amino acid descriptor sets. After normalization of all descriptor sets, the difference between a pair of descriptor sets was calculated. This difference was obtained as the difference between the distance separating a pair of AAs when represented by descriptor set 1, and the distance of the same pair when represented by descriptor set 2. This was done for all descriptor set pairs. Finally, the average difference was obtained and a full matrix was constructed, hence giving a measure of how similar different amino acid descriptors perceive amino acid structures to be.