Clustering amino acids using maximum clusters similarity

In this paper, we present a clustering method for amino acids based upon some of their physico-chemical properties e.g. volume, area, hydrophilicity, polarity, hydrogen bonding, shape and charge. Given any two clustering algorithms, the number of clusters is determined by finding the partitions of the amino acid at which the clustering similarity is maximized. The clustering similarity is measured by a similarity index corrected for chance agreement. Memberships are then found by any of the clustering algorithms used to get the maximum similarity. Our clustering method was validated since it gives the same clusters as those obtained by Stanfel's method [1] when applied to their database. We have also shown that by including an additional physicochemical property, buriability, an improvement was made since the method gives five clusters by splitting the largest ten-member cluster into two five-member clusters. This method is easy to implement and can be applied to other chemical databases such as drugs to identify structural elements required for their bioactivity.