Clustering of physico-chemical properties of amino acids
Jean-Luc Falcone, Paul Albuquerque · 2004
The data obtained from the sequencing of organisms necessitates the perfecting of methods of predicting the function and structure of proteins. For these methods to be effective, they must be based on the physico-chemical properties of the 20 amino acids (mass, charge, etc.) of which there are currently 484. Many of these properties are very strongly correlated. To regroup them, a measurement derived from the correlation is proposed. The physico-chemical properties are classified according to this distance, using the k-means and global k-means methods. A quality index has enabled the determination of of the 'natural' numbers of the groups. The results of the aggregations conform to the biochemical expectations. The centres of the groups thus produced constitute a new group of properties.