Multivariate analysis of G protein-coupled receptors

With the rapidly increasing amount of data of a bioinformatic nature, it is of great interest to find different methods to help overview, classify, identify outliers and in other ways extract information from the large amounts of data. In this paper, chemometric methods often applied in statistical process control are utilized to analyse a large number of G protein-coupled receptors. They have been investigated with respect to groupings among the sequences based on their physicochemical properties using multivariate methods. The transmembrane (TM) regions of 897 receptors were examined. The sequences were multivariately characterized using principal properties for amino acids. The methods used include principal component analysis (PCA), hierarchical principal component analysis (HPCA) and partial least squares projections to latent structures (PLS). The results show that groups of receptors belonging to the same functional class can be identified using this approach, and that the rhodopsin class of G protein-coupled receptors is very different froth other classes present in the study with regard to their physicochemical properties. Copyright (C) 2003 John Wiley Sons, Ltd.