Classification of proteins based on similarity of two-dimensional protein maps

Data reduction techniques are now a vital part of numerical analysis and principal component analysis is often used to identify important molecular features from a set of descriptors. We now take a different approach and apply data reduction techniques directly to protein structure. With this we can reduce the three-dimensional structural data into two-dimensions while preserving the correct relationships. With two-dimensional representations, structural comparisons between proteins are accelerated significantly. This means that protein-protein similarity comparisons are now feasible on a large scale. We show how the approach can help to predict the function of kinase structures according to the Hanks' classification based on their structural similarity to different kinase classes.