Relationship of the enzyme functional classes with the statistical attributes of their secondary structures

Rekha Iyer1, Sudhir Kumar2, Department of Biology, Arizona State University;, Center for Evolutionary Functional Genomics, Arizona Biodesign Institute,and the Department of Biology, Arizona State University

We examined the similarities and differences in the number, size, composition and other statistical attributes of secondary structural elements in different functional classes of enzymes. Specifically, the frequency of alpha helices and beta sheets per polypeptide chain, the alpha helix and beta strand length, the number and type of beta strands per beta sheet, and the amino acid content of alpha-helix and beta-strand termini were studied in 105 human enzymes. The results revealed enzyme-class-specific distinctions of some of these attributes, including significantly lower alpha-helical content in hydrolases, significantly greater number of parallel beta bridges per ladder in oxidoreductases and antiparallel beta bridges per ladder in transferases, and preferences for certain amino acids at alpha helix and beta strand termini. On testing the potential of these secondary structural patterns in function prediction using neural networks, an accuracy of 73% was obtained for hydrolases and transferases and 88% for oxidoreductases. The neural networks were also tested with seven human and bacterial proteins, 142 yeast proteins of known function, and 104 hypothetical proteins by using predicted secondary structures. These trials were moderately successful in identifying protein function class. This method of functional type annotation may be useful due to its computational simplicity and may be used in conjunction with structure prediction algorithms to facilitate broad functional annotation of unknown proteins and for validation purposes.