Yule Value Tables from Protein Datasets

Here, we studied systematically the association between amino acids, the constituents of protein sequences in datasets of different hierarchy, i.e. genome (human), protein type (membrane proteins), protein family (specific types of membrane receptors and transporters) and transmembrane helices versus loops (either for membrane proteins in general or family-specifically). Association was estimated using Yule’s Q statistics for pairs of amino acids within a window of size 4. Strong association between such nearby amino acids was observed in all the datasets studied, in contrast to the randomized datasets. Association strength increased as expected when the datasets were more specific. Strikingly, in transmembrane helices, associations were more negative than in any other dataset studied, suggesting that evolution of these helices requires suppression of occurrence of specific amino acid combinations within local range. The results have direct applicability to several areas of bioinformatics research, i.e. transmembrane helix boundary prediction, sequence alignment and understanding of design principles of membrane proteins in general. Data and access to the algorithms presented in this paper are available at http://flan.blm.cs.cmu.edu/