The Prediction of Amphiphilic α-Helices

A number of sequence-based analyses have been developed to identify protein segments, which are able to form membrane interactive amphiphilic α-helices. Earlier techniques attempted to detect the characteristic periodicity in hydrophobic amino acid residues shown by these structure and included the Molecular Hydrophobic Potential (MHP), which represents the hydrophobicity of amino acid residues as lines of isopotential around the α-helix and analyses based on Fourier transforms. These latter analyses compare the periodicity of hydrophobic residues in a putative α-helical sequence with that of a test mathematical function to provide a measure of amphiphilicity using either the Amphipathic Index or the Hydrophobic Moment. More recently, the introduction of computational procedures based on techniques such as hydropathy analysis, homology modelling, multiple sequence alignments and neural networks has led to the prediction of transmembrane α-helices with accuracies of the order of 95percent and transmembrane protein topology with accuracies greater than 75percent. Statistical approaches to transmembrane protein modeling such as hidden Markov models have increased these prediction levels to an even higher level. Here, we review a number of these predictive techniques and consider problems associated with their use in the prediction of structure / function relationships, using α-helices from G-coupled protein receptors, penicillin binding proteins, apolipoproteins, peptide hormones, lytic peptides and tilted peptides as examples.