Predicting peptide-receptor, peptide-protein, and chaperone-protein binding using patterns in amino acid hydrophobic free energy sequences

Much of the current study of protein organization is aimed at understanding emergent polymeric structure in three rather than one dimensions. This is largely because the weak bonds between amino acid side chains in one-dimensional peptide heteropolymers that determine their equilibrium tertiary structures involve large loop interactions between sequentially distant sites. For this reason, searches for sequential patterns seem intuitively irrelevant even though the coding for 3D protein structure is present in the 1D peptide chain, On the other hand, we have found that there are particular circumstances in which matches in sequential patterns in amino acid side chain thermodynamic properties postdict signatory modularity in the tertiary structure of protein families and polypeptide-protein interactions, such as peptide-receptor, peptide-membrane transporter, nuclear factor-protein docking, and chaperone-protein binding. Here we describe and justify our computational approach to matching sequential organization, with examples from experimentally demonstrable, polypeptide-protein interactions. Using established values for each amino acid's hydrophobic free energy, "hydrophobicity", in kcal/mol, derived from their normalized free energy of transfer from nonpolar to polar bulk phases of binary solutions,(34) we study the variations in hydrophobicity along primary sequences and seek matching patterns among them. We analyze this discrete data series using the all-poles power spectrum for shorter polypeptides. For longer polypeptides and proteins, including membrane receptors, relevant protein domains, nuclear factors, and chaperones, we first decompose the data using autocovariance matrices into orthogonal eigenvectors, compose these eigenvectors with the original series to generate eigenfunctions, and then compute their power spectra. We show examples of sequential pattern-matched peptides that postdict binding of a non-peptide (estrogen) receptor, protein (calcineurin) binding to its T-cell nuclear factor (NFAT) docking site, and chaperone (GroEL) binding to one of its target enzyme proteins (beta-lactamase). Polypeptide sequences and the associated proteins that bind them share distinguishing spectral features in their all poles power spectrum. This approach achieves practical significance because inversion of these analytic methods are now being used to successfully design de novo peptides which demonstrated their predicted physiological activity.