Deciphering the structural code for proteins: helical propensities in domain classes and statistical multiresidue information in alpha-helices

Abstract

We made several statistical analyses in a large sample of nearly 4,000 helices (from 546 redundancy-controlled PDB protein subunits), which give new insights into the helical properties of globular proteins. In a first experiment, the amino acid composition of the whole sample was compared with the composition of two helical sample subgroups (the "mainly-alpha" and the "(alpha/beta)8 barrel" domain classes); we reached the conclusion that composition-based helical propensities for secondary structure prediction do not depend on the structural class. Running a five-residue window through the whole sample, the positional composition revealed that positive and negative residues are located throughout the helices and tend to neutralize the macrodipole effect. On this basis, we analyzed charged triplets using a running five-residue window. The conclusion was that only mixed charged residues [positive (+) and negative (-)] located at positions 1-2-5 and 1-4-5 are clearly favored. In these locations the most abundant are (- -..+) and (-..+ +), and this shows the existence of side chain microdipoles, which neutralize the large macrodipole of the helix. We made a systematic statistical analysis of charged, dipolar, and hydrophobic + aromatic residues, which enabled us to work out rules that should be useful for modeling and design purposes. Finally, we analyzed the relative abundance of all the different amphipathic double-arcs that are present in helices formed by octapeptides (8) and nonapeptides (18). All of the double-arcs that make up Schiffer and Edmundson's classical helical wheel are found in abundance in the sample.