A novel, non-statistical method for predicting breaks in transmembrane helices.

We have developed a novel, non-statistical procedure for predicting possible breaks in transmembrane helices based on energy calculations. The procedure consists of stepwise elongation of the 'core' helical fragment determined by consensus results of several available prediction procedures. At each step, we calculate the conformational energies corresponding to the regular 'frozen' helical conformer of the 'core' fragment elongated by two flanking residues, E(alpha), as well as those to several options for the fragment to enter or exit the helix by changing conformations of the flanking residues, Ei. The minimal values out of Ei - E(alpha), delta(k), can be viewed as a profile of relative energies, where each minimum of delta(k) is a signal to start or to stop transmembrane helix. We suggest that boundaries of the transmembrane helix would be determined by the signals closest to the 'core' sequence in the delta(k) profiles. Our procedure was applied to prediction of the N- and C-termini for 45 transmembrane helices from the photosynthetic reaction center from Rhodopseudomonas viridis, bacteriorhodopsin and the cytochrome c oxidase from Paracoccus denitrificans. The results clearly showed that it is significantly more probable that a prediction accuracy within an error of +/- 2 residues will be obtained by our procedure than by three different statistical approaches.