We surveyed 299 high resolution, non-homologous protein crystal structures
for alpha-helix lengths and capping preferences. We find that helices show
a preference to have close to an integral number of turns. Helices can be u
sefully subdivided into either "favoured length" with 6, 7, 10, 11, 13, 14,
17, 18, 21, 22, 24, 25, 28, 29 or 31 residues, or "disfavoured length" wit
h 8, 9, 12, 15, 16, 19, 20, 23, 26, 27 or 30 residues. Favoured length heli
ces have their N and C-caps on the same side of the helix so they can lie o
n the protein surface. There is no significant difference in amino acid pre
ferences at the N terminus between favoured and disfavored length helices.
At the C terminus, favoured length helices prefer non-polar side-chains at
C4 and polar amino acid residues at C2, while disfavoured length helices pr
efer non-polar amino acid residues at C2. There are strong periodic trends
in the likelihood of terminating a helix with a Schellman or alpha(L) C-cap
ping motif. These can be rationalised by the preference for a non-polar sid
e-chain at C3 with these motifs, favouring placing C3 on the buried side of
the helix. We suggest that algorithms aiming to predict helices or C-cappi
ng in proteins should include a weight for the helix length. (C) 1999 Acade
mic Press.