A neural network has been used to predict both the location and the type of
beta-turns in a set of 300 nonhomologous protein domains. A substantial im
provement in prediction accuracy compared with previous methods has been ac
hieved by incorporating secondary structure information in the input data.
The total percentage of residues correctly classified as beta-turn or nor-b
eta-turn is around 75% with predicted secondary structure information. More
significantly, the method gives a Matthews correlation coefficient (MCC) o
f around 0.35, compared with a typical MCC of around 0.20 using other beta-
turn prediction methods. Our method also distinguishes the two most numerou
s and well-defined types of beta-turn, types I and II, with a significant l
evel of accuracy (MCCs 0.22 and 0.26, respectively).