Protein alpha -helix region prediction based on stochastic-rule learning

The authors apply their previously introduced (1992) method (the MY method) to alpha -helix region prediction for a variety of proteins which are randomly selected from the Brookhaven Protein Data Bank. The MY method produces a stochastic rule which assigns, to any region in a sequence, the probability that it is an alpha -helix region. Optimal stochastic rules are obtained by using Laplace estimation of real-valued parameters and the minimum description length principle. The experimental results show that the MY method achieved an average prediction accuracy rate of more than 70%, on more than 3000 residues in the test sequences, even when only hemoglobin sequences were used to generate examples of alpha -helix regions.