Parallel Stochastic Search for Protein Secondary Structure Prediction

Prediction of the secondary structure of a protein from its aminoacid sequence remains an important and difficult task. Up to this moment, three generations of Protein Secondary Structure Algorithms have been defined: The first generation is based on statistical information over single aminoacids, the second generation is based on windows of aminoacids –typically 11-21 aminoacids– and the third generation is based on the usage of evolutionary information. In this paper we propose the usage of naı̈ve Bayes and Interval Estimation Naı̈ve Bayes (IENB) –a new semi naı̈ve Bayes approach– as suitable third generation methods for Protein Secondary Structure Prediction (PSSP). One of the main stages of IENB is based on a heuristic optimization, carried out by estimation of distribution algorithms (EDAs). EDAs are non-deterministic, stochastic and heuristic search strategies that belong to the evolutionary computation approaches. These algorithms under complex problems, like Protein Secondary Structure Prediction, require intensive calculation. This paper also introduces a parallel variant of IENB called PIENB (Parallel Interval Estimation Naı̈ve Bayes).