Parallelization of Self-Adapting Hidden Markov Model

摘要

Biological sequence analysis is a problem of complex giant system. Though our proposed self-adapting hidden Markov model solved the training problem of the model by means of unaligned sequences, because of our applying a two-stage optimization (structure optimization and parameter optimization) algorithm for it, but it is impossible to implement on serial machine when the number and lengths of sequences increase. In order to make the self-adapting hidden Markov model can suitable in practical sequence analysis of biological macromolecules, from which to mine useful information in biological sequences, in this paper we use MPI library in the environment of WINNT, and apply parallel algorithm of the master-slave mode integrating with C language to implement the parallelization of self-adapting hidden Markov model. Such a parallelization is very effect on multiple sequences alignment.