Other
A Bidirectional Bayesian Monte Carlo Approach for Estimating Parameters of a Profile Hidden Markov Model
Applied Science Segment, Vol.1(2)
Segment Journals
2010
Abstract
Hidden Markov Models (HMMs) are statistical models frequently applied for modeling biological sequences. A Profile Hidden Markov Model (PHMM) is one of the applications of HMMs in bioinformatics. Estimating parameters of PHMM is one of the main challenges in applying them to real data in bioinformatics. In this paper, we give a brief review of two approaches named: the Baum-Welch algorithm and Bayesiyan Monte Carlo Markov Chain (BMCMC) method, for estimating parameters of PHMM. The common method for estimating parameters of HMMs usually considers the left side information of each observation. In order to improve the prediction accuracy of HMMs, in this paper we consider information on both sides of residues in a sequence for parameter estimation. The results show that using information on both sides of residues enables us to compare different methods for parameter estimation, more precisely. It is concluded that BMCMC method performs better than the Maximum Likelihood estimation. We also compare our results with those obtained when the Left-to-Right PHMM is applied.
Details
- Title
- A Bidirectional Bayesian Monte Carlo Approach for Estimating Parameters of a Profile Hidden Markov Model
- Authors/Creators
- Rosa Aghdam - Shahid Beheshti UniversityHamid Pezeshk - University of TehranSeyed Amir Malekpour - Université de MontréalSoudabeh Shemehsavar - Murdoch University, College of Science, Technology, Engineering and MathematicsMehdi SadeghiChangiz Eslahchi - Shahid Beheshti University
- Publication Details
- Applied Science Segment, Vol.1(2)
- Publisher
- Segment Journals
- Identifiers
- 991005728684507891
- Murdoch Affiliation
- College of Science, Technology, Engineering and Mathematics
- Language
- English
- Resource Type
- Other
Metrics
18 File views/ downloads
10 Record Views