HMM sampling and applications to gene finding and alternative splicing

被引:44
作者
Cawley, Simon L. [2 ]
Pachter, Lior [1 ]
机构
[1] Univ Calif Berkeley, Dept Math, Berkeley, CA 94720 USA
[2] Affymetrix, Emeryville, CA 94608 USA
关键词
suboptimal parses; sampling; hidden Markov model; conserved alternative splicing;
D O I
10.1093/bioinformatics/btg1057
中图分类号
Q5 [生物化学];
学科分类号
071010 ; 081704 ;
摘要
The standard method of applying hidden Markov models to biological problems is to find a Viterbi (maximal weight) path through the HMM graph. The Viterbi algorithm reduces the problem of finding the most likely hidden state sequence that explains given observations, to a dynamic programming problem for corresponding directed acyclic graphs. For example, in the gene finding application, the HMM is used to find the most likely underlying gene structure given a DNA sequence. In this note we discuss the applications of sampling methods for HMMs. The standard sampling algorithm for HMMs is a variant of the common forward-backward and backtrack algorithms, and has already been applied in the context of Gibbs sampling methods. Nevetheless, the practice of sampling state paths from HMMs does not seem to have been widely adopted, and important applications have been overlooked. We show how sampling can be used for finding alternative splicings for genes, including alternative splicings that are conserved between genes from related organisms. We also show how sampling from the posterior distribution is a natural way to compute probabilities for predicted exons and gene structures being correct under the assumed model. Finally, we describe a new memory efficient sampling algorithm for certain classes of HMMs which provides a practical sampling alternative to the Hirschberg algorithm for optimal alignment. The ideas presented have applications not only to gene finding and HMMs but more generally to stochastic context free grammars and RNA structure prediction.
引用
收藏
页码:II36 / II41
页数:6
相关论文
共 34 条
[1]   SLAM: Cross-species gene finding and alignment with a generalized pair hidden Markov model [J].
Alexandersson, M ;
Cawley, S ;
Pachter, L .
GENOME RESEARCH, 2003, 13 (03) :496-502
[2]   ON KTH BEST POLICIES [J].
BELLMAN, R ;
KALABA, R .
JOURNAL OF THE SOCIETY FOR INDUSTRIAL AND APPLIED MATHEMATICS, 1960, 8 (04) :582-588
[3]   COMPUTING THE N-BEST LOOPLESS PATHS IN A NETWORK [J].
CLARKE, S ;
KRIKORIAN, A ;
RAUSEN, J .
JOURNAL OF THE SOCIETY FOR INDUSTRIAL AND APPLIED MATHEMATICS, 1963, 11 (04) :1096-1102
[4]  
Durbin R., 1998, BIOL SEQUENCE ANAL
[5]  
Eddy S R, 1995, Proc Int Conf Intell Syst Mol Biol, V3, P114
[6]  
Fox B.L., 1973, INFOR CANADIAN J OPE, V11, P66
[7]  
GUSFIELD D, 1992, P 3 ANN ACM SIAM JOI
[8]   LINEAR SPACE ALGORITHM FOR COMPUTING MAXIMAL COMMON SUBSEQUENCES [J].
HIRSCHBERG, DS .
COMMUNICATIONS OF THE ACM, 1975, 18 (06) :341-343
[9]   A METHOD FOR THE SOLUTION OF THE NTH BEST PATH PROBLEM [J].
HOFFMAN, W ;
PAVLEY, R .
JOURNAL OF THE ACM, 1959, 6 (04) :506-514
[10]  
Horn R. A., 1986, Matrix analysis