Application of Hidden Semi-Markov Model to 3′ Splice Sites Identification

FENG Xiu-Cheng, QIAN Min-Ping, DENG Ming-Hua, MA Xiao-Tu, YAN Xi-Ting

(School of Mathematical Sciences, Peking University, Beijing 100871, China； Center for Theoretical Biology, Peking University, Beijing 100871,China)

Abstract: In order to improve exon level sensitivity and specificity of recent gene-finding programs, strong “search by signal” components are needed to identify splice sites, translation start and other biological signal sites. A new model for the identification of 3′ splice sites (acceptors) using Hidden Semi-Markov Model (HSMM) was introduced. This model is proved to be particularly suitable for modeling the biological structure of acceptors. When tested in Burset/Guigo dataset, this new method demonstrated an improved accuracy compared with existing method. The success of this model gives a deep understanding of the structure of acceptors and the biological process of splicing.

Key words:  Hidden Semi-Markov Model, splicing, eukaryotic gene structure prediction, EM algorithm

2004,31(5):455-458