Identification of plant messenger RNA polyadenylation sites using length-variable second order Markov model

In this paper we adopted a length-variable second order Markov model to identify plant messenger RNA poly(A) sites, and provided a common method that only relies on the experimental sequences. The efficacy of our model is showed up to 92% sensitivity and 79% specificity. This method is particularly suitable for the prediction of the poly(A) site which is lack of biological priori knowledge and has poor conservative signal characteristic, as well as for the identification of the alternative poly(A) sites in different genetic regions. Compared with other algorithms, generalized hidden Markov model needed the signal distributions and AdaBoost required the construction of signal features around the sites, our model is more versatile.