Promoter Prediction Based On Structural Properties Of DNA

Many sequencing projects have been finished in the past few years, and so it becomes increasingly important to annotate these sequenced genomes. One of the most challenging problems in genome annotation is the identification of the core promoter. The core promoter comprises the sequences that direct the initiation of transcription. Identification of core promoters is a key clue in understanding transcriptional regulation mechanism, and also the necessary steps in studying gene expression patterns and gene regulatory networks.The experiment method is expensive and time consuming in identifying the core promoters, and at the same time only a limited number of methods can mark inactive promoters. Ab initio prediction can be an alternative approach to locate promoter regions in a given genome sequence. Several algorithms have been developed to predict promoter. However, due to the diverse nature of promoter sequences, the genome-wide promoter prediction accuracy of existing prediction approaches can’t meet our needs very well. This is mainly because the information extracted by existing approaches can not effectively distinguish promoter and non-promoter sequences; At the same time, the existing approaches don’t take into account the TSS cluster in the prediction of promoter.In this paper, we present a Markov chain model based on the value of DNA structural properties, and use statistical models of promoter and non-promoter sequences, in order to better distinguish between promoter and non promoter. At the same time, we use the correlation between TSS cluster and the spectrum of structural properties to predict TSS cluster, and then use this region to weight the result of Markov chain model. This could offset the interference of neighboring sites and improve the overall effect.The empirical research show that our method is better than EP3and ProSOM, with strong generalization ability, and can achieve more balanceable effect in precision and recall, and it can achieve higher results than existing methods. Additionally, the results of our method on different chromosomes are the same, indicating that this method has strong stability.