Identification Of Nucleosome Positioning Dynamics In Multiple Samples

Nucleosomes are fundamental building units of primary chromatin structure. Nucleosome positioning refers to the relative position of DNA double helix with respect to histone octamer. The positioning has role in gene regulation through blocking the DNA sites at which proteins bind. Nucleosome positioning is dynamic with the cell line type and the cellular environment. At present, a few works have been done in identifying the nucleosome dynamics in two samples. However, it is regret that there is not a method can deal with the multi-sample problem yet. But we will inevitably encounter the problem that need to analyze nucleosome data of multi-samples (n?3), identifying the dynamics is very important in resolving epigenetic regulation mechanism.In this paper, we developed a new algorithm based on statistical model to identify the nucleosome dynamic regions in multiple samples by using the high-throughout sequencing data. The algorithm (Dimnp) consists of two main modules:calculation of nucleosome occupancy signal and identification of the dynamic regions among multiple samples. In the first module, we shift each read toward 3'direction to make it be nucleosome center, adjust the read length, remove the clonal signal, and then smooth signal. Finally we use fold change method to normalize occupancy signal. In the second module, we use chi-square test to identify the difference, and calculate the difference by single resolution, then use a P value threshold to obtain the dynamic regions. According to this algorithm, we developed two offline tools, so that people can use it in either Linux or Windows system.By drawing ROC curve and comparing to DANPOS, we evaluated the reliability and validity of Dimnp algorithm. The results showed that:Dimnp can accurately identify the dynamic regions in multiple (n?3) samples, the results between Dimnp and DANPOS maintained a good consistency. But compared to DANPOS, Dimnp can analyze multiple samples just by one time, so that it can recognize the dynamic changes more comprehensively and more efficiently.We applied Dimnp algorithm to analyze yeast genome dynamics. We analyzed four groups and the results showed that the dynamic regions significantly concentrated in the promoter region and telomere region. Besides, the genes associated with dynamic regions showed specific biological functions.In summary, we developed a new method to identify the nucleosome dynamic regions in multiple samples, established computing tools. A systematic evaluation showed that the method is feasible and has good accuracy. This paper provides methods and tools on the research of nucleosome dynamics in cell growth, differentiation, environment-specific response, cancer and other processes.