Study On Enhancer Based On Histone Modification Information And DNA Sequence Prediction

In recent decades, with the development of high-throughput sequencing technology and next-generation sequencing technology, more and more researches are focused on epigenetics. Regulation of gene expression has been the emphasis of epigenetics research, and is affected by many transcription factors, from the DNA transcription to the transcription of a series of protein late stage, then finally decides the cell growth process. Enhancer is an important and necessary cis-regulatory element, which can interact with promoter and silencers components, improve the efficiency of gene transcription and regulate gene expression. With the development and maturation of the ChIP-Chip and ChIP-Seq sequencing technology, more and more experiments began to study the relationship between enhancer and gene sequences, proteins and histone modifications. Studies have shown that the enhancer contained strong conserved sequence, and displayed specific map of many histone modifications. These histone modifications change the structural characteristics of chromosomes, thus affect the regulation of gene expression. To locate the site of the enhancer can help understand of the regulation of gene expression and promote the study of epigenetics.Because enhancers share sequence conservation, taking advantage of DNA sequence can contribute to improve the accuracy for predicting the conserved enhancer. On that account, we propose a new method which takes use of histone modification and DNA sequence information to predict enhancer. Chapter two introduces how to predict enhancer based on single SVM model, and analysis related prediction results. Consider some deficiencies of single SVM model, Chapter three improve the method. First select a training set include DHS sites. Then make use of random subspace method to construct many sub-classifiers and ensemble all classifiers results to predict enhancer.The algorithm not only consider the information of histone modification, also add the corresponding DNA sequence information to predict enhancer, then improve the accuracy of the prediction results on overlapping proportion of p300and conserved sequence. Our algorithm combines histone modifications and sequence information, and integrates a number of sub-classification to predict enhancer. To some extent, our method can reduce the noise interference, improve the stability of our model and predictive accuracy, and makes great contribution to predict the whole genome of enhancer.