The Study Of Generaliazed Increment Of Diversity Combining With Key Sites On Nuclesome Positioning Prediction Effect

Investigation of the nucleosome positioning across eukaryotic genomes is very meaningful to understand biological processes. Developing theoretical or computational methods for predicting nucleosome positioning along genome based on the massive experiment data becomes hot field of bioinformatics. Although nucleosome positioning theory has made some progress, the position of the nucleosome still cannot be accurately determined. We should improve the accuracy of nucleosome positioning by further experimental means.This thesis proposes the algorithm of Generalized Increment of Diversity, and the key sites on the core DNA, then presents some predicting nucleosome core DNA models around Generalized Increment of Diversity and the key sites combined with support vector machine (SVM). The main research is as follows:(1) The key sites are present in the thesis, then based on k-mer frequency in core DNA, this thesis raises a computational nucleosome predicting model based on the key sites combined with Support Vector Machine. The model is used to predict nucleosome DNA and linker DNA in S.cerevisiae. The model of key sites combined with ID-SVM obtains good performance with an average total accuracy of 72.56%.(2) Generalized Increment of Diversity is present in this thesis, then based on k-mer frequency in core DNA, this thesis proposes a computational nucleosome predicting model of Generalized Increment of Diversity combining with key sits, and model predicts nucleosome core DNA and linker DNA in S.cerevisiae. The gID-SVM model obtains good performance with average of accuracy of 85.35%, and the accuracy of model key sites combined with gID-SVM is 82.53%.(3) The prediction models are expanded to predict the nucleosome core DNA and linker DNA prediction in other eukaryote. Thesis uses key sites combined with Generalized Increment of Diversity to predict core DNA and linker DNA in chicken. The predication results are good. The accuracy of model key sites combined with ID-SVM is 89.03%, model gID-SVM’s accuracy is 93.13%, and model key sites combined with gID-SVM’s accuracy is 90.80%. Finally we analyze distribution of key sites in core DNA by comparison of core DNA in S.cerevisiae and chicken, and find some meaningful regulars:first, the eukaryote has university; second, the laws of the key sites are similar in different eukaryotes; the third one is the other sites on the core DNA except the key sites decide the diversity of species traits.