Research On Prediction And Visualization Of Genomic Islands

Genomic islands(GIs)are a large class of DNA segments with particular structure and function,which are usually horizontally transferred to bacteria by onjugation,conduction and transformation.GIs are characterized with their composition bias and functional feature,and their genes are associated with evolution and adaptation to niches,such as genes involved in pathogenesis and antibiotic resistance.Therefore,GI prediction become an important part of microbial genome analysis.In this dissertation,we proposed a robust prediction method and a visualization method for genomic islands,and constructed a comprehensive database of genomic islands.The main work is represented as follows:1.We summarized the prediction methods for genomic islands,and divided them into two groups.We then briefly described the representative methods in each group,and discussed their advantage and disadvantages,which provides powerful theoretical foundation for the following research.2.We proposed a robust method that used the multiscale statistical test to detect genomic island.The advantage of the algorithm is an interation of a small-scale t-test with large-scale feature selection to quantify compositional differences of a genome from that of a host.Then we introduced a multiscale segmentation algorithm to identify multi-window segments.At last,we refined the boundaries of predicted GIs using Markovian Jensen–Shannon divergence and the GC-based segmentation method.The proposed methods was evaluated with five experiments and compared it with the available competing prediction methods.The results indicate that the proposed method demonstrated better performance than existing methods,with more accurate size.3.We proposed a visualization method for genomic islands.First,we used the D3.js to design a more intuitive method to display predicted genomic islands.Then,we designed a software with graphical user interface and developed a Websever for the proposed prediction method,which can meet the requirements from all kinds of users.4.We constructed a comprehensive database of genomic islands.Hereon,we first introduced basic principle and methods of the database and collected all the data of genomic island.The aimof GIs-DB is to provide a prediction and visualization of genomic islands.