Construction Of Gene Regulatory Networks Based On Data Integration

For half a century, with the development of the research, many mathematical modelswere applied to construct gene regulatory networks, including Boolean network model,Bayesian network model, and differential equation model and so on. And these modelshave made some achievements. In the1990s, the application of high-throughputbiotechnology, such as gene chips, produced vast amounts of biological data, creating morefavorable conditions for the research of gene regulatory network construction. How toeffectively use these data becomes a hot topic of recent research. Currently, a moreeffective way is to integrate different experimental data, and then model networks on thebasis of data integration. However, existing data integration methods are still with someshortcomings, such as inadequate use of data-related features.To solve the above problems, in this thesis, a research on gene regulatory networkconstruction based on data integration was done, and the following specific research workwas completed:1. Analyzed and summarized the advantages and disadvantages of the existingmathematical models used in the gene regulatory network construction;2. An Improved data integration method was proposed. It integrated gene knockoutand perturbation data. In detail, the predictions were assigned credibility according to theiroverlap degree, in order to take full advantage of the correlation characteristics betweendifferent kinds of data. With the improved method, the coverage rate of the networkconstructed was increased;3. For improved data integration methods established above, an algorithm based onFloyd algorithm was proposed to solve the cascade problem. Longest path as the indirectpath, the algorithm could effectively distinguish direct path from indirect path by comparing their probability, and the overall forecast performance was improved;4. With the methods proposed above, a gene regulatory construction experimentsystem was built, which based on data integration. As expected, the experiment resultsverified the effectiveness of the two methods provided.Finally, a summary of the thesis was made, and later research was prospected.