Abstract: An improved data integration method is proposed, in order to solve the problem of inadequate use of data-related characteristics in the procedure of constructing gene regulatory networks based on data integration. It uses knockout data and perturbation data to make predictions respectively, gives different confidence to the predictions of the two experimental data according to their overlap degree, and priority to the part with higher overlap degree. It sorts the predictions in accordance with their confidence. To assess the performance of the two methods, Dream3(Dialogue for Reverse Engineering Assessments and Methods) datasets are used. Experimental results show that the overallscore of the improved method is 4.9% higher than the existing method.

Key words: gene regulation, data integration, knockout data, perturbation data, noise model, differential equation

摘要： 原有数据集成方法在基因调控网络构建中不能很好地利用数据之间的相关特性。为此，提出一种改进的数据集成方法。分别利用敲除数据和微扰数据进行预测，根据2种实验数据预测结果的重叠程度赋予不同的可信度，优先考虑重叠程度高的部分，按照可信度对预测结果进行排序。采用Dream3数据集与原有方法进行性能对比，实验结果表明，改进方法的总体性能比原有方法高出4.9%。

关键词: 基因调控, 数据集成, 敲除数据, 微扰数据, 噪声模型, 微分方程

CLC Number: 

• TP393