基于微阵列数据的基因调控网络构建方法研究

发布时间：2018-08-21 14:08

【摘要】：基因调控网络是指细胞内基因之间的相互作用形成的调控网络,它是生物体内控制基因表达的主要机制。构建基因调控网络是理解生命活动本质的重要手段之一,因此利用高通量实验数据,特别是微阵列数据构建基因调控网络已成为系统生物学领域的研究热点。但目前已有的基于微阵列的基因调控网络构建方法大多存在着无法确定调控方向或计算复杂度过高等问题,因此,本文提出了一种结合已有的相关性检验方法和常微分方程建模方法的基因调控网络构建方法。该方法首先利用扰动实验数据计算基因之间的皮尔森相关系数,然后通过Z分数排序的方法构建一个初始的基因调控网络。在此基础上,使用时间序列数据和常微分方程建模的方法对这个初始调控网络进行优化。在常微分方程模型建立之后,基因网络推导问题就转变成了一个模型参数估计问题。本文提出了一种内嵌禁忌搜索的粒子群优化算法,即禁忌粒子群优化来估计模型参数。在进行模型参数估计时,为了降低计算复杂性,首先利用曲线拟合方法对时间序列表达谱数据进行曲线拟合,估计出每一个时间点的微分,这样微分方程的参数估计问题就转变为一个伪多元线性回归的问题,计算时间大大降低。最后,我们使用标准测试集和真实的微阵列数据对本文所提出的方法进行了实验验证。结果表明,本文所提出的算法在构建基因调控网络时,识别的敏感性、特异性和精确性等方面均较已有方法有所提高,同时,本文方法的计算速度也较已有方法快。
[Abstract]:Gene regulatory network is a regulatory network formed by the interaction of genes in cells. It is the main mechanism of controlling gene expression in organisms. The construction of gene regulatory networks is one of the important means to understand the nature of life activities. Therefore, the construction of gene regulatory networks using high-throughput experimental data, especially microarray data, has become a hot research topic in the field of system biology. However, most of the existing methods of constructing gene regulation network based on microarray have some problems, such as the direction of regulation can not be determined or the computational complexity is too high. In this paper, a method of constructing gene regulatory network is proposed, which combines the existing correlation test method and ordinary differential equation modeling method. Firstly, the Pearson correlation coefficient between genes is calculated by perturbing experimental data, and then an initial gene regulation network is constructed by Z-score sequencing method. On this basis, the initial control network is optimized by using time series data and ordinary differential equation modeling method. After the ordinary differential equation model is established, the problem of gene network derivation is transformed into a model parameter estimation problem. In this paper, a Tabu search based particle swarm optimization (PSO) algorithm is proposed to estimate the model parameters. In order to reduce the computational complexity, the time series expression spectrum data are first fitted by curve fitting method, and the differential of each time point is estimated. In this way, the parameter estimation problem of differential equations is transformed into a pseudo-multivariate linear regression problem, and the computational time is greatly reduced. Finally, we use the standard test set and the real microarray data to verify the proposed method. The results show that the algorithm proposed in this paper is more sensitive, specific and accurate than the existing methods in the construction of gene regulation network. At the same time, the computational speed of the proposed algorithm is faster than that of the existing methods.
【学位授予单位】：东北师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：Q811.4

【相似文献】