Regularization and Noise Injection for Improving Genetic Network Models
Regularization and Noise Injection for Improving Genetic Network Models
复制标题
DOI：
10.1007/0-387-26288-1_14
复制
发表时间：
2006
期刊：
影响因子：
--
通讯作者：
E. Someren;L. Wessels;M. Reinders;E. Backer
中科院分区：
文献类型：
--
作者：
E. Someren;L. Wessels;M. Reinders;E. Backer
遗传网络建模是一个研究领域，它试图从测量到的一组基因表达中找到基因-基因相互作用的潜在网络。到目前为止，已经提出了几种不同的建模方法，例如布尔网络(Leung等人，1998)、贝叶斯网络(Friedman等人，2000)、线性网络(van Someren等人，2000a；D‘haeseleer等人，1999)、神经网络(Weaver等人，1999；Wahde和Hertz，1999)和微分方程(Chen等人，1999 b)。在这些方法中，遗传交互作用由参数模型中的参数来表示，这些参数需要从随时间推移测量的基因表达来推断。目前的微阵列技术已经导致了可以在单个阵列上同时测量其表达的基因数量的显著增加。然而，在时间进程实验中进行的测量数量并没有以类似的方式增加。因此，典型的基因表达数据集由相对较少的时间点(通常少于20个)相对于基因的数量(数千)组成。这个所谓的维度问题和测量包含大量测量噪声的事实是遗传网络建模中最基本的两个问题。通常，当从病态数据(多基因、少时间样本)中学习遗传网络模型的参数时，解变为
Genetic network modeling is the field of research that tries to find the underlying network of gene-gene interactions from the measured set of gene expressions. Up to now, several different modeling approaches have been suggested, such as Boolean networks (Liang et al., 1998), Bayesian networks (Friedman et al., 2000), Linear networks (van Someren et al., 2000a; D’Haeseleer et al., 1999), Neural networks (Weaver et al., 1999; Wahde and Hertz, 1999) and Differential Equations (Chen et al., 1999b). In these approaches, genetic interactions are represented by parameters in a parametric model which need to be inferred from the measured gene expressions over time. Current micro-array technology has caused a significant increase in the number of genes whose expression can be measured simultaneously on a single array. However, the number of measurements that are taken in a time-course experiment has not increased in a similar fashion. As a result, typical gene expression data sets consist of relatively few time-points (generally less than 20) with respect to the number of genes (thousands). This so called dimensionality problem and the fact that measurements contain a substantial amount of measurement noise are two of the most fundamental problems in genetic network modeling. Generally, when learning parameters of genetic network models from ill-conditioned data (many genes, few time samples), the solutions become