摘 要: 针对基于隐马尔科夫模型的非齐次动态贝叶斯网络(HMM-DBN)中因基因调控作用强度过度灵活而造成的网络重构精度降低的问题,提出了用参数边缘耦合方式改进HMM-DBN的方法。首先对基因调控数据进行时间分段;其次利用边缘耦合算法判断当前分段是否应该与前一段的信息交互;再次根据是否进行信息交互判断每个分段的回归参数是否耦合,结合回归参数和时间分段推断基因调控关系;最后重复上述过程直到MCMC(马尔科夫链蒙特卡洛算法)迭代完成,输出网络结构。改进后的HMM-DBN在酵母数据集与合成RAF(RAF原癌基因丝氨酸/苏氨酸-蛋白激酶)数据集上的实验结果显示,其网络重构精度达到了0.76以上,证明了该方法的有效性。 |
关键词: 非齐次贝叶斯网络;MCMC;边缘耦合;基因调控网络 |
中图分类号: TP181
文献标识码: A
|
|
Modeling of Gene Regulatory Networks with Edge-wise Coupling Parameters |
MA Mengyu1, HU Chunling2
|
( 1. School of Electronic and Information Engineering, Anhui Jianzhu University, Hefei 230601, China; 2. Department of Artificial Intelligence and Big Data, Hefei University, Hefei 230601, China)
1227554288@qq.com; huchunling@hfuu.edu.cn
|
Abstract: In order to solve the problem that the network reconstruction accuracy of Hidden Markov Model Nonhomogeneous Dynamic Bayesian Network (HMM-DBN) is reduced due to the excessive flexibility of gene regulation, this paper proposes edge-wise coupling parameters to improve HMM-DBN. First, time segmentation is carried out for gene regulation data. Then, edge-wise coupling algorithm is used to judge whether the current segment should interact with the previous one. Next, whether the regression parameters of each segment are coupled is judged according to whether information interaction is conducted. The gene regulatory relationship is inferred by combining the regression parameters and time segments. Finally, the above process is repeated until MCMC (Markov Chain Monte Carlo) algorithm iteration is completed and the network structure is output. Experimental results on yeast datasets and the synthetic RAF (Raf-1 protooncogene, serine/threonine kinase) datasets show that the network reconstruction accuracy reaches more than 0.76, which proves the effectiveness of the proposed method. |
Keywords: non-homogeneous Bayesian network; Markov Chain Monte Carlo; edge-wise coupling; generegulatory networks |