用并行计算从基因表达数据构建大规模基因调控网络

为解决大规模基因调控网络构建算法精度不高、计算时间过长的问题，提出一种从基因表达数据分析出发，并行计算和阈值限定相结合的新算法来构建大规模基因调控网络。该算法中基因间交互强度值采用条件互信息值度量，并行计算采用GPU与CPU相结合的CUDA与OpenMP架构。综合数据集的运行结果证明该算法较新的构建算法（如贝叶斯模型算法和微分方程模型算法）相比，在构建大规模基因调控网络时有更高的运算精度和更短的运行时间。

Reconstruction of gene regulatory network using parallel computational algorithm with gene expression data

In order to resolve the drawbacks of traditional gene network reconstruction algorithm, such as less accuracy and long running time, this paper proposed a novel algorithm. This algorithm used parallel computational algorithm with parallel computing and threshold measurement. It measured the interaction between genes by conditional mutual information. It used the CUDA and OpenMP structure based on GPU and CPU in parallel computing algorithm. Running tests on synthetic datasets demonstrate that this algorithm can handle large datasets with more accuracy and less running test than traditional algorithm, such as Bayesian model algorithm and differential equation model algorithm.