用FCM聚类和非参数回归方法推断基因调控网络

Inference of Gene Regulatory Network Using FCM Clustering and Nonparametric Regression Algorithm

采用FCM方法对基因表达的微阵列数据进行聚类分析,应用核回归和PP回归相结合的方法设计预报器,利用决定系数遴选父代基因集合.将这种新的组合设计方法的分析结果与相关文献的结果进行对比分析,得出了较好的推断结果;可以将这种组合方法运用于致病基因簇的搜寻.

FMC clustering is used for gene expression data, a combination of kernel regression and projection pursuit regression introduced into the design of predictors, and coefficient of determination used as a criterion in selecting the target gene's parent sets. Comparisons between the results based on the proposed method and those in the literature indicate the advantage of the proposed algorithms. The combination-design methods are recommended in finding and analyzing the carcinogenic genes clusters of human cancer.