小脑基因表达数据的模糊多尺度聚类分析

Fuzzy Multiscale Clustering Analysis of Cerebella Gene Expression Data

目的为了更好地建立符合生物学意义的基因归类,为一些未知基因的功能提出解释提供参考。方法首先对小脑组织随机抽取100组预处理后的基因表达数据,对每个由7个时间点所成的基因表达信号做多尺度分析,其次在各个尺度下运用改进的FCM算法设计了一个归类阀值,并利用模糊聚类Xie-Beni指数得到了最优聚类数并实现各个尺度下小脑组织基因的聚类,并把每一层对应的聚类结果输出到文本文件,最后找出各层聚类结果完全一致的基因进行归类并进行生物学解释。结果得到的小脑组织基因最优聚类数为3类,通过分类结果对照发现,各类中的大多数基因生物学意义接近。结论运用多尺度分析并结合FCM算法应用于基因聚类是有效的,结果具有一定生物学意义,能对生物学基因聚类及基因功能解释具有一定指导作用。

Objective In order to establish genetic classification in according with biological significance,and give reference to interpret some unknown gene＇s function.Methods First of all we did a multiscale analysis toward cerebella gene expression signal,subsequently we used an improved FCM clustering algorithm and design a classification threshold in various scales,then we used fuzzy clustering Xie-Beni index to achieve the optimal number of clusters and accomplish the clustering of cerebella genes of various scales,and each of class corresponding gene label was exported to txt file,finally we found out the genes which were classified exactly the same in every layer and were conducted their biological explanations.Results The optimal number of clusters of cerebella genes was 3 categories,and we according to the classification results comparison,we found that majority of genes in various types had close biological significance.Conclusion It is effective to gene clustering where use multiscale analysis combine FCM algorithm,the result has certain biological significance,it can give guidance in biological gene clustering and explaining gene function.