多组学数据分析中关联网络融合ANF算法及应用研究

Application of Affinity Network Fusion Algorithm in Multi-omics Data

目的引入关联网络融合(affinity network fusion,ANF)方法对多组学数据进行整合分析,并应用于肿瘤分子分型研究。方法模拟产生两组学数据,改变总体差异大小等情况比较多种多组学整合方法的效果。实例分析中目标人群选择TCGA数据库中对铂类药物敏感并拥有mRNA和甲基化两个组学的卵巢癌患者,目标基因是TCGA数据库和ImmPort数据库中共有基因,目标甲基化位点是目标基因对应的所有甲基化位点。使用ANF、SNF、K-means、系统聚类和iCluster五种方法比较聚类效果。结果模拟实验提示存在总体差异的两亚型间差异较小时ANF方法的效果明显优于其他方法。实例分析结果表明,通过ANF方法进行多组学数据整合得到的分子分型较单组学得到的分子分型有更好的生物学意义且多组学聚类效果优于其他方法。结论ANF方法可以应用于多组学数据整合分析,具有实际应用意义。

Objective Affinity network fusion(ANF)method was applied to integrate multi-omics data to explore its application in research of diseases subtypes.Methods ANF was compared with SNF,iCluster,K-means and hierarchical cluster by analysis of simulation experiments and actual multi-omics data in Ovarian cancer chemotherapy sensitive population.Results In simulation experiment,by changing the disparity and the percentage of difference variables,it was determined that when the overall disparity of subgroups was small,the clustering effect of ANF method was better than other methods.Actual analysis shows that the molecular typing obtained by ANF method is more biologically significant than the molecular analysis obtained by other methods and its single analysis.Conclusion ANF can be applied to multi-omics data integration analysis and has practical significance.