基于Alpha-NMF的AD样本分类及特异性基因选择方法

由于基因表达谱数据的高噪声、高维性、高冗余以及数据分布不均匀等特点使得在分析过程中仍然有很多挑战性问题。基于该目的,将一种无监督学习方法--非负矩阵分解方法,应用到基因表达谱数据中,挖掘出与AD相关的信息基因。然而标准NMF算法其效率较低,并且在基因表达数据的应用有效性低。为了适应该领域的需求,采用了Alpha-NMF算法。该算法能够有效的克服标准NMF算法的缺陷,获得较好的实验结果。多次运行Alpha-NMF算法,选取分类准确率和稳定性最优的实验结果,对其集合基因设定一阈值,筛选出集合基因中大于该阈值的信息基因。最后通过基因功能分类以及生物功能结构图来验证所提炼出的特异性基因的有用性和可靠性。

Classification and specific genes selected by nonnegtive matrix factorization

Gene expression profiling contains large quantities of information about cellular process.While the high noise,high dimensionality and inhomogeneity of gene expression profiling made our research become more difficulty.The nonnegative matrix factorization method was applied to the gene expression profiles,and it could dig out specific genes related with AD.However,standard NMF algorithm can＇t be used well in gene expression data.The improved algorithm Alpha-NMF can effecticely overcome the defect of the standard NMF algorithm,and achiece good results.Choose the best trial and set a certain threshold with geneset to extract specific genes.Then genetic function classification and biological function chart used to verify the usefulness and reliability of specific genes.