基于ICA的AD样本的相关基因研究

为了克服每次运行独立成分分析算法（Independent Component Analysis，ICA）所得到不同的ICs对最后所选择的AD基因造成的不良影响，提出了一种应用ICA选择基因的改进方法。首先运行多次ICA算法，然后通过比较相关系数选出与AD相关的ICs，最后依据基因出现的频数选择相关基因。通过该方法寻找到39个在每个与AD相关的ICs中都过表达的基因，且其中有一个基因位于X染色体上，通过理论分析和实验结果表明该算法的确是一种可行性的方法。

Abstract:

 To overcome the adverse impact of getting different ICs from each running of ICA algorithm to the final choice of AD genes， the paper proposed a modified method of choosing genes by ICA. After several running of the ICA algorithm，according to the comparison of correlation coefficients the genes related with AD were selected based on frequency of the genes appearing. Using this method， 39 over-expression genes in each ICs related with AD were found， and one of them was located on chromosome X. Theoretical analysis and experimental results show that the algorithm is a feasible method.