Abstract: The acquisition process of gene expression dataset mixed with noise easily,the noise can interfere with data expression correctly which will affect their further analysis and research.The Empirical Mode Decomposition(EMD) denoising of using median-calculation to estimate noise standard deviations that exists some drawbacks and affects the denoising effectiveness.Under the EMD,the fractional Gaussian noise(fGn) model can provide more accurate estimation way of noise standard deviations,denoising on this model can reduce white and colored noise,which will enhance the denoising effectiveness.So a denoising scheme based on fGn is proposed on the basis of median-calculation EMD denoising and done the denoising analysis on colon cancer gene expression dataset.Experimental results show that,the values of signal-to-noise,noise rejection ratio,t-test,etc.in advanced method have a certain superiority compared with median-calculation EMD denoising,which can be used as an reference means of denoising to the gene expression dataset.

Key words: gene expression dataset, Empirical Mode Decomposition(EMD) denoising, noise standard deviations, fractional Gaussian noise(fGn), colon cancer

摘要： 基因表达数据集获取过程中容易掺杂噪声成分,噪声会干扰数据的正确表达从而影响其后期的分析与研究。基于中值计算法估计噪声标准差的经验模态分解(EMD)去噪存在一定的不足,从而影响去噪效果。分数阶高斯噪声(fGn)模型可提供EMD下较为准确的噪声标准差估计方法,在该模型下去噪可减少白色及有色噪声,进而增强去噪效果。因此在中值计算EMD去噪基础上,提出一种基于fGn模型的去噪方案,并对结肠癌基因表达数据集做去噪分析。实验结果表明,相比中值计算EMD去噪方法,改进方法的信噪比、噪声抑制比、t检验等值具有一定的优势,可作为基因表达数据集去噪的一种参考方案。

关键词: 基因表达数据集, 经验模态分解去噪, 噪声标准差, 分数阶高斯噪声, 结肠癌

CLC Number: 

• TP391