基于小波变换特征提取的代谢组低浓度标志物的筛选

Screening the Potential Important Biomarkers in Low Concentration Based on Wavelet Transform

目的利用连续小波变换对代谢组色谱数据进行变换分析,获得重要的低丰度肿瘤标志物。方法将代谢组一维色谱数据通过不同多尺度连续小波变换获得二维系数图像,再使用random forest(RF)算法对其图像进行模式识别和特征提取,筛选潜在的低浓度生物标志物。结果针对色谱数据经过连续小波变换处理后,RF建模判别效果较原始色谱数据有很大的提高。结论针对色谱数据进行小波变换后的特征提取结果,对质谱数据的某一段进行重点分析,能够筛选出重要的低丰度肿瘤标志物,具有重要研究价值。

Objective Metabolomics chromatographic mainly used to screening the potential important biomarkers in data analyzed by continuous wavelet transform （CWT） was low concentration. Methods The one-dimensional chromatographic data was transformed into two-dimensional image using CWT with different scales. Random Forest （RF） algorithm is applied to do the pattern recognition and feature selection, with the purpose of screening the potential important biomarkers in low concentration. Results According to chromatographic data after applying the continuous wavelet transform, the RF discrimination effect is improved greatly compared to the original data. Conclusion According to the feature extraction results of chromatographic data after being transformed by wavelet transform, screening the potential important biomarkers in low concentration could be obtained after selective analyzing mass spectral data with certain retention time, which had important research value.