厦门大学学报（自然科学版）2016，Vol.55Issue(4)：564-569,6.DOI:10.6043/j.issn.0438-0479.201601006
核磁共振波谱信号分析中的尺度缩放新算法
A New Variable-scaling Method for NMR Spectral Data Analyses
摘要
Abstract
Variable scaling,which aims to improve the performance of subsequently multivariate model,is a crucial procedure in metabolomic data analysis.However,most of the existing variable-scaling methods were carried out independently rather than taken into consideration the subsequently statistical model.Therefore,it is difficult for the optimal statistical model to be achieved.This pa-per proposes a model-adaptive-scaling method(MAS)for metabolic profiling analyses.The proposed method updates scaling coeffi-cients of variables by upgrading the preselected statistical model.A real-world nuclear-magnetic-resonance-based(NMR-based)meta-bolic profiling set was used to evaluate the proposed method and to compare with other three commonly-used scaling methods,i.e., unit variance(UV)scaling,Pareto scaling,and variable stability(VAST)scaling.Experimental results show that the proposed method outperforms other scaling methods in preserving the molecular information of spectra,enhancing important variables,and promoting the predicative ability of the preselected multivariate model.关键词
尺度缩放/磁共振/波谱信号/多元统计分析Key words
scaling/magnetic resonance/spectral signal/multivariate statistical analysis分类
化学化工引用本文复制引用
刘新卓,邓伶莉,王婉兰,沈桂平,许晶晶,董继扬..核磁共振波谱信号分析中的尺度缩放新算法[J].厦门大学学报（自然科学版）,2016,55(4):564-569,6.基金项目
国家自然科学基金(81371639,31372546) （81371639,31372546）
福建省自然科学基金(2015Y0032) （2015Y0032）
厦门大学校长基金(20720150018) （20720150018）
东华理工大学科研基金(DHBK2015308) （DHBK2015308）
