AIMTo search for and application of new method for data processing in metabonomic studies.

METHODSThe paper proposed that in the processing of metabonomic data, robust PCA method can be used to diagnose outliers; and unstable variables judged by comparison between difference within class and difference among classes should be excluded before data analysis; moreover, the data should be properly scaled before further processing. The proposed methods were used to preprocess metabolomic data of four genotypes of the Arabidopsis thaliana plants.

RESULTS AND CONCLUSIONThe outcome demonstrated that the application of these methods can obviously improve clustering and biomarker identifying results.