ISOTOPE ASSISTED METABOLOMICS OF ARABIDOPSIS

点击翻译按钮获取中文摘要

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. Nuclear Magnetic Resonance (NMR) has emerged as an important analytical tool in the field of metabolomics. Virtually all of the current methods for statistical analysis of complex NMR spectra rely on a smoothing procedure (binning) to condense spectral complexity and correct for variability in peak location. In this paper, we show that the traditional binning methods are a major source of experimental variability. In response to this, we introduce Automated Filtering of NMR Spectra (AFNS). AFNS is a processing program that uses a rolling binning algorithm, multiple binwidths, and t-statistic based filtering as a means of identifying significant features in complex spectra. As an initial application of this program, we analyzed proton spectra of liver extracts from two strains of mice: the C57BL6/J (B6) and BTBR, in both genetically obese (ob/ob) and lean (+/+) conditions. Although lean mice from both strains are healthy, the BTBR-ob/ob mice develop severe diabetes. In contrast, B6-ob/ob mice remain resistant to diabetes despite obesity. When traditional binning methods were used, proton spectra from B6(+/+ and ob/ob) and BTBR(+/+ and ob/ob) could not be linearly separated by Principal Components Analysis (PCA) and the distribution of individuals was strongly influenced by the selection of binwidth. In contrast, spectra processed with AFNS revealed that B6(+/+ and ob/ob) and BTBR(+/+ and ob/ob) mice have distinct molecular profiles that can be linearly separated from one another along a single principal component. We attribute the success of AFNS to reduced binwidth related variability and a more robust selection of spectral features.