Lecture
Addressing data dimensionality and structure in Metabolomics: issues and solutions
- at -
- ICM Saal 2
- Type: Lecture
Lecture description
While early metabolomic studies relied mainly on NMR, hyphenated methods involving separation techniques and mass spectrometry (GC-MS, LC-MS, CE-MS) have now been demonstrated to be powerful and complementary analytical techniques. Due to the ever-increasing number of signals that can be measured within a single run by modern platform, datasets not only become gradually larger in size, but also more intricate in their structures. Challenges related to making use of this wealth of data include extracting relevant elements possibly spread across different tables, reducing dimensionality, summarizing dynamic information in a comprehensible way and displaying it for interpretation purposes. Metabolomics constitutes a representative example of fast moving research fields taking advantage of recent technological advances to provide extensive sample monitoring. Due to the wide chemical diversity of metabolites, several analytical setups are often required to provide a broad coverage of complex samples or situations. Classical hypothesis-driven approaches are no longer applicable to such data collection and dedicated data analysis strategies have to be used. Multivariate methods based on the computation of latent variables or components, such as principal component analysis (PCA) and partial least squares (PLS) regression constitute potent tools to provide compact data representations and diagnostic tools for the detection of relevant variables. Nevertheless, most of these approaches lack the ability to fully exploit more complex data structures such as (1) multifactorial, (2) longitudinal and (3) multiblock setups. As presented in this lecture through examples, dedicated approaches, able to cope with the inherent properties of these MS metabolomic datasets are therefore mandatory for harnessing their complexity and provide relevant information. In that perspective, chemometrics has a central role to play in the choice of the appropriate methodology.
