Abstract
Foodomics studies deal with multifactorial problems which are analyzed with advanced multivariate analytical instruments and thus multivariate data handling methods are required to explore the data for latent information and data structures. The modern data analytical platforms generate vast amounts of data in a very short time and the analyst risk the challenge to be flooded with noninformative data. This is particularly problematic when the amount of variables with orders of magnitude exceeds the number of objects investigated. PCA is a very efficient method to investigate the information perturbation by different data cleaning and preprocessing methods. If the experimental design is aimed at differentiating between classes, additional classification methods have proven to be very powerful for metabolomics data. However, when implying a priori knowledge in the model step, the outcome must be rigorously validated before biological interpretation and presentation of the results. If data are multidimensional (e.g., many samples measured with GC--MS) and coeluting peaks are present, multiway models can be applied to perform unique peak deconvolution. Knowledge of chemometric methods combined with a biological understanding will provide the best platform for getting most from foodomics studies but also to avoid approaching data incorrectly and making nonvalidated interpretations.
| Original language | English |
|---|---|
| Title of host publication | Foodomics : advanced mass spectrometry in modern food science and nutrition |
| Editors | Alejandro Cifuentes |
| Number of pages | 32 |
| Publisher | Wiley |
| Publication date | 4 Mar 2013 |
| Pages | 507-538 |
| Chapter | 19 |
| ISBN (Print) | 9781118169452 |
| ISBN (Electronic) | 9781118537282 |
| DOIs | |
| Publication status | Published - 4 Mar 2013 |