Statistical recoupling of variables for the identification of candidate biomarkers and perturbed metabolic networks. Application to whole organism NMR - CNRS - Centre national de la recherche scientifique Accéder directement au contenu
Communication Dans Un Congrès Année : 2011

Statistical recoupling of variables for the identification of candidate biomarkers and perturbed metabolic networks. Application to whole organism NMR

Laetitia Shintu
Martial Piotto
  • Fonction : Auteur
Lyndon Emsley
  • Fonction : Auteur
  • PersonId : 925398

Résumé

We have developed an automated variable-sized bucketing procedure based on statistical relationships between neighboring variables, obtained by traditional 0.001ppm wide bucketing, to recouple into a cluster the points belonging to metabolic NMR signals. This Statistical Recoupling of Variables (SRV) procedure acts as an efficient noise-removing filter and identifies new variables, which have physical, chemical or biological meaning. This allows a large reduction of the number of variables and thus counters the "curse of dimensionality" observed in the "-omics" sciences. SRV can be used in combination with multiple hypothesis testing corrections to evaluate the statistical significance of variations observed on the metabolic phenotypes discriminating the populations under study. It is thus possible after SRV to extract candidate biomarkers whose single variations are sufficient to sustain the discrimination. In a more comprehensive approach, SRV can be associated with Statistical Total Correlation Spectroscopy (STOCSY) to yield a recoupled form (R-STOCSY) and obtain a clear 2D pseudo-spectrum displaying spin and metabolic correlations. These correlations can then be identified as metabolic connectivities. By considering the shortest path length between the metabolites involved in these correlations, it is possible to visualize the perturbed metabolic network associated with a given pathophysiological condition, and to assess the robustness of the overall analysis. These approaches together provide a powerful way to secure the interpretation of metabonomics data. In the case of perturbation of minor intensity, we added an orthogonal filter to focus on the effect under scrutiny, which would otherwise be diluted in the complexity of the data set. We illustrate this ability in whole organism NMR based studies. HRMAS NMR analysis of entire nematodes and cell pellets can lead to the functional genomics of C. elegans and a better understanding of the tumorogenesis pathophysiology in the MEN1 cancer syndrome.
Fichier non déposé

Dates et versions

hal-00980051 , version 1 (17-04-2014)

Identifiants

  • HAL Id : hal-00980051 , version 1

Citer

Benjamin J. Blaise, Cécile Vercherat, V. Navratil, Annie Lacheretz-Bernigaud, Laetitia Shintu, et al.. Statistical recoupling of variables for the identification of candidate biomarkers and perturbed metabolic networks. Application to whole organism NMR. 7th International Conference of the Metabolomics Society, Jun 2011, Cairns, Australia. ⟨hal-00980051⟩
90 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More