University of Hertfordshire

From the same journal

By the same authors

  • Jacqueline Hallmann
  • Silvia Kolossa
  • Kurt Gedrich
  • Carlos Celis-Morales
  • Hannah Forster
  • Clare B. O'Donovan
  • Clara Woolhead
  • Anna L. Macready
  • Rosalind Fallaize
  • Cyril F. M. Marsaux
  • Christina-Paulina Lambrinou
  • Christina Mavrogianni
  • George Moschonis
  • Santiago Navas-Carretero
  • Rodrigo San-Cristobal
  • Magdalena Godlewska
  • Agnieszka Surwiłło
  • John C. Mathers
  • Eileen R. Gibney
  • Lorraine Brennan
  • Marianne C. Walsh
  • Julie A. Lovegrove
  • Wim H. M. Saris
  • Yannis Manios
  • Jose Alfredo Martinez
  • Iwona Traczyk
  • Michael J. Gibney
  • Hannelore Daniel
  • Food4Me Study
View graph of relations
Original languageEnglish
Pages (from-to)2565-2573
JournalMolecular Nutrition and Food Research
Journal publication dateDec 2015
Early online date30 Sep 2015
Publication statusPublished - Dec 2015


SCOPE: A high intake of n-3 PUFA provides health benefits via changes in the n-6/n-3 ratio in blood. In addition to such dietary PUFAs, variants in the fatty acid desaturase 1 (FADS1) gene are also associated with altered PUFA profiles.METHODS AND RESULTS: We used mathematical modeling to predict levels of PUFA in whole blood, based on multiple hypothesis testing and bootstrapped LASSO selected food items, anthropometric and lifestyle factors, and the rs174546 genotypes in FADS1 from 1607 participants (Food4Me Study). The models were developed using data from the first reported time point (training set) and their predictive power was evaluated using data from the last reported time point (test set). Among other food items, fish, pizza, chicken, and cereals were identified as being associated with the PUFA profiles. Using these food items and the rs174546 genotypes as predictors, models explained 26-43% of the variability in PUFA concentrations in the training set and 22-33% in the test set.CONCLUSION: Selecting food items using multiple hypothesis testing is a valuable contribution to determine predictors, as our models' predictive power is higher compared to analogue studies. As unique feature, we additionally confirmed our models' power based on a test set.

ID: 9476165