Machine learning methods for metabotyping and personalized nutrition

 |  25.11.2022

INSA UB – Instituto de Investigación en Nutrición y Seguridad Alimentaria

Av. Prat de la Riba, 171

Dr. Cristina Andrés-Lacueva

Website of the group showing the members and previous projects. The mission of our group is to gain insights in the relationship between foods and health using a Nutrimetabolomics approach. In Nutrimetabolomics, there are four main goals: a) to discover and validate biomarkers of dietary intake to predict dietary exposures from objective measurements, b) to find biomarkers mediating the effects of diet in health to explain inter-individual variability in dietary responses, c) to dissect the metabolic pathways linking diet, gut microbiota metabolism of dietary compounds, host metabolism, and risk of disease, and d) to discover metabotypes and their biomarkers to design tailored-dietary interventions for a personalized health care.


The growth of metabolomics in recent years has had a strong impact on the field of nutrition, leading to the development of what is known as personalized nutrition, which, like precision medicine, attempts to determine the appropriate dietary intervention for everyone based on his/her metabolomic profile, also known as metabotype.

In this study, multivariate and Machine Learning methods for the construction of metabotypes will be reviewed and applied to simulated and real data arising from several European projects from our group and to a recently funded project named “Food4Brain” (PID2020-114921RB-C21). An important feature of metabotyping is that it is common to look for metabotypes characterizing a certain phenotype such as obesity or cardiovascular risk. We will aim to determine metabotypes associated with decreased brain health assessed through image, psychological and cognitive tests. Potential results will oscilate between the «discovery of new groups», which would be an unsupervised classification approach, and the finding of groups that explain certain characteristics of the individuals, which has a supervised analysis component.

Eventually if existing methods don’t prove to work well with this ambiguity, a new approach, like supervised clustering will be developed and made available to the community.

Job position description
The person to work on this project is expected to have a mixed background, such as a BSc in nutrition, chemistry or related life-sciences and an MSc in Biostatistics and/or Bioinformatics, which makes his/her proficient for working and analyzing quantitative data using statistics, machine learning and bioinformatics tools.