Location: Jean Mayer Human Nutrition Research Center On Aging
Project Number: 8050-10700-001-002-R
Project Type: Reimbursable Cooperative Agreement
Start Date: Apr 1, 2023
End Date: Dec 31, 2023
Objective:
ARS will conduct research for the following two specific aims:
1) Develop prediction models for plasma TG by applying AI and deep learning-based techniques to genomic data and dietary variables/measurements.
2) Predict individual responses in TG levels (as a proxy for cardiovascular risk) to a simulated increase in a Mediterranean diet score.
Approach:
To accomplish these objectives, ARS will use the subset of participants from the UK BioBank who have whole genome genotype (selected SNPs =1 million) and at least two 24-hr dietary assessment questionnaires (n=95,777). For objective 1, this subset will be split randomly into 50%, 20%, and 30% for training, validation, and testing. In step 1, using the Generalized Multifactor Dimensionality Reduction approach, we will conduct a genome-wide scan to identify gene, diet, gene-gene, and gene-diet (and lifestyles) interactions that contribute to TG levels in the training set. Selected variants and diet phenotypes will be combined as the best set of markers for deep learning. In step 2, using the convolutional neural networks learning approach, we will develop optimal models to predict the risk of hypertriglyceridemia based on identified factors from step 1. The prediction method will be cross-validated in the validation set. In step 3, the accuracy of the resulting prediction model will be estimated in the test dataset with the correlation between predicted and observed values.
For objective 2, we will use the best-developed model from above and simulate an increase in Mediterranean diet score (MDS) for 1000 subjects with high (i.e., n=500) or low TG (i.e., n=500) and low MDS. For validation, the predicted TG levels of the tested subjects (n=100) will be compared with those measured in 100 subjects consuming a high MDS diet and matched for age, sex, anthropometric characteristics and selected genotypes, biomarkers.
