The role of epigenetic modifications in obesity is poorly understood. Some associations between DNA methylation in cord blood and BMI in later childhood have been demonstrated, but the causality of these associations requires further investigation. Any causal relationships between epigenetic modifications and obesity could operate in either direction - DNA methylation may alter transcriptional regulation and hence affect obesity, or alternatively obesity may lead to changes in DNA methylation. We propose to study this by examining the associations between changes in DNA methylation and changes in adiposity in the Avon Longitudinal Study of Parents and Children (ALSPAC).

The associations between changes in adiposity and changes in epigenetic modifications will be assessed using measurements of adiposity and DNA methylation at birth and ages 7 and 15/17 years. Adiposity will be assessed by weight and height at all ages, and by DXA-assessed fat mass at 7 and 15/17 years. The associations will be investigated in multiple ways in order to compare methodological approaches to investigating changes in methylation. We will:

1. Identify the 100 methylation sites that change the most between each pair of sequential time points, and assess whether change in methylation at these sites is associated with change in adiposity

2. Perform principal components analysis (PCA) of changes in all 450k methylation sites to identify patterns of change in methylation, and relate the scores from these principal components to changes in adiposity

3. Perform partial least squares (PLS) analyses - a method that has similarities to PCA. PCA identifies combinations of variables that explain the greatest proportion of variance in those variables. In contrast, PLS identifies combinations of variables which explain the greatest proportion of variance in an outcome (in this case change in adiposity)

4. Perform latent class analysis of changes in methylation and relate these to change in adiposity

If any associations between changes in DNA methylation and changes in adiposity are identified, we will use a genetic risk score for obesity as an instrument for obesity changes, and perform a Mendelian Randomisation analysis to assess the direction of causality.