Background:

In this study we will investigate how mothers' choose their partners. We will investigate this matching process using the exemplar of the relationship between parents' heights. Parents' heights are more associated than would be expected if individuals mated randomly within the population (Silventoinen et al. 2003). Height is a highly heritable phenotype. We do not know how much of the association of partners' heights is due to a genetic correlation, and how much is due to individuals from similar environmental backgrounds matching. For example, the associations of partners' heights could be because they chose to mate with individuals with similar backgrounds. We will extend the Mendelian randomisation framework to address these questions.

ALSPAC is an ideal dataset to investigate this hypothesis because it has detailed longitudinal data on both mother and fathers of the cohort children and genome-wide data on a large sample of the mothers.

Aims:

We will estimate the associations between mothers' height and height allele scores and their partners' height and observable characteristics using ALSPAC data.

Hypothesis: What are the associations between mothers' height and genotypic scores for scores for height and their partners' observed height and characteristics?

Exposure: The mothers' height at birth of their child and a weighted allele score of 697 recently reported height variants will be used as the exposure.

Outcomes: The outcomes will be the fathers' height, education, family income, paternal grand-parents education.

Confounding variables: first eight principal components of population stratification.

Outline:

People do not choose their partners at random: partners tend to be more similar than would be expected if partners were randomly chosen. We do not know if these associations are because people directly match on particular characteristics, height for example, or because partners match on a third confounding factor, such as their socio-economic background. We will investigate these associations in ALSPAC. There are now 697 genetic variants which are known to affect height (Wood et al. 2014). We will use these variants as instrumental variables for mothers' height in a Mendelian randomisation analysis.

In this study we will investigate the observed association between the mothers' height and their partners' phenotypes and compare the size of this association to the association implied by the genetic variants for height. If individuals actively match on height then we would expect similar associations between the observed height and the genetic variants for height. However, if the observed associations between mothers' height and their partners' phenotypes are the result of a third, confounding factor, then we would expect to find weaker evidence of associations between the genetic variants and the partners' characteristics. To maximise power we will construct weighted allele scores of the height genetic variants.