A graphical tool for exploring SNP-by-environment interaction in case-parent trios

We propose a data-smoothing method for exploring statistical interaction between a single nucleotide polymorphism (SNP) and non-genetic risk factors in case-parent trios. Our smoother can be used as a diagnostic tool for checking for the presence or the form of the interaction. The smoother arises from a case-only analysis conditional on parental genotypes. Conditioning on parental genotypes helps to protect against the false impression of interaction that traditional case-only analyses can give when genotypes and non-genetic risk factors are not independent in the population. We discuss the theoretical motivation for the smoother, and illustrate its use with simulated data. We show that the effect of the SNP would have been missed if the interaction suggested by the smoother had not been modelled.