After our announcement in August that we would be making GxE Regression available in SVS, we were pleased to receive feedback that this was exactly what our customers had been wanting. Being able to account for environmental effects or gene effects as interacting with the SNPs was essential to those researchers working with GWAS. Unfortunately, this did not help our customers who were also working with related samples and known environmental interactions that needed to be tested. So, we were asked to add GxE interaction terms into Mixed-Model Analysis.

We are pleased to announce that we will shortly have an updated Mixed-Model Analysis script available for you to download! (Once it is available, we will update this post with the download link.)

In this regression model, the SNPs, environment factors and interaction terms are treated as fixed effects and the kinship or relationship between samples is treated as a random effect.

For example, consider a model that adjusts for gender, BMI and the interaction of BMI against the genetic information.

We simply multiply the interaction term(s) against the SNP being included in the mixed-linear model and include this as a fixed effect in the MLM. Instead of using the regression beta for the individual marker tested, the beta for the interaction between the environmental factor and the SNP being tested is used.

Here is the result from our test dataset (click on the image to enlarge):

As you can see, by testing the interaction between BMI and the SNPs, we found a signal in a gene not detected with EMMAX, with BMI and Gender as fixed covariates.

From a genome wide view, you can see the signal profiles are more similar between the EMMAX with covariates and EMMAX without, and both are different when testing the interaction term. The largest signal, when not including the interaction term, is for a SNP that is not significant when including the interaction term (click on the image to enlarge).

Signal in MAPK14 and SLC26A8 not present when including the interaction term (click on the image to enlarge):

If you have questions or any other suggestions about where you would like to see future improvements in SVS, please send us an email to [email protected] and let us know.