Bubble plot of p-values indicating significant orthogonal GxE interactions. There were 16 significant orthogonal GxE interactions when testing the genetic effects of 12 diabetes-related metabolic traits modulated by eight covariates. Likelihood ratio tests were used to compare the full model with the RxE only model for each of six independent data sets, and p-values were meta-analyzed using the Fisher method. The size of dot reflects its significance; the bigger the more significant. For example, the BMI-ALC pair is most significant (p-value = 1.86E-10), indicating that ALC significantly modulates genetic effects of BMI phenotypes. WHR, waist-hip ratio; TC, Total cholesterol; systolic, systolic blood pressure; sodium, sodium in urine; potassium, potassium in urine; LDL, low density lipoprotein cholesterol; HDL, high density lipoprotein cholesterol; HbA1c, haemoglobinA1C; CRP, C reactive protein; BMI, body mass index; Age, Age at recruitment; ALC, alcohol intake frequency; smk, smoking status; PA:summed, a summed physical activity; PA:walk, physical activity walking; PA:moderate, physical activity moderate intensity; PA:vigorous, physical activity vigorous intensity; Diet, healthy dietary scores.