Among the 20 types of missense rare variants we detected in the exon 5 cytoplasmic loop region, 12 were observed only once (MAF=0.025%), with the most frequent missense rare variants having an MAF of 0.25%. Although the joint analysis of all rare variants provided sufficient power to detect differences between ND cases and controls, this approach assumes the same direction of effect (protective, deleterious or neutral) for all of the variants on ND risk, otherwise it could adversely affect power (45). To address this concern, we conducted a bioinformatic analysis to predict the effect of each nonsynonymous rare variant on protein function. This analysis suggested that eight of the 20 types of rare variants were predicted to be damaging to α4 subunit function. Restricting the comparisons to the aggregate effect of these eight variants confirmed the results of the analysis of all 20 nonsynonymous rare variants.
