Objective: Understanding genetic dissection of vascular dementia (VD) has been quite limited by single gene analyses. We simultaneously analyzed multiple genes in their association with the susceptibility to VD to examine the genetic susceptibility controlled by their additive and epistatic effects.
Methods: Two hundred and seven VD patients and age/sex matched 207 controls were genotyped for 71 candidate genes that were expressed in the human brain. Multifactor dimensionality reduction was used to select the best genetic variants. Then, entropy decomposition was applied to assess the selected variants' individual and interactive genetic contributions to the susceptibility to VD.
Results: The best two-locus, three-locus, four-locus candidate models resulted by the multifactor dimensionality reduction analysis showed the significance in both training and testing accuracy estimates. The largest estimate (10 out of 10) of cross-validation consistency was obtained for the two-locus model that included family with sequence similarity 134 member B (FAM134B, rs10041159) and tumor necrosis factor receptor superfamily, member 19 (TNFRSF19, rs9317882). The subsequent entropy decomposition analysis showed that individual genetic variants eliminated the uncertainty of 3.02 and 6.25% in case-control status, and the entropy largely decreased with the presence of their interaction. Especially, the interaction among genetic variants included in the final model rendered the additional decrease (32.95%) of entropy.
Conclusion: The FAM134B and TNFRSF19 showed a dramatically strong synergistic epistasis in explaining the genetic dissection of the susceptibility to complex VD.