UCP-4 (aka SLC25A27) does give one the willies.
Variants near rs9472817
"... of mitochondrial uncoupling protein 4 in the development of vascular demyelinization of the white matter of the brain, referred to as leukoaraiosis (LA). The mUCPs are presumed to be of great importance in the regulation of the mitochondrial membrane potential (MMP) and the cellular energy metabolism."https://www.ncbi.nlm.nih.gov/pubmed/20545631
Same variant near rs9472817 for Multiple sclerosis:
"...of mitochondrial uncoupling protein 4 (mUCP4) can give rise to the development of MS, since mUCP4 is presumed to be of great importance in the regulation of the MMP and cellular energy metabolism. The clinical and genetic data on 120 relapsing-remitting MS patients and 250 neuroimaging alteration-free subjects were analyzed. The rs10807344 CC genotype proved to exert a protective effect against the occurrence of MS (neuroimaging alteration-free controls, 58%; MS group, 33%; P < 0.0000089; OR, 0.32; 95% CI: 0.2-0.56, P < 0.005). https://www.ncbi.nlm.nih.gov/pubmed/19536655
"...These neuronal UCPs are expressed in various brain tissues and may exert a neuroprotective effect against increased oxidative stress. We found modest associations between schizophrenia and the four tag SNPs, rs660339 (odds ratio (OR) = 1.330; P = 0.0043) and rs649446 (OR = 0.739; P = 0.0069) in UCP2, and rs10807344 (OR = 0.622; P = 0.0029) and rs2270450 (OR = 0.704; P = 0.0043) in UCP4, all of which were statistically significant even after correcting for multiple comparisons."https://www.ncbi.nlm.nih.gov/pubmed/17066476
None of these were recent large scale GWAS, though it does show that the AD result would not be a throw of the dart.
Modern GWAS are agnostic about what the genes involved might be, so they need huge samples to overcome the multiple